Analysis of Hybrid Management Strategies Master of Science Thesis ERIK KARLSSON MARTIN WILNER Department of Signals and Systems Division of Automatic Control, Automation and Mechatronics CHALMERS UNIVERSITY OF TECHNOLOGY Göteborg, Sweden, 2008 Report No. EX085/2008 Abstract In the automotive industry nowadays a lot of focus is set on environmental friendly ve- hicles, with the primary aim of reducing emissions and pollutants. The most efficient way of doing this is to increase the fuel economy, i.e. to make the vehicle drive longer on the same amount of fuel. Hybrid electric vehicles have been shown to be a popular way of achieving this, without compromising the performance and reliability present in the conventional vehicles. The idea behind this kind of vehicle is to combine two different power sources, typically a conventional fuel driven internal combustion engine and a battery driven electric machine, and use both of them to provide tractive forces to the vehicle. Adding this second power source results in a higher complexity of the powertrain, which requires a sophisticated control system to manage the power flow in the system, henceforth denoted the hybrid management strategy. This master thesis is conducted at ZF Friedrichshafen AG, which currently investigate two different hybrid management strategies for hybrid electric vehicles in a simulation environment. An analysis of these two strategies is done for different vehicles setups, where fuel consumption will be regarded as the primary factor. Before the analysis is conducted, the simulation environment is extended with some different implementations, preventing errors to occur. More parts of the strategies and some gear shift strategies are implemented as well. When the performance of the hybrid management strategies are studied, it is seen that both strategies perform well, giving similar results. Though they differ in the behavior, they manage to get similar results. An important factor to consider is the parameterization of the strategies, as this influences the behavior and results significantly. The biggest difference between the two strategies has to do with how they manage the energy flowing in the battery. As a last step, the two strategies are tested with a second gear shift logic, and it is seen that both of the strategies are able to perform well. i ii Acknowledgements First of all we would like to thank our two supervisors at ZF, Jochen Köhler and Notker Amann, for their help and interesting discussions throughout the thesis. A special thanks to Mohsen Elsayed, Johannes Kaltenbach, Maged Khalil and Christian Mittelberger for their help. We also thank our examiner at Chalmers, Professor Bo Egardt. We want to thank all the friends in Friedrichshafen who supported and helped through- out the thesis, especially during the spare time. Also friends and family at home who have supported. From Erik: a very special thank you to Ayse, who supported me a lot with everything. iii iv Abbreviations ECMS Equivalent Consumption Minimization Strategy EM Electric Machine HEV Hybrid Electric Vehicle HMS Hybrid Management Strategy ICE Internal Combustion Engine LUT Look-Up Table OEL Optimal Efficiency Line SML Strategic Mode Logic SoC State of Charge TD Torque Distribution v vi Contents Abstract i Acknowledgements iii Abbreviations v List of Figures xii List of Tables xiii 1 Introduction 1 1.1 General Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Problem Description and Objectives . . . . . . . . . . . . . . . . . . . . . 2 1.3 Outline of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 2 Hybrid Electric Vehicles 5 2.1 Hybrid Electric Vehicles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.2 Series Hybrid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.3 Parallel hybrid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.4 Components in the Parallel Hybrid . . . . . . . . . . . . . . . . . . . . . . 7 2.4.1 Battery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.4.2 Clutch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.4.3 Fuel tank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.4.4 Internal Combustion Engine . . . . . . . . . . . . . . . . . . . . . . 7 2.4.5 Electric Machine . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.4.6 Transmission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.5 Impact of Hybridization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 3 Hybrid Management Strategies 11 3.1 Hybrid Management Strategies . . . . . . . . . . . . . . . . . . . . . . . . 11 3.1.1 Strategic Mode Logic . . . . . . . . . . . . . . . . . . . . . . . . . 12 3.1.2 Torque Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . 13 3.1.3 State of Charge Controller . . . . . . . . . . . . . . . . . . . . . . . 13 3.1.4 Brake Management . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 3.2 Rule-Based Hybrid Management Strategy . . . . . . . . . . . . . . . . . . 13 3.2.1 Strategic Mode Logic . . . . . . . . . . . . . . . . . . . . . . . . . 13 3.2.2 Torque Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.2.3 State of Charge Controller . . . . . . . . . . . . . . . . . . . . . . . 18 3.3 Table-Based Hybrid Management Strategy . . . . . . . . . . . . . . . . . . 19 3.3.1 State of Charge controller . . . . . . . . . . . . . . . . . . . . . . . 19 vii viii CONTENTS 3.3.2 Torque Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . 20 3.3.3 Strategic Mode Logic . . . . . . . . . . . . . . . . . . . . . . . . . 21 3.4 A Global Optimal Control Strategy: Dynamic Programming . . . . . . . . 22 3.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 4 The Simulation Environment 25 4.1 Simulation Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 4.1.1 Implementations and Improvements . . . . . . . . . . . . . . . . . 26 4.2 Vehicle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 4.3 Driving Cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 4.4 Level of the State of Charge . . . . . . . . . . . . . . . . . . . . . . . . . . 28 4.5 Driver Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 4.6 Driver Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 4.7 Benchmark Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 4.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 5 Simulation with Rectified Layout 35 5.1 Improving the Original Rule-Based Strategy . . . . . . . . . . . . . . . . . 35 5.1.1 Calculation Error when ICE Torque Decrease and EM Boost Oc- curred Simultaneously . . . . . . . . . . . . . . . . . . . . . . . . . 35 5.1.2 State of Charge Controllers Synchronization Issues . . . . . . . . . 36 5.1.3 Tuning the Strategic Mode Logic . . . . . . . . . . . . . . . . . . . 37 5.1.4 Tuning the Torque Distribution . . . . . . . . . . . . . . . . . . . . 37 5.2 Comparison Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 5.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 6 Simulation for Various Design Layouts 45 6.1 Battery Size Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 6.1.1 Comparison Results . . . . . . . . . . . . . . . . . . . . . . . . . . 45 6.2 Battery Size and Vehicle Weight Reduction . . . . . . . . . . . . . . . . . 48 6.2.1 Comparison Results . . . . . . . . . . . . . . . . . . . . . . . . . . 48 6.3 No Electrical Driving . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 6.3.1 Comparison Results . . . . . . . . . . . . . . . . . . . . . . . . . . 51 6.4 Other Ideas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 6.4.1 Hybrid Drive When Braking . . . . . . . . . . . . . . . . . . . . . 53 6.4.2 Gradient Based Method for Deciding the Torque Distribution Pa- rameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 6.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 7 Implementation of a Second Gear Shift Strategy 59 7.1 Gear Shift Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 7.1.1 Speed-Based Gear Shift Strategy . . . . . . . . . . . . . . . . . . . 59 7.1.2 Optimal Efficiency Line-Based Gear Shift Strategy . . . . . . . . . 60 7.1.3 Tuning of the Gear Shift Strategies . . . . . . . . . . . . . . . . . . 61 7.2 Comparison Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 7.3 General Analysis of the Gear Shift Strategy . . . . . . . . . . . . . . . . . 62 7.4 Table-Based Strategy With Adapted Gear Shift Strategy . . . . . . . . . . 66 7.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 CONTENTS ix 8 Conclusions and Recommendations 69 8.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 8.2 Recommendations for Future Work . . . . . . . . . . . . . . . . . . . . . . 70 Bibliography 73 Appendix i A State of Charge Level i x CONTENTS List of Figures 1.1 Toyota Prius model NHW20, a mass produced HEV released in 2004 by the Toyota Motor Company. . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2.1 A schematic of the series hybrid. . . . . . . . . . . . . . . . . . . . . . . . 6 2.2 A schematic of the parallel hybrid. . . . . . . . . . . . . . . . . . . . . . . 6 2.3 Example of an efficiency map (left) and fuel consumption (right) of a diesel engine. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2.4 Example of an efficiency map for an EM, for motor mode (positive torques) and generator mode (negative torques). The arrows indicate increasing efficiency, where the highest efficiency point is about 90%. . . . . . . . . . 9 3.1 The structure of the HMS as implemented in the simulation environment. 11 3.2 Simplified version of the stateflow diagram, representing the SML for the rule-based strategy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 3.3 Four arbitrary working points of the ICE are selected, to show that they can be moved to an area with a higher efficiency. . . . . . . . . . . . . . . 15 3.4 The curves representing the features in the rule-based TD. The ICE torque increase and the ICE torque decrease curves are set to arbitrary values. . 16 3.5 SoC controller for the table-based strategy. . . . . . . . . . . . . . . . . . 19 3.6 The two tables used in the TD (left) and the SML (right) for the table- based strategy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 3.7 Surface plot for the SML and the TD of the table-based strategy. . . . . . 22 4.1 The structure of the simulation environment at ZF, together with some important inputs and outputs. . . . . . . . . . . . . . . . . . . . . . . . . 26 4.2 A dead-lock situation which could occur in the system, if the mode of the vehicle was changed during a gear shift phase. . . . . . . . . . . . . . . . . 26 4.3 Small delivery truck used for the simulations [15]. . . . . . . . . . . . . . . 27 4.4 Three cycles that are used in the simulations with the small delivery truck. 28 4.5 Algorithm to balance the SoC level, presented as a flowchart. . . . . . . . 30 4.6 Basic parts of the artificial driver used in the simulations. . . . . . . . . . 32 5.1 The incorrect version of the EM Boost and the ICE Torque Decrease interaction (left), and the correct version (right). . . . . . . . . . . . . . . 36 5.2 Operating points of the EM, shown on the ICE efficiency map for the Japan5 cycle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 5.3 Simulation results for the rule-based strategy (solid blue) and the table- based strategy (dash-dot red), showing the SoC, ICE torque, EM torque, vehicle mode and cycle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 xi xii LIST OF FIGURES 5.4 The energy flow in the battery for the rule-based strategy (left) and the table-based strategy (right). Green boxes indicate recharging and blue boxes discharging. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 5.5 Operating points of the ICE, where the height of the boxes represents fuel consumption. The rule-based strategy (left) and the table-based strategy (right). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 6.1 Simulation results for the rule-based strategy (solid blue) and the table- based strategy (dashed red), showing the SoC, vehicle mode and cycle. . . 46 6.2 The SoC for the two HMSs plotted together with the desired SoC for the table-based strategy, the vehicle modes and the cycle. . . . . . . . . . . . 47 6.3 Operating points of the ICE, where the height of the boxes represents fuel consump- tion. The table-based strategy with vehicle weight 7100 kg (left) and vehicle weight 2500 kg (right). . . . . . . . . . . . . . . . . . . . 49 6.4 The SoC for the two HMSs plotted together with the desired SoC for the table-based strategy, the vehicle modes and the cycle. . . . . . . . . . . . 50 6.5 The SoC and the fuel consumption from the Japan5 route. The base value for the SoC controller in the table-based strategy is at 52%. . . . . . . . . 52 6.6 The SoC and the fuel consumption from the Japan5 route. The rule-based strategy uses the gradient based method for its TD, and the base value for the SoC controller in the table-based strategy is tuned to 45%. . . . . 53 6.7 Simulation data from a braking sequence, for Electric Drive (solid blue) and Hybrid Drive (dashed red). . . . . . . . . . . . . . . . . . . . . . . . . 54 7.1 Gear-shifting map of the speed based strategy, where dotted lines repre- sents a downshift and solid lines represents an upshift. . . . . . . . . . . . 60 7.2 The implementation of the OEL-based shift strategy. . . . . . . . . . . . . 61 7.3 Gear operating points of the rule-based shift strategy (top) and the OEL- based shift strategy (bottom), for positive powers in hybrid driving. . . . 64 7.4 The fuel consumption of the ICE together with three typical power lines. 64 7.5 Gear operating points of the rule-based shift strategy (top) and the OEL- based shift strategy (bottom), for negative powers in electric driving. . . . 65 7.6 The EM efficiency map together with three typical power lines and two arrows used in an example. . . . . . . . . . . . . . . . . . . . . . . . . . . 66 A.1 Four plots of the algorithm which gives the best balanced SoC and the fuel consumption. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii List of Tables 3.1 Overview of the HMSs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 5.1 Vehicle specifications for the first simulations. . . . . . . . . . . . . . . . . 35 5.2 Fuel consumption results as a function of the degree of ICE torque increase for two cycles, expressed relative to the base value of 100% set for zero ICE torque increase. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 5.3 Fuel consumption results for the two HMSs for three different cycles, ex- pressed relative to the base value of 100% set for the conventional vehicle. 42 6.1 Vechicle specification for the second simulations, with reduced battery size. 45 6.2 Fuel consumption results for the two HMSs for three different routes, expressed relative to the base value of 100% set for the conventional vehicle. 46 6.3 Vehicle specification for the third simulations, with reduced battery size and vehicle weight. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 6.4 Fuel consumption results for the two HMSs for three different routes, expressed relative to the base value of 100% set for the conventional vehicle. 49 6.5 Vehicle specification for the fourth simulations, where no electric driving is possible. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 6.6 Fuel consumption results for the two HMSs for three different routes, expressed relative to the base value of 100% set for the conventional vehicle. 51 6.7 Fuel consumption results for the rule-based strategy based on the script, expressed relative to the base value of 100% set for the conventional vehicle. 56 7.1 Vehicle specifications for the simulations with the second gear shift strategy. 62 7.2 Fuel consumption results for the two HMSs for three different routes, expressed relative to the base value of 100% set for the conventional vehicle. 63 7.3 Fuel consumption results for the two HMSs for three different routes, expressed relative to the base value of 100% set for the conventional vehicle with the OEL-based gear shift strategy. . . . . . . . . . . . . . . . . . . . 63 7.4 Fuel consumption results for the two HMSs for three different routes, expressed relative to the base value of 100% set for the conventional vehicle with the OEL-based gear shift strategy. . . . . . . . . . . . . . . . . . . . 66 xiii xiv LIST OF TABLES Chapter 1 Introduction This chapter gives some background to the topic, and presents the idea behind the HEV (Hybrid Electric Vehicle). The problem description of this work is then discussed, as well as the main objectives. The chapter finishes with the outline of the report. 1.1 General Introduction In 2007, 16.1 million new cars were sold in the U.S. [4]. In [3] it is stated that about 97% of these were so called conventional cars, meaning that they have one propulsion system, namely an ICE (Internal Combustion Engine). The most common types of fuel used for the ICE are gasoline and diesel, which both are refined versions of crude oil. When this fuel is used, exhaust emissions and pollutants are released into the air. The pollutant that attracts most of the attention is CO2, which is one of the greenhouse gases. There is a constant pressure from organizations and governments on the automotive industry, as well as public concern, to reduce their impact on the environment. This has led to in- vestments in more environmental friendly vehicles emitting less emissions and pollutants. A better fuel economy, i.e. a low fuel consumption, is strongly correlated to a lower release of emissions and pollutants of the vehicles. Installing a catalyst in the vehicles is a step toward reducing pollutants, but to reduce the amount of needed fuel would be to get to the source of the problem. Another factor emphasizing the importance of a better fuel economy is the increase in oil prices during the recent years. A better fuel economy can be achieved in many ways, e.g. by reducing vehicle weight, increasing efficiencies in the drive systems or even driving more environmental friendly. Another approach is to change to a power source that consumes less fuel, or even no fuel at all, such as compressed air or fuel cells. A battery is also one possible power source, and this alternative was quite popular as the electric vehicle was in series production in the 1990’s. But mainly due to its high costs, long battery charging times and limited driving range, the electric vehicle’s popularity quickly declined. An alternative to exchanging the ICE completely, is to combine it with a second power source. In this way the benefits of the ICE can still be exploited, in combination with the benefits of the other power source. This kind of vehicle, which uses two different power sources, is known as a hybrid vehicle. Examples of a hybrid vehicle could be an electric bicycle, which uses both an engine and muscle power, or a boat that has an engine and sails. The type of vehicle that is most interesting for more fuel effective purposes is called an HEV, which combines the long driving range of an ICE with good fuel economy and low exhaust emissions of a battery. To be able to get the energy from the battery, an EM (Electric Machine) is 1 2 1.2. PROBLEM DESCRIPTION AND OBJECTIVES used, and the second power source is usually referred to as the EM. As stated in [16, 19], fuel consumption improvements of even up to 25% is achievable by the HEV compared to conventional vehicles, also resulting in reduced emissions and pollutants. The HEV, compared to the electric vehicle, does not depend on external charging of the battery and works well within the existing infrastructure for fueling at gas stations. One HEV that has gained much attention the last decade is the Toyota Prius, depicted in Figure 1.1. The Prius model is the first mass produced HEV and is currently market-leading in its segment [20]. It was introduced in the U.S. in the year 2000, and has steadily increased its sales every year with the releases of new models. Figure 1.1: Toyota Prius model NHW20 [5]. The concept of the HEV is more complex compared to that of a conventional vehicle, much due to the multiple power sources. This will give more possibilities to operate the powertrain, but also results in a more complicated control system. Since two power sources are available, they have to interact with each other in an appropriate way. Hence, a crucial part of every HEV is the HMS (Hybrid Management Strategy), consisting of supervisory controls, managing the power in the system and the interaction of the ICE and the EM. A well designed HMS will surely increase the possibilities of improving the fuel economy, which is the most important task of the HEV. 1.2 Problem Description and Objectives This thesis will investigate the functionality and performance of two existing HMSs, currently being developed at ZF Friedrichshafen AG. They are two separate strategies, and an analysis of them will be conducted. One of them is based on look-up tables which has been generated from optimization, where the fuel consumption is minimized using the ECMS (Equivalent Consumption Minimization Strategy). The other strategy has a more heuristic rule-based structure, and is less complex primarily for easier un- derstanding, tuning and fault detection. ZF is interested in the differences between the two strategies, and wants to investigate what features are important to consider when designing an HMS. New ideas for the rule-based strategy are encouraged, and the results of these implementations will be studied. Another important part of the study is to see how the two HMSs are able to cope with various design layouts, such as different vehicle setups and gear shifting strategies. This will give an indication of their adaptability and robustness. To be able to conduct this study and to be confident in the results, the existing simulation environment for the HEV has to be improved, and extended with some implementations. CHAPTER 1. INTRODUCTION 3 Considering this problem description, the main objective of the thesis can be summa- rized in the following formulation: Improve and extend the current simulation environment, so that the system is working properly. This is done by implementations preventing errors to occur, and also im- plementations of new parts in the two HMSs. Then analyze and compare the HMSs, and study their effect on the fuel consumption, as well as evaluate the robustness when changing to various design layouts. 1.3 Outline of the Thesis The remaining report is organized as follows. Chapter 2 explains more about HEV, its common setups and its most important components. In Chapter 3 the two HMSs that will be analyzed are described, and Chapter 4 gives an overview of how the simulation environment is set up, together with some new implementations. Chapter 5 gives the first simulation results, and the comparison between the two HMSs is discussed. Simulation results from various design layouts are presented in Chapter 6, followed by Chapter 7 which presents a new gear shift strategy. Finally the conclusions and recommendations are given in Chapter 8. 4 1.3. OUTLINE OF THE THESIS Chapter 2 Hybrid Electric Vehicles This chapter introduces the concept of the HEV and presents its most typical configura- tions. A short description of the important parts of such a system is given, with focus on the two power sources, the ICE and EM. The chapter ends with a summary part. 2.1 Hybrid Electric Vehicles As mentioned earlier, an HEV includes two power sources to propel the vehicle, an ICE and an EM. Hence there are two energy storage systems in the vehicle; a gasoline or diesel tank for fueling the ICE, and a battery for the EM. Often the HEV uses the ICE as a dominant power source and the EM as an assisting source, in order to provide a better fuel economy and lower emissions. Due to the low energy density of the battery compared to that of diesel or gasoline, the EM will give power for a shorter time than the ICE. This means, that in order to get a battery within the same “energy-range” as the fuel, the size and weight of it would have to be increased significantly, which is not an efficient solution for vehicles. When it comes to the storage systems, the fuel tank for the ICE is just like in a conventional vehicle, and is refueled at the gas stations when getting empty. On the other hand, the storage device for the EM, the battery, is recharged while driving the vehicle. This can be done in two ways; either by using extra power from the ICE, or by using free energy available when decelerating the vehicle. Hybrid powertrains can be designed in many different ways, e.g. depending on how the vehicle will be utilized, configuration of the power sources, the number of clutches etc. In this work the two most common schemes of the HEV are emphasized, namely the series hybrid and the parallel hybrid. 2.2 Series Hybrid In a series hybrid, the ICE is not mechanically coupled to the transmission’s input shaft, but charges all its power to a battery. The battery then supplies power to a number of EMs, which are mechanically coupled to the transmission’s input shaft providing it with power. A series configuration with two EMs is shown in Figure 2.1. The first EM is used to charge the battery with power from the ICE, and the second EM is coupled to the transmission’s input shaft, providing power from the battery. An advantage of this configuration is that the ICE can be optimized to work in the vicinity of its best operating point, without considering the actual driving situation. For example when doing “stop and go” driving, which usually happens in urban areas, the ICE is exposed to fast transients leading to many working points. As the number of working points 5 6 2.3. PARALLEL HYBRID increase, optimizing the ICE usage becomes more difficult, eventually resulting in a low average efficiency of the ICE. In this situation the series configuration has the possibility of letting the ICE work at its optimal working point, since it is not dependent on the road conditions. One negative effect of the series setup is the need for two EMs, which are costly components. Another is the loss off energy, due to all the energy conversions which are present between the ICE and the transmission’s input shaft. IC E B attery E M 2 Transm ission Transm iss ion ’s inpu t sha ft E M 1 W heel S teering w hee l Fuel tank Figure 2.1: A schematic of the series hybrid. 2.3 Parallel hybrid In the parallel hybrid depicted in Figure 2.2, both the ICE and the EM are mechanically coupled to the transmission’s input shaft, on the contrary to the series hybrid. There is also the possibility to decouple the ICE with a clutch. When the ICE is coupled, the power from the sources are added in order to fulfill the demanded power. Since the power from the ICE acts directly on the transmission’s input shaft, it does not need to go through all the power conversions as in the series setup, and a higher efficiency for the driveline is achieved. When the ICE and the EM are both coupled, they run at the same speed. The advantage of the parallel hybrid can be realized especially for long highway driving. In [9], it is stated that when driving at high velocities, the ICE is already working at points with high efficiencies, and the losses in the drivetrain between the ICE and the road are minimal. The parallel configuration is considered to be more efficient than the series one, when looking at a general driving pattern with both city and highway driving [6, 13]. Another benefit is that only one EM is needed, resulting in reduced costs, weight and space requirements. IC E B atte ry EM Transm issionC lu tch Transm ission ’s inpu t shaft W hee l S teering w heel Fue l tank Figure 2.2: A schematic of the parallel hybrid. In this thesis study the parallel hybrid is used, and the components which build up the parallel hybrid is explained in the following section. CHAPTER 2. HYBRID ELECTRIC VEHICLES 7 2.4 Components in the Parallel Hybrid 2.4.1 Battery The battery is the electrical energy storage, and usually the type Li-ion, NiMH or super capacitors. The battery is a so called bi-directional storage device, since it can charge and discharge itself when the vehicle is driving. An important part of the battery is its SoC (State of Charge), which tells how much energy [J] is available. The SoC should always be kept within a certain range of the battery capacity, e.g. 30% - 70%, in order to prolong its lifetime and increase its efficiency. When the battery level is starting to get either too low or too high, measures need to be taken to secure the limits from being reached. This, as will be seen later on, is done by the HMS, since it is responsible for controlling the SoC of the battery. If, for some reason, the SoC should reach very low or high levels, there are built-in safety features in the battery which disables it. 2.4.2 Clutch The clutch is the component that connects and disconnects the ICE to the transmission’s input shaft (not to be confused with the more commonly known clutch present in the transmission, which is used to shift gears). When the clutch is engaged (closed) the ICE is coupled to the transmission, providing torque. As the clutch is disengaged (opened), the ICE is disconnected from the driveline, and the demanded power of the driveline is provided only by the EM. 2.4.3 Fuel tank The fuel tank is the energy storage for the ICE, like diesel, gasoline, natural gas or hydrogen. This energy storage does not inherit the bi-directional functionality as the battery does, and is refueled when not driving. 2.4.4 Internal Combustion Engine The ICE is the main power unit in the powertrain, converting chemical energy from a fuel/air mixture into mechanical energy on the driveline. Depending on the specific setup of the parallel hybrid, the ICE can be turned off during driving and/or during standstill of the vehicle, to e.g. save fuel and reduce noise. An essential part of any HMS is to understand the characteristics of the ICE, and especially its efficiency. This will be a frequently discussed element throughout the chapters when covering the HMS. The model of the diesel engine specifies the fuel consumption in [g/h] as a function of its operating point, which is defined by two values; the engine speed ωICE [rad/s] and the engine torque τICE [Nm]. The fuel consumption is represented by a two dimensional matrix, which can be used to form fuel maps of the engine. A common representation of the fuel map is the engine efficiency, taken as the ratio of the mechanical output power and the chemical input power. The chemical input power is determined from the fuel’s lower heating value, which tells how much heat is released during combustion, in other words the energy content of the fuel. The lower heating value for diesel is typically 44.5 [MJ/kg]. The engine efficiency, ηICE [-], is computed as in (2.1) ηICE = Pm Pch = Pm lhvf · fc [−] (2.1) where Pm [W] is the mechanical output power, Pch [W] the chemical input power, lhvf [J/kg] the lower heating value and fc [kg/s] the fuel consumption. Another common 8 2.4. COMPONENTS IN THE PARALLEL HYBRID representation of the fuel map is the brake specific fuel consumption, which in fact is inversely proportional to the efficiency of the ICE. Example of an efficiency map is presented as a contour plot in Figure 2.3(a). The torque is normalized to the maximum torque of the discussed power source, and this kind of normalization is used for all torque axes in the following plots. Speed [rad/s] N or m al iz ed to rq ue [% N m ] 100 150 200 250 0 25 50 75 100 Efficiency [%] τ max [% Nm] OEL [% Nm] (a) Efficiency map, optimal efficiency line and the maximum torque of a diesel engine. The arrows indi- cate increasing efficiency, where the highest efficiency point is about 40%. Speed [rad/s] N or m al iz ed to rq ue [% N m ] Fuel consumption [g/h] 100 150 200 250 0 25 50 75 100 (b) Fuel consumption of a diesel engine, where the arrow indicates increasing fuel consumption. Figure 2.3: Example of an efficiency map (left) and fuel consumption (right) of a diesel engine. Also shown is the OEL (Optimal Efficiency Line), which describes the best torque and speed combination for any given power. The power, P , is computed from these two variables, as in (2.2) P = τ · ω [W ] (2.2) where τ [Nm] is the ICE torque and ω [rad/s] is the angular speed of the ICE’s crankshaft. Most relations expressed in this work will be torque-oriented, since the strategies are well explained using torques. Given the actual output shaft speed and demanded torque from the driver, is it still possible to change the ICE working point. The ICE speed is controlled by the gear shifts in the gear box, whereas the ICE torque is controlled by the HMS. To get closer to the OEL, which clearly would be advantageous in an efficiency perspective, the oper- ating point of the ICE can be moved. By combining an increase/decrease of the speed and torque, the working point can be moved over the torque-speed plane. To move the operating point by increasing/decreasing the torque is an important functionality of the HMS, which will be discussed more in Section 3.2.2. The fuel consumption of an ICE can also be shown in a speed-torque plane, as in Fig- ure 2.3(b). Here it is seen that the fuel consumption increases as the torque and the speed increases. It is understood that the idea of keeping a high efficiency of the ICE does not necessarily mean a low fuel consumption. Keeping low speeds and low torques will give a low fuel consumption, but on the other hand the efficiency of the ICE might be very low, which brings a contradictory effect. CHAPTER 2. HYBRID ELECTRIC VEHICLES 9 2.4.5 Electric Machine The EM is the second power unit in the powertrain, and functions both as a motor and a generator. The motor mode is used when the EM supplies tractive force to the vehicle by converting the electrical energy in the battery to mechanical energy at the driveline, leading to a discharge of the battery. The generator mode is enabled when the EM recuperates mechanical energy from the vehicle during braking, called regenerative braking, or during a recharge from the ICE. This energy is converted to electrical energy charging the battery. The efficiency of the EM is also described by an efficiency map, shown in Figure 2.4 for both positive and negative torques (used as a motor or generator). Speed [rad/s] N or m al iz ed to rq ue [% N m ] 50 100 150 200 250 300 −75 −50 −25 0 25 50 75 100 Figure 2.4: Example of an efficiency map for an EM, for motor mode (positive torques) and generator mode (negative torques). The arrows indicate increasing efficiency, where the highest efficiency point is about 90%. 2.4.6 Transmission The transmission device (gear-box) is used for adjusting the speed of the driveline to a desired speed of the wheels. It consists of a set of gears with different conversion ratios, and yields a torque-speed conversion from a (usually) higher speed of the driveline to a lower speed at the wheels, but with a higher torque. Included in the transmission device is also the clutch enabling the gear shifts (not to be confused with the previously discussed clutch, which couples and decouples the ICE from the driveline). 2.5 Impact of Hybridization Due to the flexibility and the additional components of the parallel hybrid powertrain compared to that of a conventional vehicle, some objectives and benefits can be listed, such as the following: • The ICE can be sized for a lower torque, since the EM can be used for boosting [16, 19]. Though, the maximum vehicle velocity should not be increased when 10 2.6. SUMMARY adding the EM, since this velocity always has to be achievable, even without the EM (due to manufacturer specifications). • Less wear on the brakes and reduction of waste energy due to regenerative braking. • Longer lifetime of the ICE, since unfavorable situations are avoided due to the power assisting EM. • Smaller and lighter fuel tank. • Electric driving, where the fuel consumption is at idle consumption or zero. • Less noisy operation since the ICE can be turned off during standstills. To use the EM for regenerative braking is a very important functionality of the HEV, and will be explained further in the next chapter. 2.6 Summary This chapter gives an introduction to the HEV and the most important components of the parallel configuration. The efficiency maps of the ICE and the EM are discussed, and their importance for the coming chapters is emphasized. It is realized that a HEV has a more complex structure compared to a conventional vehicle, but that it also gives more flexibility in how to design and control the system as a whole. The benefits of this will be shown later in Chapter 5, 6 and 7, when the HEV is simulated. Chapter 3 Hybrid Management Strategies In this chapter the idea behind the HMS is discussed together with its layout, followed by a description of the existing HMSs that will be studied in this work. Finally a summary part is given. 3.1 Hybrid Management Strategies Regarding the previous sections, it is obvious that there is the need of a supervisory control system for the hybrid powertrain, to let it operate in an efficient way. It is mainly the two power sources, the ICE and the EM, and their interaction that need to be controlled. This control system is called the HMS, and is a crucial part of every HEV. To put it in a compact formulation, the objective of the HMS in this work is the following: • To control the mode of the vehicle, i.e. mainly to set the status of the ICE (on/off). • To control the cooperation of the ICE and the EM, i.e. to decide the torque split between them. The goal of the HMS may be to e.g. minimize the fuel consumption or the emissions, or a combination of both. In this thesis the HMSs are optimized against minimizing the fuel consumption. The HMS is divided into four separate segments; SML (Strategic Mode Logic), TD (Torque Distribution), SoC Controller and Brake Management. The main structure of this division is seen in Figure 3.1, together with one important output from the SML and two from the TD. The four parts, even though they are separated, strongly interact and affect each other. Their functionalities are described below. H ybrid M anagem ent S trategy Stra tegic M ode Log ic Torque D istribution SoC C ontro lle r Brake M anagem ent Inputs M ode of the vehic le IC E torque EM torque Figure 3.1: The structure of the HMS as implemented in the simulation environment. 11 12 3.1. HYBRID MANAGEMENT STRATEGIES 3.1.1 Strategic Mode Logic In this part the decision about the mode of the vehicle is taken, where the most used modes are; Electric Drive with ICE off, Electric Drive with ICE on and Hybrid Drive. All of these modes describe different driving situations, and should be appropriately se- lected. The strategic mode of the vehicle is directly correlated to the status of the ICE, which can either be on or off. To control the status of the ICE is an important factor and doing so has a great potential to save fuel. Below the three modes are described more in detail, with their respective torque calculations. Electric Drive with ICE off In this mode the vehicle is propelled only by the EM, and the clutch in Figure 2.2 is disengaged. It does not require any torque from the ICE, and therefore it is turned off. All the demanded torque by the driver is taken from the EM, and the equations describing this mode are shown in (3.1) τICE = 0 [Nm] τEM = τdriver [Nm] (3.1) where τICE [Nm] is the torque from the ICE, τEM [Nm] is the torque from the EM and τdriver [Nm] is the demanded torque by the driver. To be in this mode will definitely have benefits regarding the fuel consumption and the emissions, since no fuel is consumed. A typical condition for switching the ICE off is when the vehicle comes to a standstill, and no torque is demanded by the driver. Electric Drive with ICE on This mode is similar to the one previously discussed, since the vehicle is propelled by using only the EM and the clutch in Figure 2.2 is disengaged. The difference is that the ICE is kept running at the idle speed. The two equations describing this mode are stated in (3.2) τICE = τICE, idle + τload [Nm] τEM = τdriver [Nm] (3.2) where τICE, idle [Nm] is the required idle torque, and τload [Nm] is the extra torque required for some applications such as the steering. Keeping the ICE on obviously leads to higher fuel consumption and more emissions compared to turning it off, but this mode might be necessary, since many vehicles need the ICE to power e.g. the air condition and the steering. There is also an energy loss when turning the ICE on, and turning it off should be avoided if it needs to be turned on again shortly after, which e.g. can happen if the battery is very low. Hybrid Drive In this mode the ICE and the EM are both used to propel the vehicle, and the clutch in Figure 2.2 is engaged. The sum of their torque contribution results in the demanded torque by the driver, as in (3.3) τdriver = τICE + τEM [Nm] (3.3) CHAPTER 3. HYBRID MANAGEMENT STRATEGIES 13 This is the most elaborate mode, as the torque needs to be split between the two power sources. There are infinitely many ways of combining the two power sources, and some- how the split has to be computed. This is where the importance of the TD is understood, as it is responsible of computing this split. 3.1.2 Torque Distribution The task for this part of the HMS is to decide the torque distribution for the system, when the mode Hybrid Drive is active. It is determined how to operate the two torque paths in the most favorable way, to satisfy the torque demand by the driver. In some situations it might be advantageous to take all the torque from the ICE, and nothing from the EM, typically when the SoC of the battery is very low. This would mean that the HEV is operated as a conventional vehicle, as the EM is not used. On the other hand, when the battery is fully charged, the best might be to supply most of the torque to the driveline from the EM, and only a small amount from the ICE. These kind of decisions are made in the TD part. The strategy can be designed in numerous ways, and is an important part of the HMS, in order to get a good performance. 3.1.3 State of Charge Controller This part is responsible for controlling the SoC of the battery, in such a way that it is kept within the allowed limits. The two analyzed HMSs have different ways of doing this, as will be seen later. 3.1.4 Brake Management The brake management system is needed for the HEV, in comparison to the conventional vehicle. It should make sure that the driver gets what he demands when braking, at the same time letting the components in the HEV operate in an efficient way. The two HMSs have the same brake management, and therefore there is no focus on this part when doing the comparison. In Section 4.6 the brake management is discussed further. 3.2 Rule-Based Hybrid Management Strategy One of the two HMSs is the rule-based strategy, as mentioned in Section 1.2. The structure of this follows the one illustrated in Figure 3.1, and here the three parts SML, TD and SoC controller are explained. 3.2.1 Strategic Mode Logic The SML is represented by a discrete-event system1, and implemented as a stateflow chart in the Simulink environment. A simplified version is presented in Figure 3.2. Each mode of the vehicle is represented by a specific state, in which an action can be performed when activated. In order to switch mode (state), the transition, illustrated by a line, needs to be fired. The transitions can be e.g. time functions and/or value functions, and represent conditions for changing the mode, such as demanded power or standstill detection. When the transition is fired, a new state is entered, and the vehicle switches mode. Every state can have many transitions connecting to different states, and are fired in a prioritized order if they should occur at the same time. 1For more information about discrete-event systems, see [2] and [7]. 14 3.2. RULE-BASED HYBRID MANAGEMENT STRATEGY Electric Drive ICE Off entry: do action 3 Electric Drive ICE On entry: do action 2 during: timer += tSample Hybrid Drive entry: do action 1 [Initial conditions] [Small power demand & High SoC] [Large power demand & timer > delay] 2 [Large power demand | Low SoC] 2 [Small power demand] 1 [Standstill detection & !Low SoC] 1 Figure 3.2: Simplified version of the stateflow diagram, representing the SML for the rule-based strategy. Depending on the current driving conditions, the SML will choose one of the modes Electric Drive with ICE off, Electric Drive with ICE on or Hybrid Drive for the vehicle. To make the switch between the different modes is a crucial point of the HMS, and surely much fuel can be saved if the mode changes are done correctly. One of the main variables deciding the switch is the torque demanded by the driver, in a relation to the current energy state of the battery. The power demand is computed from the driver as in equation (2.2). Then a time is computed, which tells how long the battery will be able to provide this power for the vehicle, if all the demanded power is given by the EM. Since this variable tells how long the vehicle can be driven electrically (either Electric Drive with ICE on or Electric Drive with ICE off), it can be compared to thresholds deciding the switching of the modes. The logic can be explained as in the following: • If the vehicle is able to drive electrically for more than 20 seconds, considering the current torque demand and SoC, switch to the mode Electric Drive with ICE on. • If the vehicle is able to drive electrically for less than 10 seconds, considering the current torque demand and SoC, switch to the mode Hybrid Drive. Typically the torque demand from the driver is high for accelerations, and therefore the time the vehicle will be able to drive electrically is short. Hence the SML will choose the mode Hybrid Drive, where the ICE can provide the high torque demand. If the vehicle is driven in a downhill, the demanded torque by the driver is very low (even negative), and electric driving should be chosen. Many variables are used in the logic deciding the mode of the vehicle, though the torque demand from the driver in relation to the SoC is the most important. Examples of other variables deciding the switching between modes are standstill detection, parking brake, ICE temperature, vehicle velocity, gear shift, etc. It can also happen that when going downhill, the battery becomes fully charged, and no more energy can be stored. In that case it might be better to switch from electric driving to hybrid driving, and use the engine braking functionality of the ICE. It should be remembered that the TD in the HMS is only active in the mode Hybrid Drive, as the other modes only use the EM for traction. Therefore, the TD is much dependent on the behavior of the SML, and the SML can be seen as a deciding function over the TD. CHAPTER 3. HYBRID MANAGEMENT STRATEGIES 15 3.2.2 Torque Distribution The idea behind the TD in the rule-based strategy is to have the possibility to always operate the ICE at a high efficiency. This can be done by moving the actual operating point closer to the optimal working point as in Figure 3.3, to increase the efficiency, and the methodology is known as load leveling. Speed [rad/s] N or m al iz ed to rq ue [% N m ] 80 100 120 140 160 180 200 220 240 260 280 0 25 50 75 100 Efficiency [%] τ max [% Nm] OEL [% Nm] Figure 3.3: Four arbitrary working points of the ICE are selected, to show that they can be moved to an area with a higher efficiency. In [1], it is thoroughly described that for a vehicle which is ICE dominated, like this one, the overall system efficiency is expected to be higher if the ICE is considered more than the EM, when designing the control system. Hence the focus will be set on improving the conditions under which the ICE is working. The TD for the rule-based strategy is computed in two steps as in (3.4) τEM = τEM, EM Boost + τEM, ICE Torque Increase + τEM, ICE Torque Decrease + τEM, Regenerative Braking [Nm] τICE = τdriver − τEM [Nm] (3.4) The torque from the EM is calculated from four different parts, namely EM Boost, ICE Torque Increase, ICE Torque Decrease and Regenerative Braking. They are best ex- plained with the help of Figure 3.4 in combination with the previously discussed ICE efficiency map, Figure 2.3, and is done in the following. EM Boost This part will assist the ICE when a high torque is demanded by the driver, especially when doing accelerations and e.g. hill-climbing. In Figure 3.4, the ICE is shown in the speed-torque plane, with the maximum torque line and the maximum driver torque line. The maximum torque is depending on the specifications of the ICE, and it will not be able to provide a torque over this line. In this work, the maximum driver torque line is set so that it coincides with the highest value of the maximum torque line of the ICE. The torque from the EM due to boost, τEM, EM Boost [Nm], is given in (3.5) 16 3.2. RULE-BASED HYBRID MANAGEMENT STRATEGY τEM, EM Boost =  τdriver, max − τICE, max [Nm] for τdriver > τdriver, max τdriver − τICE, max [Nm] for τICE, max < τdriver < τdriver, max 0 [Nm] for τdriver < τICE, max (3.5) where τICE, max [Nm] is the maximum torque of the ICE. As soon as the driver demands a torque above the maximum torque line, the EM boost function will be enabled. If the driver demanded torque is below the maximum torque line, the EM boost function is disabled. 80 100 120 140 160 180 200 220 240 260 280 0 25 50 75 100 Speed [rad/s] N or m al iz ed to rq ue [% N m ] τ ICE, max OEL τ driver, max τ ICE Torque Increase τ ICE Torque Decrease EM Boost ICE Torque Decrease ICE Torque Increase Figure 3.4: The curves representing the features in the rule-based TD. The ICE torque increase and the ICE torque decrease curves are set to arbitrary values. ICE Torque Increase This function, as its name implies, will increase the torque of the ICE. This is to take more torque than is demanded by the driver, and use it to charge the battery. The extra torque from the ICE can be seen as an equivalent negative torque from the EM, due to the computations of the ICE torque in equation (3.4); as the EM torque decreases, the ICE torque increases. The idea behind this feature has to do with the efficiency map of the ICE, as explained in Section 2.4.4. In Figure 3.4 the OEL and the ICE torque increase line are both shown. If the ICE is operated at a low efficiency below the ICE torque increase line, it could be advantageous to increase the torque to reach a higher efficiency, and store the extra energy in the battery. If the driver demands a torque below the ICE torque increase line, the torque of the ICE will be increased to the ICE torque increase line. Since this will result in a higher torque (and power) from the ICE than the driver demanded, the extra torque will be used to recharge the battery, which is formulated in (3.6) τEM, ICE Torque Increase ={ τdriver − τICE Torque Increase [Nm] for 0 < τdriver < τICE Torque Increase 0 [Nm] for τdriver > τICE Torque Increase (3.6) where τICE Torque Increase [Nm] is the value of the ICE torque increase line. If the driver demanded torque is above this line, it results in no torque increase. The value from CHAPTER 3. HYBRID MANAGEMENT STRATEGIES 17 equation (3.6) is saturated in a way that it always stays negative (negative EM torque gives positive ICE torque). To increase the torque of the ICE will lead to higher fuel consumption, but the hope is that more than this fuel can be saved at a later stage, by using the energy it managed to store in the battery during the ICE torque increase. It should be noted that increasing the torque of the ICE and save the extra energy in the battery will not necessarily be beneficial. This is due to the losses in the electrical system which occurs when converting the mechanical energy from the ICE to electrical energy for the battery. The extra energy taken from the ICE might be “lost” due to all the conversions, which would mean that the ICE torque increase was not useful. In this case the functionality of the torque increase should be disabled, and the EM should recharge the battery only by regenerative braking. ICE Torque Decrease This function can be seen as the opposite of the previous discussed one, as it decreases the torque of the ICE, i.e. takes less torque from the ICE than is demanded by the driver, and uses torque from the EM instead. If once again consulting equation (3.4), it is understood that a lower torque of the ICE is represented by a higher torque from the EM. This function also relates to the ICE efficiency map, but works in another region compared to the ICE torque increase. In Figure 3.4, the ICE torque decrease line is shown. If the driver demands a torque above this line, the torque of the ICE is lowered to the value of the line. In this way, the inefficient operating points present in the top left corner of the ICE efficiency map can be avoided (though this engine has quite high efficiency there also). The equations are given in (3.7) τEM, ICE Torque Decrease = τICE, max − τICE Torque Decrease [Nm] for τdriver > τICE, max τdriver − τICE Torque Decrease [Nm] for τICE Torque Decrease < τdriver < τICE, max 0 [Nm] for τdriver < τICE Torque Decrease (3.7) where τICE Torque Decrease [Nm] is the value of the ICE torque decrease line. If the driver demanded torque is below this line, it results in no torque decrease. In equation (3.7), the value of the ICE torque decrease is saturated in a way that it always stays positive (positive EM torque gives negative ICE torque). In contrast to the ICE torque increase, this function will give a lower fuel consumption when activated. Since the ICE efficiency map is non-linear, different amounts of fuel can be saved depending on the different working points. It is therefore important to do the ICE torque decrease when much fuel can be saved, with a little amount of electrical energy. Regenerative Braking This is perhaps the most important part of the HEV and its fuel consumption savings. The regenerative braking is enabled when the vehicle is braking. e.g. at traffic lights and downhill slopes. For a conventional vehicle, in braking situations the kinetic energy of the vehicle is disposed in the mechanical brakes at the wheels, converting the energy into waste heat. Instead of using the mechanical brakes, the HMS will use the EM to take up the energy and recharge the battery. The braking of the vehicle is represented by a negative torque demanded by the driver, and the equation for the regenerative braking is given in (3.8) τEM, Regenerative Braking = { τdriver [Nm] for τdriver < 0 0 [Nm] for τdriver > 0 (3.8) 18 3.2. RULE-BASED HYBRID MANAGEMENT STRATEGY where τEM, Regenerative Braking [Nm] equals the negative driver demanded torque. If the driver demanded torque is positive, it results in no regenerative braking. Since this part is not concerned about the ICE efficiency, no figure is shown in this case. Final EM torque computation Each of the parts are limited such that they do not interfere with the other parts, and they are active in their region defined by the lines. Some parts might be active simul- taneously, but it is guaranteed that their individual contributions are limited to their operational region. The result of adding these four features together is the EM torque of the TD, previously formulated in equation (3.4). The tuning parameters of the rule-based TD are the lines of the ICE torque increase and the ICE torque decrease, and should be set to values giving a good control of the ICE efficiency. The EM boost and the regenerative braking will be active as previously discussed. 3.2.3 State of Charge Controller Up till now, the discussion of the ICE torque increase and the ICE torque decrease has only been seen as a way to reach a better efficiency of the ICE. But the two features have a second purpose as well, as they will be the SoC controller of the rule-based strategy. Not only will they care about the efficiency of the ICE, but also about the SoC level of the battery. Consider the following two examples: • If the SoC level is getting too low, the charge of the battery needs to be increased. The best thing would be to do it by regenerative braking, but it is not known when a situation like that will occur. To be sure not to deplete the battery, the ICE needs to be used to charge the battery. This is done by taking more torque from the ICE than the driver demands, and charge it into the battery. Therefore the ICE torque increase function is enabled, as it does exactly this thing. • If the SoC instead is getting too high, it would be preferable to use the battery as much as possible, and the ICE as little as possible. Therefore the ICE torque decrease is enabled, and less torque is taken from the ICE, with the purpose to use the battery more and save fuel. As realized from the example above, the ICE torque increase and the ICE torque de- crease have two missions; one of them is to increase the efficiency of the ICE, and the other one is to keep control of the SoC level. It can be argued that the effect of the ICE torque increase/decrease should be stronger when the SoC is getting to its low and high extremes. Hence, the increase and decrease lines are stronger and more effectful in these cases, and the focus is set on controlling the SoC. If the SoC would actually reach its extremes, the increase and decrease are very strong to force the SoC away from these points, to prevent battery damage. In this case the efficiency of the ICE is disregarded, as saving the battery is more important. When the SoC is in a normal range (zone), not close to the low or high extremes, no actual control of the SoC is required and it is allowed to move freely. In this case the focus is set on the efficiency of the ICE. CHAPTER 3. HYBRID MANAGEMENT STRATEGIES 19 3.3 Table-Based Hybrid Management Strategy The second HMS follows the same layout as the rule-based strategy, as it also consists of the parts SML, TD, SoC controller and brake management, though they are implemented in another way (except the brake management). Some parts of the table-based strategy are implemented from the Dymola software into the Simulink environment, as they not yet existed. The strategy is mainly based on look-up tables, containing data from the ICE, EM and other important components of the vehicle. These tables are optimized by using the ECMS method, explained further down in the section about the TD. First the SoC controller is explained, then the TD and last the SML. 3.3.1 State of Charge controller This HMS has a dedicated SoC controller, which objective is to control the SoC level in a predictive way, by adjusting the SoC of the battery to a specified level. The prediction is based on future recuperation possibilities, which become available when the velocity of the vehicle increases. A feedback controller, Proportional-Integral (PI) controller type, is implemented to track a certain set value, which changes throughout the route. Its implementation is shown in Figure 3.5. + - Desired SoC Vehicle Model PI- controller Error signal SoC controller SoCdes Control signal SoCact TD SML SoCu Reference signal Figure 3.5: SoC controller for the table-based strategy. The SoCdes should be followed without major oscillations, and should (eventually) be reached exactly, which implies that integral action is needed in the controller. The error signal is computed as SoCdes − SoCact, and the control signal SoCu is used as input to two tables in the SML and the TD. Since this control signal is used for both the SML and the TD, potential conflicts between them are avoided as the SoC is increased or decreased. A desired SoC level is set to a base value, e.g. 52%, which can be lowered as the velocity of the vehicle increases. The expression for the reference SoC, SoCdes, is given in (3.9) SoCdes = β − Ek = β − m · v2 2 [%] (3.9) where β is the base value expressed in [J], Ek [J] the kinetic energy, also fully expressed with the mass m [kg] and the velocity v [m/s]. The reason for the desired SoC to change, is that the kinetic energy of the vehicle increases with the velocity (squared), and this energy has a potential to be recuperated and converted to electric energy when braking, charging the battery. The desired SoC is lowered to increase the charging capacity of the battery for the regenerative braking, so that it does not overcharge. It is always better to use the energy in the battery than not be able to recuperate the available braking energy. Therefore, sometimes the EM will be used more at working points that will 20 3.3. TABLE-BASED HYBRID MANAGEMENT STRATEGY discharge the battery, even if it is not optimal from a TD point of view. In [18], another kind of predictive strategy is used for simulations with an HEV. There the route of the vehicle is assumed to be known a priori, and a prediction is made of when to empty the battery, to assure sufficient absorbing capacity for the coming regenerative braking. This can be seen as a kind of extreme SoC controller, as it exactly knows when to lower the desired SoC to discharge the battery ahead of the decelerations. In real life, the route and the driving conditions are not known in advance, and therefore the predictive SoC controller based on kinetic energy used here is a good compromise, and has potential to lower the fuel consumption. 3.3.2 Torque Distribution The torque distribution for the table-based strategy is based on the so called ECMS, which is a local optimization strategy considered to be relatively easy to realize. The variable subjected to the optimization is the fuel consumption, i.e. the fuel consump- tion is minimized. It is based on the fuel flow in the system, and is explained more below. ECMS The ECMS is an instantaneous minimization strategy, which means that it tries to optimize the fuel consumption at every time instance. The result is the optimal torque distribution between the ICE and the EM, which uses the minimum amount of fuel at each time instance. One could first think that only using the battery would lead to no consumption at all, and that it would be the best idea. However, the problem is that the battery will be depleted fast, and only the ICE can be used for further driving. Thus, in order not to deplete the battery at a fast rate, it has to be recharged during the drive, either by means of regenerative braking which does not consume any fuel, or by power supplied by the ICE, which does consume fuel. Since the charge sustaining behavior of the battery is a criteria, the EM will consume fuel in some sense, when it uses power from the ICE to recharge the battery. But, since the energy taken from the battery does not give an indication of how much fuel it is equivalent to, the fuel consumption from the two sources is not directly comparable. Therefore the energy taken from the battery will be converted to an equivalent amount of fuel, so that it can be compared to that of the ICE. The equation that will be used for the optimization describes the instantaneous total fuel flow in the system, ṁtot f (t), as in (3.10) ṁtot f (t) = ṁf(ωICE(t), τICE(t)) + ṁeq f (ωEM(t), τEM(t)) [kg/s] (3.10) where ṁf is the ICE fuel flow and ṁeq f is the equivalent fuel flow, representing the en- ergy taken from the battery. When doing the calculations for the EM’s equivalent fuel consumption, the efficiencies of the components have to be taken into account. This is because some power will be lost, for example when using the ICE to charge the battery (converting from ICE to EM, EM to battery, battery to EM and finally EM to trans- mission’s input shaft), instead of taking it directly to the transmission’s input shaft. Since these efficiencies depend on future working points, which are unknown, they will be represented by mean values. The ECMS is concluded in the following statement: • Depending on the required speed and torque at a point in time, an optimal control signal is available which distributes the torque between the ICE and the EM in such a way that the total instantaneous fuel consumption is minimized. CHAPTER 3. HYBRID MANAGEMENT STRATEGIES 21 The ECMS is computed off-line, and the values of the optimal control signal is stored in a table used in the on-line implementation. Implementation of the ECMS The values from the ECMS is used in a four dimensional table, to compute the torque split between the ICE and the EM, illustrated in the left of Figure 3.6. The inputs are the demanded torque by the driver, the speed of the transmission’s input shaft, the actual SoC and the desired change of SoC coming from the SoC controller. The desired SoC change, SoCu, gives an indication of the ICE torque increment that is necessary to reach the desired SoC. The output will be the desired torque of the ICE, τICE [Nm], which is subtracted from the driver demand to make out the EM torque, τEM [Nm]. Both equations are shown in (3.11) τICE = LUToutput(τdriver, ωinput, SOCact, SOCu) [Nm] τEM = τdriver − τICE [Nm] (3.11) where LUToutput [Nm] is the output from the four dimensional table. A surface plot describing the TD is shown in Figure 3.7, where the SoC inputs are fixed to a certain value. For every combination of speed and driver demanded torque τdriver, there is a defined torque taken from the ICE, τICE. Though this strategy is computed in an optimal way, it does not necessarily mean that it will perform optimal under conditions other than those used for the optimization itself. The optimization is dependent on e.g. vehicle setup and estimated mean efficiency values, and if these conditions are changed, it will affect the strategy. The ECMS tables used for this work are computed in another software and its special optimization conditions will not be the same when it is simulated. This will of course affect the results, and should be kept in mind when doing the comparison. 4D Look-Up Table Demanded torque Transmission’s input shaft speed Actual SoC Desired SoC change ICE torque driverτ actSoC uSoC 3D Look-Up Table Transmission’s input shaft speed Actual SoC Desired SoC change ThresholdactSoC uSoC Figure 3.6: The two tables used in the TD (left) and the SML (right) for the table-based strategy. 3.3.3 Strategic Mode Logic The principal part of the SML is a table, shown in the right of Figure 3.6. Its output is a threshold which will be compared to the driver demanded torque, and make a decision about the status of the ICE. The expression is shown in (3.12) ICE Status = { on for τdriver > threshold + hysteresis off for τdriver < threshold - hysteresis (3.12) where both time and amplitude hysteresis is added to avoid frequent toggling. The inputs to the table are the speed of the transmission’s input shaft, the actual SoC and the desired change of SoC coming from the SoC controller. The three inputs will influence the decision in the following way: 22 3.4. A GLOBAL OPTIMAL CONTROL STRATEGY: DYNAMIC PROGRAMMING • Input shaft speed: The characteristics of the components in the system are depen- dent on different input shaft speeds. At low ICE torques the fuel efficiency is low and Electric Drive will be desired instead of Hybrid Drive. This means that the fuel will be used in a more efficient way. • Actual SoC: For a low actual SoC, the decision will be biased towards keeping the ICE turned on, and vice versa for a high actual SoC. • Desired SoC change: If a positive change is desired from the SoC controller, the decision will be biased towards keeping the ICE turned on (recharging power is taken from the ICE to increase the SoC). On the other hand, if there is a negative change desired from the SoC controller, the decision will be biased towards turning off the ICE. The surface plot in Figure 3.7 can be used to describe the SML as well. Here the high- lighted red curve shows the threshold value when the ICE is turned on/off. 0 25 50 75 100 84 178 272 0 25 50 75 100 Normalized τ driver [% Nm]Speed [rad/s] N or m al iz ed τ IC E [% N m ] Figure 3.7: Surface plot for the SML and the TD of the table-based strategy. 3.4 A Global Optimal Control Strategy: Dynamic Pro- gramming A commonly used approach in literature to find the global optimal HMS is dynamic programming, which is based on Richard Bellman’s principle of optimality [12, 11, 8]. The control problem is stated as minimizing a cost function, such as fuel consumption, and dynamic programming is used to find the optimal way to use the ICE and the EM. The optimal strategy is obtained through recursive computation, where the equations are solved backwards in time. The final state of the system is used as a starting point for the algorithm, and the best way (out of all possible ways) to get to this state from the starting state is chosen as the optimal control law. For every instance in time many different control signals are plausible, but only one is optimal. Due to all these possible control signals which have to be calculated and evaluated, lots of computation effort is required. This long computation time is a great disadvantage of the global optimization approach and hence not applicable for real-time implementation in the vehicle. The other reason for its lack of implementation possibilities is that the route which the vehicle will drive needs to be known beforehand, to be able to make use of the recursive feature of CHAPTER 3. HYBRID MANAGEMENT STRATEGIES 23 the algorithm. This is not the case for normal driving, though research is being made that is using traffic and road information from GPSs. Since this work is focused on the comparison between two realizable control strategies, the dynamic programming approach is not implemented. 3.5 Summary In this chapter the concept of the HMS is explained, and the essential parts are described. An HMS’s purpose is to use the power sources, ICE and EM, in a way that fulfills the demanded performance, and the HMSs are optimized against fuel consumption in this work. The parts in the HMS are the TD, SML, SoC controller and brake management. The TD is the part which decides the split of the torque between the ICE and the EM, and will be active when the vehicle is in the mode Hybrid Drive. The SML handles the switching of the modes in the vehicle, which can be any of the three Electric Drive with ICE off, Electric Drive with ICE on and Hybrid Drive. The SoC controller handles the charge level in the battery, to protect it from damage and optimize the performance. The two HMSs which will be analyzed in this work are extensively discussed, and their implementations are given. The rule-based strategy has a more transparent configura- tion, which is based mainly on optimizing the operation of the ICE. The table-based strategy is on the other hand obtained through optimization, where the ECMS gives the tables which are used. In this strategy also an explicit SoC controller is implemented. To get an overview of the two HMSs, a table of their features is shown in Table 3.1. Table 3.1: Overview of the HMSs. HMS Strategic Mode Logic Torque Distribution SoC controller Rule-based Changes mode de- pending on the cur- rent torque demand in relation to e.g. the battery charge and allowed EM torque. The main part is a stateflow, which switches mode de- pending on the in- puts. Based on engi- neering intuition and component efficiencies. The TD is divided into different functions; ICE Torque Increase and Decrease, EM Boost and Regener- ative Braking. The outputs are then summarized. Included in the TD and the SML. Table-based Changes mode de- pending on the rela- tion between the out- put from the look-up table and the driver demanded torque. Look-up table based on ECMS optimiza- tion. Explicit PI - con- troller which sends signals to the TD and the SML. 24 3.5. SUMMARY Chapter 4 The Simulation Environment This chapter presents the simulation environment used in the work, like driver, driving cycles, etc. Some specific implementations and fixes in the simulation environment are discussed. In the end of the chapter is a summary part. 4.1 Simulation Environment The software used for the simulations and the data analysis is Matlab and its simula- tion tool Simulink. Some parts, for example the model of the vehicle, is build up in the Modelica language, and is only available in Dymola. These Dymola parts will be called on by Simulink, and are integrated in the Simulink environment. The simulation environment at ZF is a big and complex system, with a large number of parameters and variables. This system has been developed in a way so that the step from simulations to real vehicle should go as smooth as possible. Therefore, many functions and features which are used in the simulation environment are very complex for simula- tion purposes, but are still implemented for an easier transition to the real vehicle. This sometimes caused difficulties when analyzing and extending the simulation environment, as the structure of the system was not completely known. The system is constructed in the way shown in Figure 4.1, which can also be seen as a hierarchy of the decision making process. The part Hybrid Management Strategy has been discussed in the previous chapters, and consists of the two HMSs that will be analyzed. It is the highest level, where all requests should be allowed and there is no consideration of how it later will be realized. This is the part which will be implemented in the real vehicle. The outputs are sent to another part, considered as a “black-box” in this work since the functions in here are unknown, due to confidentiality reasons. Here, the requests from the HMS will be realized through operational signals, if it is allowed by the hardware. As this part was concealed, it sometimes created problems during the debugging, since it was unclear how the output from the black-box was computed. Finally the outputs are forwarded to the third part, containing the modeled vehicle. 25 26 4.1. SIMULATION ENVIRONMENT H ybrid M anagem ent S tra tegy M ode of the vehic le EM torque ICE torque C lutch s ignals Battery s igna ls C om ponent constrains Adjusted IC E /EM torque Adjusted m ode o f the veh ic le Actual ICE torque and sta tus Velocity of the vehic le Acce leration /brake signalsActual EM torque Stra teg ic M ode Logic O perational Layer (”B lack Box”) Protective Functions Conversions V ehic le M odel G ear Shift S trategy Vehic le Dynam ics and D river Torque D istribution SoC Controller B rake M anagem ent Figure 4.1: The structure of the simulation environment at ZF, together with some important inputs and outputs. 4.1.1 Implementations and Improvements The simulation environment is extended, e.g. two gear shift strategies are implemented (more in Chapter 7). Improvements of the simulation environment are also implemented and two examples are discussed further down, the other ones are left out. Avoiding Dead-Lock During the simulations of the system, it is realized that the vehicle can easily get stuck in different states, like a kind of “dead-lock”. This often results in that the vehicle is unable to provide power to the wheels, or shift gears, and the simulation crashes. The cause of this is that too many different state changes are requested by the system simultaneously, and it can not treat them all in a proper way. This has mainly to do with problems in the implementation of the operational layer and the interaction with the vehicle model. An example of a typical dead-lock is shown in Figure 4.2, where the vehicle mode and the gear shifts are shown. 0 1 2 3 4 5 6 7 8 9 10 Time [s] Vehicle Mode OK Shift time Gear 6th Gear 5th Gear 4th Gear Electric Drive Dead−lock Hybrid Drive Figure 4.2: A dead-lock situation which could occur in the system, if the mode of the vehicle was changed during a gear shift phase. For every gear shift, there is a certain shifting time which is necessary to engage the new gear. If a mode change would occur during this shifting time, it could result in a com- plete dead-lock of the system. Since the main focus is set on the Hybrid Management Strategy part, the solution is to make sure that the requested mode change is put on hold until the shift is completed. It can be thought of as a queuing system, where the requested state transitions are organized and only allowed when certain conditions are met. This might lead to a disadvantage for the HMS, if a requested mode is not allowed CHAPTER 4. THE SIMULATION ENVIRONMENT 27 for some seconds. Derating of the EM The initial HMS had a part for limiting the negative torque requested by the EM, which purpose was to prevent negative torque on the transmission’s input shaft at standstill. If that was not done, it could cause undesired negative acceleration for the vehicle, which should be avoided both for comfort and safety reasons. Also, another very important reason for limiting the negative EM torque is that it inflicts mathematical errors in the Dymola model, and as a result the simulation crashes. The limitations of the EM torque is necessary, but should be used as seldom as possible, only close before and at standstill situations. One of the implemented improvements is that the EM torque is not limited if the acceleration is over a specified positive threshold, to make sure that the EM only is limited when the vehicle is going to a standstill. Previously, only the speed of the input shaft was considered, which did not take care of the positive acceleration. Another problem was that the TD calculated the different torques without any knowledge of when the EM was limited. This could lead to the following problem: • The ICE torque is increased and the EM torque decreased, leading to a battery recharge. After this, the negative EM torque is derated, which results in an unnec- essary increase of the ICE torque, since it would be equalized by the mechanical brakes, and not used for recharging the battery which was the intention. After the fixes, this problem could no longer occur. 4.2 Vehicle The simulated vehicle is a small delivery truck, similar to the one seen in Figure 4.3. Its components include, among others, an ICE, an EM, clutches, a battery, vehicle driveline and mechanical brakes. The dynamics of the vehicle and tires are also present, and parameters such as drag coefficient, mass and rolling resistance. The vehicle is not a fully electrical vehicle, meaning that it is not able to drive electrically when the ICE is turned off. Therefore the mode Electric Drive with ICE off can only be selected when the vehicle is at standstill, meaning that the electric driving is done in the mode Electric Drive with ICE on. Figure 4.3: Small delivery truck used for the simulations [15]. 28 4.3. DRIVING CYCLES 4.3 Driving Cycles In order to run the simulations for the vehicle, a predefined driving cycle is used. The cycle is specified by a desired velocity of the vehicle, for different points in time of the simulation. The cycle can be designed in a countless of ways, but should preferably represent a real driving pattern which is typical for the specific vehicle. A somewhat more synthetic cycle, with constant acceleration and deceleration parts, will also be used. This cycle is especially good to study the behavior of the vehicle, to easier find unwanted effects and faulty behavior, since it is easy to know the wanted behavior. The three driving cycles which the vehicle is tested on are shown in Figure 4.4. In this work, one important aspect is to test the behavior of the HMSs on different cycles, since it is very easy to sub-optimize the HMS to perform well on one certain cycle. The more cycles that are tested, the more confident one can be about the performance of the HMS. Though, simulating many and long cycles is a tedious task, and the selection ultimately has to be limited to a few. 0 20 40 60 80 100 120 140 160 180 0 50 100 Sort3 V el oc ity [k m /h ] 0 200 400 600 800 1000 0 50 100 Customer V el oc ity [k m /h ] 0 200 400 600 800 1000 1200 1400 1600 1800 0 50 100 Japan5 Time [s] V el oc ity [k m /h ] Figure 4.4: Three cycles that are used in the simulations with the small delivery truck. 4.4 Level of the State of Charge An important remark considering the simulations is the charge of the battery, the SoC. In order to get comparable results from two different simulations, the SoC level for each of the simulations has to be well balanced. This means that the simulation should end with an SoC value that is close to the start value, i.e. it should have a small ∆SoC. This is formulated in (4.1). |∆SoC| < ε (4.1) where |∆SoC| [% units] is the absolute value of the ∆SoC, and ε [-] is a threshold. The threshold should preferably be as small as possible. If the |∆SoC| would be too large, the CHAPTER 4. THE SIMULATION ENVIRONMENT 29 results of the fuel consumption could not be accurately studied. Consider the following example: • Simulation 1 starts with SoC of 50%, and finishes with SoC of 60%. It has increased the SoC of 10%, which has increased the total fuel consumption, due to higher power from the ICE. • Simulation 2 starts with SoC of 50%, and finishes with SoC of 40%. It has decreased the SoC of 10%, which has decreased the total fuel consumption, due to lower power from the ICE. If the fuel consumption of the simulations were to be compared, it would give misleading results. The first simulation would have a disadvantage of its increased SoC level, and the second simulation would have an advantage of its decreased SoC level. If there was a linear relation between the fuel consumption and the ∆SoC, a balanced SoC would not have been as crucial. In that case, a factor x could be used to “scale” the actual fuel consumption, taking the ∆SoC into account. But, since there is no linear dependency of the two variables, i.e. the factor x does not exist, a balanced SoC is very important for the analysis. In order to minimize the ∆SoC, an algorithm has been developed to compute an optimal start SoC, which tries to minimize the |∆SoC| in (4.1). Since the main part of the analysis will be to compare the fuel consumption from different simulations, the fuel consumption itself will be used as an important variable in the algorithm. Some criteria for the algorithm are shown below, and the algorithm is schematically presented as a flowchart in Figure 4.5. • A maximum of 15 simulations will be run (though for the simulations done in this work, a balanced SoC was found before 15 simulations). • Two simulations, one with a positive ∆SoC and one with a negative ∆SoC, should occur. • If the fuel consumption for the last iteration is only differing slightly compared to the best value up until “now” (which is given from the best ∆SoC), a good start SoC was found. In the algorithm, several simulations are iterated. For every simulation, it is checked if the criteria for a balanced SoC level is reached, otherwise it continues to iterate (maxi- mum 15 times). A new start SoC is computed after every iteration and used as a start SoC for the next iteration, with the aim of giving a more balanced SoC level. The result of the algorithm will be the best start SoC, and the most accurate value of the fuel con- sumption (interpolated between the two best iterations) used for the comparison. The algorithm is explained further in Appendix A. 4.5 Driver Model One important aspect which has to be considered is how the vehicle will respond to the demand from the driver. The additional components in the HEV powertrain will give the vehicle further possibilities to react on the driver’s demand, though, since the driver is used to a conventional vehicle, the response should be similar for the HEV. This is important for safety reason, but also for comfort. The features that the driver expects 30 4.5. DRIVER MODEL Start: in itia te param eters S im ula te system Save end SO C and fuel consum ption Calculate de ltaSO C IS deltaSO C < old deltaSO C ? Save deltaSO C as the best one YES IS deltaSO C positive ? NO IS pos itive deltaSO C < old positive de ltaSO C ? YES Save positive deltaSO C and its fuel consum ption YES Save negative deltaSO C and its fuel consum ption IS negative deltaSO C > old negative deltaSO C ? NO YES IS deltaSO C’s zero line crossed ? NO N O Interpolate fue l consum ption between the two best s im ula tions YES N O IS # sim ulations >= 15 O R (deltaSO C ’s zero line crossed = TR U E AN D interpolated fuel consum ption not changing ) ? YES NO Decis ion (ask question ) Activ ity (do action ) Boxes in the flowchart Finished : a good start SO C w as found Save fue l consum ption N ew start SO C = current start SO C - spec ified value TRU E if a positive and a negative deltaSO C is available for first s im ula tion : end SO C specified va lue = if deltaSO C ’s zero line is crossed for th is s im ula tion : ½ of last deltaSO C for other s im ula tions : som ewhat h igher than last deltaSO C V ariab les in the flowchart if deltaSO C ’s zero line has been crossed : last deltaSO C deltaSO C ’s zero line crossed : N ew start SO C = current start SO C + spec ified value Figure 4.5: Algorithm to balance the SoC level, presented as a flowchart. CHAPTER 4. THE SIMULATION ENVIRONMENT 31 are primarily the vehicle response from the acceleration and brake pedals. When pressing the accelerator pedal, a positive torque should accelerate the vehicle, and if the accelerator pedal is released, a low negative torque should be the result. The torque should be specified for a certain angle of the accelerator pedal, when pressed, and an ICE angular speed. It is similar when releasing the accelerator pedal; the negative torque should be the same for a certain pedal angle and an ICE angular speed. The response from the vehicle, when the brake pedal is pressed down, should be fast and predictable for the driver. To make the HEV work in a predictable way, an angle on the brake pedal should affect the vehicle with a particular negative force, which is similar to a negative torque on the transmission’s output shaft. In the real vehicle, used for testing, the EM torque is only connected to the accelerator pedal, because of the actual configuration. The simulation model is therefore imple- mented in the same manner. When the accelerator pedal is pressed, a positive EM torque can be used by itself, or in combination with the ICE, to fulfill the demanded torque. A negative torque from the EM, which occurs when releasing the accelerator pedal, will be interpreted as engine brake by the driver. The difference between the engine brake for the HEV in the simulations, and a real conventional vehicle, is the force of the engine brake. A conventional vehicle can only brake with the ICE, but the HEV can also use the maximum negative torque of the EM. The brake pedal in the HEV model has no major differences to that in a conventional vehicle; all negative force from the brake pedal is performed by the mechanical brakes. The driver implemented in the model will only use the mechanical brakes if the engine brake, with the ICE and the EM, can not fulfill all the braking force. In this way the maximum energy is recuperated. Situations which need the mechanical brakes occur, in electric driving, when the negative demanded torque is below the minimum allowed EM torque, or when the battery is fully charged. In hybrid driving it occurs when the negative demanded torque is below the minimum allowed EM + ICE torque. The reason for this implementation is that the real vehicle has a mechanical connection between the brake pedal and mechanical brakes, just like in a conventional vehicle. As mentioned above, it is important that the HEV behaves in a predictable way. There- fore, changes in the SoC should not influence how the vehicle responds to the driver’s demand. A typical example is the EM boosting functionality. It requires energy from the battery, but when the SoC is too low, the EM boosting is not available anymore. This might be dangerous if the driver does not get the expected power, e.g. in an over- taking situation. A similar problem occurs when braking with negative EM torque; if the SoC is too high, indicating a full battery, the EM braking is not available. To solve the problem with the EM boost complication, it can be chosen to only allow torques up to the maximum ICE torque, to make sure that it is always possible to fulfill the driver demand regardless of the SoC. If it is chosen to allow higher torques than the maximum ICE torque, which is the case for the simulated vehicle, it can be advantageous to keep the SoC at a high level, so that the EM boost functionality always can be used. Also, the highest torque possible should not be too high, so that the EM boost will discharge the battery fast. 32 4.6. DRIVER IMPLEMENTATION To solve the problem with an unexpected behavior, when the driver requests a negative torque, a brake management system should be implemented. When the driver wants to brake and presses down the brake pedal, a brake management should analyze if this can be accomplished with the EM, or if also the mechanical brakes are needed. The brake management should make sure that the driver gets the negative torque he expects and at the same time minimize the use of the mechanical brakes. 4.6 Driver Implementation A part of the simulation environment is the artificial driver, which is intended to resemble the driving style of a real driver. The purpose of the driver is to track the desired velocity of the vehicle for every time instance, as predefined in the cycle previously discussed. This is done by adjusting the accelerator and brake pedal. The modeled driver is implemented with a PI-controller, seen in Figure 4.6. vref + - vact e Predefined route Vehic le M odel Torque conversion P I- contro ller E rror Dem anded torque ur y D river H M S Figure 4.6: Basic parts of the artificial driver used in the simulations. The controller takes the desired velocity from the predefined driving pattern as a refer- ence signal, and subtracts the actual velocity from the vehicle model. This results in the error signal, e, also given in (4.2) e = desired velocity− actual velocity (4.2) The error signal is sent to the PI-controller and then converted to a torque, which makes out the requested torque by the driver, at the transmission’s input shaft. This torque is used as an input signal to the HMS and the vehicle model, which finally sends out the actual velocity of the vehicle. The proportional and the integral gain in the driver had to be tuned, making it follow the desired trajectory in a good way. There is mainly one difference observed in behavior compared to that of a real driver. It has to do with the gear shifting, especially for accelerations. When shifting a gear, the driver shows a large increase in requested torque, which probably would not occur for a real driver. The reason for this is because the clutch is disengaged during a gear shift, and there is no torque on the driveline propelling the vehicle at that moment. The driver reacts on this loss of velocity, and increases its demanded torque. This would not be the case for a real driver, since it is accepted to loose some speed when shifting the gear, and the real driver would not react on this. Since the driver will not be able to follow the cycle perfectly and its behavior is depen- dent on the behavior of the system, it will affect the simulations differently. Therefore, it is important to make the driver follow the cycle well, so that it does not influence the CHAPTER 4. THE SIMULATION ENVIRONMENT 33 analysis to a great extent. 4.7 Benchmark Model In order to understand the benefits of the HEV more clearly, and relate it to somewhat “common facts”, a conventional vehicle will be used as a benchmark model. It is created by changing the settings in the model of the HEV. The conventional vehicle will not use the EM, and the ICE is the only power source. Since the ICE always needs to be on give the torque demanded by the driver, the SML and the TD are not necessary, and will therefore be disabled. Thus, it will be possible to do an absolute comparison of the two HMSs, where the conventional vehicle works as a base. The analysis will be more clear, if a comparison is made to a more “familiar” type of vehicle. 4.8 Summary In this chapter the structure of the simulation environment is given. It is implemented in Simulink, except for parts of the vehicle model which uses Dymola. Some implemen- tations are given which have to be done to be able to simulate the vehicle. Also two important improvements are discussed, where the first one avoids dead-lock situations. The second one is the limiting of the negative EM torque when coming close to stand- stills, in order to avoid discomfort for the driver and simulations to crash. The type of vehicle which is simulated is a delivery truck, and can only turn off the ICE at standstill. Therefore, the potential fuel savings of driving electrically are not as high compared to if the ICE could be turned off during driving. The driving cycles which the vehicle will be simulated on are given. To compare the fuel consumption between different configurations, an algorithm is implemented. The algorithm tries to balance the start and the end SoC for every simulation, to not let the difference in SoC influence the fuel consumption results. The vehicle should behave in a predictable way for the driver, which is important both for safety and comfort reasons. The vehicle response on the acceleration and brake pedal should not be dependent on factors outside the driver’s knowledge. The HEV should re- spond similarly to the driver demand, as in a conventional vehicle, since the driver is used to a conventional vehicle. The implementation of the driver is mainly a PI-controller, which tracks the velocity profile available in the cycle information data. A conventional vehicle has been implemented, which only uses the ICE as a power source. This is done to make the comparison easier between the two HMSs, because the conventional can be used as a benchmark. 34 4.8. SUMMARY Chapter 5 Simulation with Rectified Layout In this chapter the delivery truck is simulated with the two HMSs on different cycles, and the results are studied. Firstly the original rule-based strategy is analyzed, in order to see how it is performing, and possible improvements are implemented. Both the SML and the TD are tuned to give the rule-based strategy a good performance. Secondly the improved rule-based strategy is compared to the table-based strategy, and the result is discussed. The chapter ends with a summary part. 5.1 Improving the Original Rule-Based Strategy The specifications for the vehicle used in the first simulations are given in Table 5.1. It represents a rather large battery and a fully loaded truck. Table 5.1: Vehicle specifications for the first simulations. Li-Ion Battery Capacity: 27 Ah Vehicle Total mass: 7100 kg Hybridization Electric Drive with ICE on and Hybrid Drive Gear shift logic Speed-based When the original set up of the rule-based strategy is simulated, some unwanted effects are observed, which need to be treated. The two most important are listed below, together with their respective solutions. 5.1.1 Calculation Error when ICE Torque Decrease and EM Boost Occurred Simultaneously Problem As explained in Section 3.2.2, the TD has four functions that computes a part of the EM torque, which are then added together resulting in the total EM torque. Since the outputs from these four functions are added, it is important that they do not interfere with each other, as by doing so will ruin the purpose of the separate functions. It is dis- covered that the two functions EM Boost and ICE Torque Decrease interfere with each other, when the driver demands a torque over the maximum torque of the ICE. In the left part of Figure 5.1, a working point is specified above the maximum torque line. Since the ICE can not provide this torque, the EM Boost function will decrease the working point to the τmax line. If the ICE Torque Decrease function is enabled simultaneously, it lowers the working point to the specified ICE Torque Decrease line. The sum of these 35 36 5.1. IMPROVING THE ORIGINAL RULE-BASED STRATEGY two results in a working point which is too low, and not the one intended by the ICE Torque Decrease. Solution This is solved by limiting the effective region of the ICE Torque Decrease, to torques lower than τmax only, hence making sure that they do not occur at the same time. The result is shown in the right part of Figure 5.1. 80 100 120 140 160 40 100 Speed [rad/s] N or m al iz ed to rq ue [% N m ] Incorrect version τ max ICE Torque Decrease line EM Boost contribution ICE Torque Decrease contribution New working point (incorrect) Old working point Total decrease 80 100 120 140 160 40 100 Speed [rad/s] N or m al iz ed to rq ue [% N m ] Correct version τ max ICE Torque Decrease line New working point (correct) EM Boost contribution ICE Torque Decrease contribution Old working point Total decrease Figure 5.1: The incorrect version of the EM Boost and the ICE Torque Decrease interaction (left), and the correct version (right). 5.1.2 State of Charge Controllers Synchronization Issues Problem The TD has one system for controlling the SoC, and the SML another one. This means that these two parts can have different opinions of how the SoC should be affected. The SML calculates how many seconds the EM can provide the demanded torque, if kept constant, until the battery would reach its lower al