Parametrization of Lithium-Ion Battery Cell Model and Test Rig Development for BMS Application

Examensarbete för masterexamen
Master's Thesis
Övrigt, MSc
Dolk, Victor
Rhedin, Pontus
Abstract The rising demand for electric vehicles has driven significant advancements in lithium ion battery technology. A critical function of the Battery Management System (BMS), central to electric vehicle performance, is the estimation of the State of Charge (SOC). Kalman filtering is a widely used method for SOC estimation. Additionally, estimating battery degradation over time, known as the State of Health (SOH), is vital for understanding how the battery will age. Another important aspect of the BMS is to estimate the State of Power (SOP) to limit the battery’s allowable currents to prevent damage. Developing an accurate battery model requires measurement data from a real battery cell, and since lithium ion battery cells are temperature dependent, these measurements must be conducted across a wide range of temperatures to capture this dependency. This thesis focuses on the development of a measurement rig, enabling the collection of data necessary for the parameterization of the battery model. This model is then used to estimate SOC, SOH, and SOP. This is conducted for two different lithium ion cells with the same cell chemistry. The simulation results show that the dual Resistance Capacitance (RC) Thévenin Equivalent Circuit Model (ECM) with thermal dependency is beneficial due to the trade off between low model complexity and high accuracy. The parameters used for the model were based on multiple measurements conducted for eight different temperatures. The measurement methods were chosen and evaluated with time efficiency and high accuracy in mind. The parameters were obtained by using different curve fitting methods with respect to open circuit voltage OCV, SOC and temperature. The terminal voltage measurement from the measurement equipment was used to verify model and parameterization accuracy where the mean modeling error for the blue cell is 0.1050% while median error for the grey cell is 0.1317%. Both simulation and parameterization output were satisfactory. The SOC results on real data sets show that Kalman filtering is a robust and beneficial method since the estimated SOC for both filters were within the 3% SOC error bounds. The Unscented Kalman Filter (UKF) performs better in the presence of high uncertainties and nonlinearities compared to the extended Kalman filter (EKF). However, both filters performed similarly due to easy nonlinearities and common nominal conditions. Therefore the operating conditions will decide which estimation algorithm to be used. The implemented SOP algorithm estimates the maximum and minimum current to be drawn at every time instance with respect to SOC and voltage limits. The algorithm also provides highly accurate power estimations for 5, 10, and 30 seconds into the future. The implemented SOH algorithm based on cycling measurement shows a clear degeneration trend for both parameters it has been based on. The SOH estimation is a function of changes in OCV and capacity with respect to a number of cycles.
Keywords: Battery, Lithium, ECM, SOC, SOH, Estimation, SOP, Kalman, Measurement, Simulation, UKF, EKF.
Arkitekt (konstruktör)
Geografisk plats
Byggnad (typ)
Teknik / material