Impact of Converter Size Asymmetry on Power Transfer in Railway Power Systems

Abstract

This thesis investigated the impact of converter station capacity asymmetry on power transfer for different system impedances. The scope of this thesis was to investigate the significance of the adjustable static converter parameters, in order to minimize active power losses in the railway power system. This thesis only considers the static frequency converter (SFC), and since an SFC solely consists of power electronics and is modeled to mimic the behaviors of a rotary frequency converter (RFC) through software, the parameters are adjustable. This means that it was possible to adjust the quadrature reactances for instance, which would not be possible if the RFC were modeled. The Y-bus matrices and the load profiles were made in Matlab whilst the optimization and load flow models were constructed in the General Algebraic Modeling System (GAMS), which was used to optimize the system with the objective function to minimize active power losses. Various cases were simulated where the converter parameters were optimized for both the autotransformer (AT) - and booster transformer (BT)-systems to understand the system for representative cases. The models were used for studying realistic cases as well as cases with large asymmetry in terms of power capacity and different load locations between substations. This was done to shed further light regarding the philosophy of local production, whether it should be prioritized or not. The results show, for both the AT- and BT-system, that if the load is located in the middle of the line, it is optimal for the substations to share the load as evenly as possible. If the load is located closer to one substation than the other, then the system should prioritize local production regardless of the substation capacities. If there is a large asymmetry between the substations, in terms of installed total power capacity, it is more optimal to control the load-sharing with the no-load angle than other parameters. Optimization of the no-load angle reduced power losses for all cases, particularly when the substations had highly asymmetric power capacities. However, it was shown that the no-load voltage should always be kept as high as possible. The results also show that the quadrature reactances should be as low as possible in order to minimize losses, this is possible since the SFC quadrature reactances can be set in the software controlling the power electronics. In regards to the active- and reactive power droops, it was shown that with optimization they were not able to reduce the losses significantly. In contrast, it was shown and discussed that the active power droop, when solely optimized, was able to reduce the voltage drops and decrease the overall voltage variations. However, during the full optimized cases, the droop factor did not give such significant change since it was probably de-prioritized. It was also discussed that the reactive power droop got very small for the BT-system when the no-load voltage also were optimized, which probably is a result of the high no-load voltage and limited load model. In order to reach clarification regarding the droop constants, a load model with higher complexity would be needed.