Modelling Of High-speed Railway Traction Power Supply System Based On Subspace Identification

With the notable development of high-speed railway industry,traditional traction transformers in high-speed trains would gradually be replaced by the high-power vehicular power electronic transformers(PET)for the sake of lightweight and energy saving.To ensure the reliability of PET under different operating circumstances,it’s crucial that its control design guarantees solid robustness.Therefore,full investigations regarding the train-network coupling mechanism and factors affecting the system’s uncertainty are undoubtedly deemed preliminaries.Nowadays,however,the majority of proposed stability criteria are merely valid in regards to specific train-network structuring and parameters,and most studies concerning the coupling stability adopt simplified models for the network-side,which can’t eventually demonstrate the impact of location changes,faults and other factors on the entire system.A more comprehensive mathematical model of long-distance,multi-conductor traction power supply system is required for a deep dive into coupling and robust stability analyses.Firstly,two mechanism-based modelling methods of traction power supply system are illustrated: time-domain simulation model of chain circuits and topology-based state-space model of modularized structure.Following the elaboration of their principles and algorithms,a typical traction power supply arm is taken as example,with its simulation model and topology-based state-space model constructed.Advantages and limitation of the two models for train-network coupling analyses are respectively elucidated in this case.Secondly,a new identification modelling method of traction power supply system is proposed.Theories of subspace identification method with application of geometric projection are expounded,and hence the uniformed framework for three classical algorithms is deduced.Then the equivalent reduced mathematical model of traction power supply system is established from input-output of specific identification experiments.The evaluation criteria for identification results take into consideration the reasonable trade-off of accuracy and simplification.Quantitative descriptions of relationship between order decision and model precision verify the reliability of identification results in terms of reflecting the original system’s characteristics in both time domain and frequency domains.Last but not least,with identification tests conducted under different operation situation,including different train locations along with various fault types,precision of identification models in different orders has been discussed,from which a uniformed order decision is concluded.Consequently,uncertainty from these factors can be attributed as parameter variation without order changes.Thus the uncertainty of traction power supply system can be measured by gap metric,providing useful reference for further measurements of train-network uncertainty and even the related fault-tolerant control of PET.