Research On Digital Twin Modeling Method Of Urban Rail Train For System Power Flow State Deduction

Power flow is an important feature reflecting the steady-state operation of urban rail power supply system,its dynamic change is usually deduced and analyzed based on the model of train and power supply network.The establishment of high-precision train model is the key premise to ensure the accuracy of system power flow calculation.It aims to calculate the power and motion position of the train at each time accurately as the input condition for power flow calculation of power supply network.However,at present,the train modeling for power flow deduction usually adopts ideal driving strategy,solidified empirical parameters and rated passenger load,which leads to the deviation between the calculation results and the real situation.Aiming at the problem of low consistency of traditional train modeling,based on the advanced concept of actual state feedback and information physical fusion in digital twin,a high fidelity digital twin modeling method of urban rail train for system power flow state deduction was proposed in this thesis,which effectively improves the accuracy of the train model and provides an accurate and effective support means for power flow analysis and optimization of traction power supply system.Firstly,the electrical and kinematic characteristics of the train was deeply studied in this thesis,and the system equivalent model of urban rail power supply network and train for power flow deduction was established.Focusing on the accuracy loss of traditional train modeling caused by single particle simplification and fixed driving strategy,a train kinematics model based on multi-particle force analysis and actual speed curve input was constructed,which improves the consistency between the model and the actual train in operation mode.Then,aiming at the error accumulation problem of train model caused by solidifying empirical resistance parameters,a fidelity evolution method of train model parameters based on swarm intelligence optimization algorithm was proposed.Through the closed-loop tracking and comparison of measured data and simulation data under digital twin system,the parameters were inversed and optimized,and the on-line correction of model resistance parameters was realized.The effectiveness of the proposed method was verified according to the actual subway line data,which showed that the parameter correction had a good effect on improving the accuracy of the train model.Finally,aiming at the calculation deviation of the train model operation state caused by the rated load setting,a passenger load modeling method of urban rail train based on similar day extraction and deep neural network was proposed.Through the quantification of the environmental factors,the correlation analysis of daily feature similarity and deep regression training,a train passenger load model reflecting the influence of actual environmental factors was constructed,which realizes the setting of train load conditions for accurate deduction.It had been verified that the passenger load model proposed in this thesis effectively solves the calculation error caused by the rated load setting of the traditional train modeling,and ensures the accuracy of the deduction of the train running state.