Suppression Study For Low Frequency Oscillation Of Electrical Quantities In Vehicle-Grid System Based On The Model Predictive Control

With the rapid development of high-speed railway,the scale of railway network is continually expanding in China.Besides,as more and more high-speed electric multiple units(EMUs)and high-power electric locomotives have been put into operation,a vehicle-grid coupling system with complex interactions is formed.So far,the low-frequency oscillation(LFO)phenomenon of vehicle-grid coupling system has appeared in many places,which directly influences the safety and stability of the railway system operation.Therefore,it is necessary to carry out the study on its mechanism and corresponding suppression measures.Due to the control of grid-side converter is the key factor that affects LFO,and considering that the model predictive control has significant advantages in converter control,a multi-variable optimal controller based on state estimation which can realize the state feedback linearization is designed first for the purpose of LFO suppression.To better deal with the nonlinear characteristics of pulse width modulation(PWM)rectifier in EMUs and highlight the advantages of model predictive control,then,a decoupling predictive current control and a model predictive direct power control are realized.To improve the robustness of model predictive control,a robust model predictive control is achieved by constructing a disturbance estimator.Through the simulations,analyses and comparisons,the improvement in control performance and LFO suppression capability of the proposed methods are evaluated.Firstly,the equivalent circuit of the PWM rectifier is obtained by the modeling of the traction drive system in the EMUs.Then,the state space model of PWM rectifier is constructed and simplified by analyzing the equivalent circuit.Based on the simplified model,the detailed realization process is carried out:construct the set value calculation module to calculate the set values of the states;Design linear quadratic optimal controller to achieve the optimized feedback of the states;Deduce the closed-loop state observer to estimate the state quantities.Finally,a multivariable optimal controller based on state estimation is completed.The stability of the method is verified by the eigenvalue analysis in the closed-loop state space model of PWM rectifier,which involves the number of vehicles and the equivalent impedance of the traction network when multiple vehicles access traction network.Compared with the traditional transient direct current control(TDCC),the Matlab/Simulink simulation and RT-LAB platform validation show that the proposed method has better control performance and LFO suppression effect.Then,in order to better deal with the nonlinear characteristics of the PWM rectifier in EMUs and make the input current and power well tracked,combining the finite-control-set characteristic of the PWM rectifier,a decoupling current prediction model and a power prediction model are deduced based on the dynamic characteristic equation in the AC side of the PWM rectifier.By adopting the error of d and q current components with their set values as cost function,the decoupling predictive current control is realized.Similarly,the model predictive direct power control is also achieved by adopting the error of active and reactive powers with their reference values as cost function.The optimal control value is obtained by minimizing the cost functions,and then output.The control performance of the two methods is compared by Matlab/Simulink simulation,and the model predictive direct power control is further verified in RT-LAB platform.Finally,in order to improve the robustness of model predictive control,the uncertainty of the circuit parameters is taken into account when constructing the PWM rectifier prediction model.Meanwhile,the quadratic sum of the error of the d and q current components with their references is taken as the cost function.By constructing the state space equation of the disturbance estimator,the closed-loop eigenvalues are deduced and the distribution maps of the eigenvalues are drawn to complete the calculation and selection of the key parameters.Finally,the robust model predictive control based on the disturbance estimator is realized.The current prediction values are calculated from the improved prediction model,and the control voltage variation is obtained by minimizing the cost function.The control voltage at present interval can be obtained by summing the variation with the control voltage in the last interval.Compared with the traditional TDCC and previously designed methods,the dynamic,the steady state performance and LFO suppression effect of the proposed method are demonstrated by simulations.