Research On Robust Model Predictive Control Method Of LCL-type Grid-connected Inverter

For the sake of enhancing the traditional power electronic energy supply structure,building an environment-friendly power electronic energy supply system,the renewable energy-distributed generation system(RE-DPGS)has been vigorously developed.As the bridge of electric energy conversion between power generation unit and power grid,it is necessary to study the robustness of grid-connected converters(GCIs).In this paper,the LCL GCIs are taken as the control plant,and implement the finite control set-model predictive control(FCS-MPC)method.The major research content contains two aspects.On the one hand it is to upgrade the performance of LCL GCIs when the filter parameters are perturbed,which is the platform-model parameters are mismatched.On the other hand,it is to go deep into discussing the above issue and consider the method to improve the robustness of the FCS-MPC controller in the case of parameter mismatch in the noise environment.This paper first analyzes the causes of filter parameter mismatch,it majors due to the abnormal temperature rises and aging of filter elements.Then,analyze the influence of parameter mismatch on the grid side current prediction error.By constructing a mathematical expression,the relationship between parameter error and grid-side current prediction error is derived.Aiming at the issue that parameter mismatch impacts the quality of grid side current,this paper provides two online filter parameter identification methods to adaptively update the model parameters and make the model parameters fit the LCL GCIs platform.The first method is the state feedback parameter identification method.The principle of this method is to establish the filter parameter estimation model in discrete time according to the state value relationship of LCL GCIs,so as to realize the online identification of filter inductance and capacitance parameters.The second method is the root mean square gradient descent optimization(RMSprop-GDO)parameter identification method.The method defines the error vector according to the errors between the sampling state vector and the predicted state vector,and takes the 2 norms of the error vector as the target function,the gradient direction of the objective function is taken as the change direction of the parameter matrix,through the self-tuning algorithm to continuously iterative until the error between the sampling state value and the predicted state value is minimized.At this time,the filter model parameters in the prediction model are the real working parameters,so as to realize the online high accurate identification of the filter inductance and capacitance parameters of the GDO method.Then,choose the RMSprop algorithm to dynamically optimize and adjust the learning rate of the GDO algorithm,and improve the identification speed of the GDO parameter identification algorithm.Finally,the effectiveness of the proposed parameter identification method is tested by the three-phase LCL GCIs hardware platform.A fuzzy Kalman filter-model predictive control method is proposed to solve the problem that the sampling data of model predictive control is too much and the quality of grid side current is affected under noise environment.This method only needs to sample grid voltage and grid side current.Compared with Kalman state estimation,it mainly improves the accuracy of state estimation under time-varying noise environment.Kalman state estimation is an online state estimation algorithm with grid voltage,gridside current and inverter output voltage as input and machine-side current,grid-side current and capacitor voltage as output.Finally,the online estimated state value is fed into the model predictive control.The Kalman state estimation can be divided into three parts,namely prediction,update and correction,and the state estimation is iteratively obtained at each time.Then,the measurement noise covariance matrix R in the Kalman state estimation is adjusted online by the fuzzy algorithm to realize the fuzzy Kalman state estimation algorithm with adaptive noise regulation.Finally,the effectiveness of the fuzzy Kalman state estimation algorithm is verified on the hardware platform of the threephase LCL grid-connected inverter.