Research On Fractional Hammerstein Model And Model Predictive Control Of PEMFC

Proton exchange membrane fuel cell(PEMFC)is a new energy power generation device,which has the advantages of low temperature start-up,fast response,low noise,etc.,and has a good development prospect.Affected by multiple variables in actual work,there is a non-linear phenomenon in the power generation process,and the fractional order is also a characteristic when it is in the working state.The prerequisite for efficient battery development and utilization is to establish an accurate electrical characteristic model and design a suitable controller.This paper adopts the subspace identification method,introduces the theory of fractional calculus,establishes a Fractional Order Hammerstein State Space(FOHSS)model based on an improved optimization algorithm,and proposes a Rolling Window-based Subspace Model Predictive Control(RWSMPC)strategy can not only maintain the normal operation of the battery but also improve the battery power generation performance.The main research contents of this paper are as follows:(1)Combined with multivariate input and output data,the fractional order and nonlinear characteristics of the battery are analyzed,and the FOHSS model of the battery is established.Analyze its working principle and characteristics,and use mutual information method to select input and output variables.A method of fractional subspace identification based on variable step length and short memory is proposed to reduce the calculation amount of fractional differential while ensuring calculation accuracy.The N-information Criterion(NIC)criterion is introduced to determine the order of the system.The simulation results show that the FOHSS model based on the improved identification algorithm based on the input and output data can describe the actual input and output relationship.(2)Aiming at the unknown parameters in the identification algorithm,an Adaptive Differential Evolution-Black Hole(ADE-BH)algorithm is proposed.The concept of black hole is introduced into the DE algorithm,the operating parameters are adaptively operated,and the Euclidean distance is introduced as the theoretical basis for updating the target vector.The ADE-BH evolutionary algorithm is introduced into the identification algorithm,and the simulation results show that after the ADE-BH evolutionary algorithm is used to optimize the unknown parameters in the identification process,the established FOHSS model describes the electrical characteristics of PEMFC more accurately.(3)Based on the established FOHSS model of the battery,the RWSMPC strategy is designed.According to the input and output data matrix,the subspace predictor equation is derived and introduced into the objective function.The Alternating Direction Multiplier Method(ADMM)is used to solve the optimal control rate of the SMPC with constraints.In view of the time-varying characteristics of the battery,a rolling window is introduced to update the subspace predictor.In order to prevent the loss of effective information,the determination coefficient is used to determine whether the model needs to be updated.The simulation results show that in the case of model mismatch,the effect of tracking and controlling the electrical characteristics of PEMFC based on the RWSMPC strategy is better.