Research On Current Control Of Single-phase LCL Grid-connected Inverter Based On ADP

As an important part of the energy feedback system,the grid-connected inverter directly affects the output power quality of the grid-connected system.The inverter control strategy proposed in this paper can effectively suppress the harmonic components of the grid-connected current,improve the stability of the grid-connected system,help alleviate energy shortages,and develop renewable energy.The proposed algorithm has important social and economic significance.This article mainly studies the following:(1)Firstly,the development status of grid-connected inverters is introduced,and the topology of grid-connected inverters is classified.In order to solve the problem of poor adaptability and single control characteristics of existing inverter control strategies,approximate dynamic programming will be proposed.The ADP)method is applied to the field of inverter control.The research progress of the approximate dynamic programming(ADP)method is briefly introduced.The application of ADP algorithm in grid-connected inverter control is explained in the literature,and the practical significance of the control algorithm made in this paper is explained.(2)Introduce the principles of dynamic programming(DP)and approximate dynamic programming(ADP),classify ADP algorithms according to different evaluation network design methods,and introduce six typical ADP algorithm control schemes,according to the evaluation network input port.The control scheme is divided into an Action-Dependent(AD)control method and a non-execution dependent control method with or without control input;it is classified into Heuristic(HDP)according to whether the output port contains a derivative value or a partial value derivative.,Double Heuristic(DHP)and Hybrid(GDHP)control schemes.However,the above typical ADP control scheme must first design and optimize the evaluation network module,and there is no uniform method for the evaluation network design,and the training of the module increases the computing burden of the system.In order to simplify the ADP control scheme,this paper changes the approximate method of approximate dynamic programming,converts the infinite series of dynamic programming into finite series,collects multiple inverter output current values to calculate performance indicators,thus avoiding evaluation.The training process of the network module.(3)Design the LCL filter parameters,establish the approximate discretization system model and hybrid system model of the grid-connected inverter system,and identify the parameter matrix of the hybrid model,and build the grid-connected inverter according to the identified hybrid model.Model network.The parameter identification algorithm uses a particle swarm algorithm.Search algorithm algorithms such as particle swarm optimization have strong global optimization ability in low-dimensional space,and the requirements for fitness function are not high.The traditional least squares algorithm can only predict the state value of the next moment.The fitness function can only use the linear function,and the particle swarm algorithm can delay the prediction time and build a highly sensitive nonlinear fitness function.Improve the accuracy of the identification parameters.However,the traditional particle swarm optimization algorithm not only includes positional parameters,but also includes velocity parameters.The particle identification efficiency is low.In order to reduce the order of the particle swarm optimization algorithm,an improved particle swarm optimization algorithm with only positional parameters is proposed to improve the identification efficiency and reduce the efficiency.The algorithm parameters are used.For the problem that the traditional particle swarm algorithm is easy to fall into premature convergence and the global search ability is poor,the improved algorithm introduces random parameters obeying the exponential distribution to ensure that the particles always have a certain global search ability and use the improved particle swarm optimization algorithm.The hybrid model of the inverter is identified,and the model network module of the ADP method is built.(4)Design the action network module of ADP algorithm,and use the optimization algorithm to train the action module;firstly,design the topology structure of the action network module,and design the completed action network controller combined with the evaluation module and model module to build The sequence neural network of the ADP control scheme optimizes the motion network controller using the BPPT training algorithm.At the end of the paper,the simulation experiment platform of the system is built in simulink,and the neural network controller based on ADP control method is compared with PI control.The simulation experiment sets various working conditions under different sampling frequencies,different system parameters,grid voltage distortion,etc.The result proves that the controller based on the approximate dynamic programming algorithm can effectively suppress the grid-connected current harmonics,effectively improve the reliability of the grid-connected inverter and improve the output power quality of the inverter.