Circulating Current Suppression And Frequency Control Method Of Parallel Inverters In Microgrid Based On Reinforcement Q-Learning

In recent years,new energy sources such as solar energy and wind energy have been connected to the grid on a large scale in the form of distributed power sources to provide electrical energy to the grid.In distributed photovoltaic power generation,inverters need to be used to convert DC power to AC power and transmit it to the grid or AC load.When the microgrid operates independently,since there is no large grid to provide voltage and frequency compensation for the microgrid,it is necessary to control the parallel connection of the microgrid Inverter to ensure the stability of voltage and frequency in the microgrid.However,in the operation of parallel inverters,there are factors such as inconsistent line impedance with the common bus,changes in power device parameters,and sudden load changes,which will cause the microgrid system to generate circulating current,the inverter output power cannot be evenly divided,and the operating frequency of the microgrid fluctuates and other issues.Aiming at the parallel inverter system in the microgrid,this paper combines the Reinforcement Q-Learning algorithm with droop control method and virtual complex impedance method,and carries out research on circulating current suppression,power sharing and frequency control between parallel inverters.Firstly,for the parallel circulation of inverters,based on the droop control method,the three-phase voltage inverter is modeled,the power module of the droop controller,the voltage and current double closed-loop module are designed,and the parallel system of two inverters is built.Designing a simulation experiment to test the suppression effect of the droop control method on the circulating current of the inverter parallel system when the line impedance is the same and the line impedance is different.The experimental results show that the droop control method has good circulation current suppression effect on inverter parallel systems with the same line impedance,but it has poor circulation current suppression effects on inverter parallel systems with different line impedances.Secondly,in view of the poor circulation suppression effect and the inability to divide the output reactive power of parallel inverters in the droop control method under different line impedances,this paper proposes a virtual complex impedance method,which introduces virtual complex impedance in the droop control method and realizes the compensation of the inverter output line impedance.Then designing a simulation experiment to verify the feasibility of the method.Considering that the line impedance parameters are unknowable in practical applications,and the premise of the application of the virtual complex impedance method is the known line impedance parameters,this paper combines the Reinforcement Q-Learning algorithm with the virtual complex impedance method to achieve the compensation of the inverter output line impedance,and achieve the effect of parallel inverter circulation suppression and power sharing.Simulation experiments prove the effectiveness of this method.Finally,aiming at the frequency control problem of microgrid,this paper proposes a frequency control method based on Reinforcement Q-learning algorithm.In a parallel inverter system using droop control,the frequency will be affected by load fluctuations.In this paper,the Q-Learning algorithm is combined with the droop control method to compensate the rated output active power of the parallel inverter and adjust the frequency of the microgrid to realize the control of the frequency of the microgrid.Finally,the results of the simulation experiment show that the method has better frequency control effect.