A Distributed Two-Layer Frequency Compensation for Islanded Microgrids Based on Q-learning and PI Controllers

Frequency instability generates significant challenges to the stability of the system. To solve the frequency deviation problem, the traditional secondary control uses PID controller to achieve frequency compensation for the primary control, but simultaneously the traditional PID controller has disadvantages such as poor dynamic performance and the need for manual tuning of parameters. The problems mentioned above will lead to poor compensation accuracy. To address such issue, this paper proposes a new frequency compensation scheme that divides the traditional frequency secondary control into two layers, the first layer uses an improved PID controller that considers the average value of output frequency of all distributed generators, and the second layer is based on Q-learning technology to compensate again. The proposed scheme shortens the response time and improves the control accuracy, and effectiveness is verified in MATLAB/Simulink.