Adaptive Control Strategy for Inertia and Damping of Virtual Synchronous Generator Based on CNN-LSTM Data-Driven Model

Abstract

With the application of distributed power sources, the stability of the power system has been dramatically affected. Therefore, scholars have proposed the concept of a virtual synchronous generator (VSG). However, after the system is disturbed, how to make it respond quickly and effectively to maintain the stability of the system becomes a complex problem.To address this problem, a frequency prediction component is incorporated into the control module of the VSG to enhance its performance. The Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) model is used for frequency prediction, ensuring that the maximum energy capacity released by the storage system is maintained. Additionally, it guarantees that the inverter's output power does not exceed its rated capacity, based on the predicted frequency limit after the system experiences a disturbance. The advantage of real-time adjustment of inverter parameters is that the setting intervals for inertia and damping can be increased. The selection criteria for inertia and damping can be derived from the power angle oscillation curve of the synchronous generator. Consequently, an adaptive control strategy for VSG parameters is implemented to enhance the system's frequency restoration following disturbances. The validity and effectiveness of the model are verified through simulations in Matlab/Simulink.