Real-Time Transient Optimization of Grid-Forming Inverters via Cascaded Neural Network-Driven Damping Coefficient

Grid-forming inverters (GFMIs) are crucial in enabling power grids with high penetration of renewable energy sources. However, replacing synchronous generators with inverter-based distributed energy resources introduces stability challenges, primarily due to reduced system inertia. One significant challenge is the dynamic response of GFMIs during disturbances and anomalies, such as changes in line impedance. This paper proposes a decentralized control strategy that integrates a feed-forward neural network (FNN) and a cascaded feed-forward neural network (CFNN) to enhance the dynamic response of GFMIs during line impedance disturbances. The first FNN estimates the grid’s line impedance based on the active power, frequency, and rate of frequency change. At the same time, the cascaded neural network predicts the optimal damping coefficient in the primary controller to improve the transient behavior of GFMIs. The effectiveness of the proposed approach is validated through multiple case studies utilizing MATLAB/Simulink.