Recurrent Neural Network Driven Load Forecasting to Enhance Disturbance Mitigation in frequency regulation of power system

The increasing integration of renewable energy sources and the unpredictable nature of modern power systems pose significant challenges for maintaining effective load frequency control (LFC). Conventional LFC methods struggle to manage the fast-changing dynamics and variability in load demand, leading to frequency instability. This paper addresses these challenges by introducing a novel solution that leverages machine learning (ML) techniques, specifically recurrent neural networks (RNN), to forecast load demand in control areas. The forecasted load is then fed into a three-degree-of-freedom proportional-integral-derivative (3DOF-PID) controller, which enhances the system’s ability to regulate frequency. The proposed controller is tested on a two-area, four-machine power system under step load changes, demonstrating significant improvements in dynamic response. Additionally, real-time load data from the Delhi power grid is used to validate the controller’s effectiveness in handling realistic load variations. This work underscores the potential of ML-based LFC strategies to improve stability and efficiency in modern power systems.