Reinforcement Learning for Load Frequency Control of Renewable Integrated Systems

The participation of renewable energy resources in modern power systems has been increasing due to their positive impacts on diversifying power supplies with environmental benefits. Despite these advantages, the system's vulnerability to frequency deviation increases due to the low inertia and intermittent characteristics of these sources. This paper proposes a simple Integral controller in which the frequency control parameter is tuned by reinforcement learning using a Deep Deterministic Policy Gradient model under changing inertia conditions due to the effects of renewable energy resources. The system's frequency will return to its nominal value faster by the improvement of secondary control. The simulation results of the proposed method show significant improvement in the system's frequency control under constantly changing inertia and load. The proposed method has a significant potential to improve the system's robustness under constantly changing inertia and load conditions.