Application of neural network based variable fractional order PID controllers for load frequency control in isolated microgrids

A dependable and highly effective load frequency control (LFC) is a critical requirement of modern power systems, and it necessitates special consideration for the proper operation of microgrids (MGs). It can be even more critical for the isolated MGs, which are remotely operated in rural areas, due to the increased penetration of renewable energy sources (RESs) including fuel cells (FC), solar and wind energy, and integrated energy storage systems (ESSs). As a result of technological advancements in control and computation tools, it is crucial to develop new solutions for the LFC system and ensure the power quality of these hybrid MGs (HMGs). This chapter addresses the application of (fractional-order) proportional-integral-derivative ((FO)PID) controllers with variable gains and orders to frequency deviation control (FDC). In a coordinated control strategy between ESSs and RESs, our objective is to design an FDC system based on the variable PID-type controllers which are tuned by a neural network (NN) in a multiagent structure. To achieve this, a reinforcement learning (RL) technique is applied to train the NN-based tuners. This automatically tuned NN-based variable (FO)PID (V(FO)PID) controller associated with RL enhances the performance of isolated HMGs under different levels of RES penetration and load disturbances. By capturing all tuning knobs of the V(FO)PID controller in a self-tuning technique, the proposed approach not only can provide a desirable performance related to the LFC problems such as recovery time or frequency stability but also can be used over a wide range of operating conditions due to its flexibility in configuring the integration and derivative actions toward improving the frequency stabilization performance.