An Enhanced NN-based Load Frequency Control Design of MGs: A Fractional order Modeling Method

Abstract

Developing accurate mathematical models for microgrid (MG) components is the initial step before implementing various load frequency control (LFC) strategies and analysis. In this regard, different high-order models associated with different nonlinearities have been included to increase the modeling accuracy resulted in a performance improvement in the LFC techniques. Nevertheless, these high-order nonlinear models pose some potential problems such as obstacles in the analytical description of the system and control problem along with its high computational complexity. In this light, the fractional order based models are deployed to effectively balance the model accuracy and analytical complexity. First, two fractional order components (energy storage system and fuel cell) are arranged in a controlled coordinated strategy to enhance the frequency stability. Then, two artificial neural network (ANN) controllers are deployed for each components in a multi-agent framework. To accomplish this step, a multi-agent stochastic reinforcement learning optimization is applied to train the two controllers. Test results on an isolated MG with fractional components validate the efficacy of the coordinated LFC strategy.