Machine Learning and Modified Cheetah Optimizer-Based Energy Management System for Cost-Effective Microgrid Operation

The use of renewable energy sources (RESs) in modern electric grids poses considerable challenges regarding the balance of supply and demand and the cost-effective operation of the grid. In this paper we present the decentralized energy management system (EMS) to reduce operational costs, to cope with uncertainties of the renewable generation and load demand. The solution involves employing a machine-learning-based approach, the multilayer perceptron artificial neural network (MLP-ANN), to develop precise forecasts of photovoltaic and wind turbine generation, ambient temperature, and load demand. Of the two tested training algorithms (Levenberg-Marquardt (LM) and resilient backpropagation (RP), LM had the best predictive performance. The newly proposed Modified Cheetah Optimizer (MCO) algorithm eliminates the common local trapping and premature convergence issue possessed by the traditional optimization methods to optimize generation scheduling. The integrated EMS based on MCO achieves an impressive reduction of 26.8% in operational cost in comparison with traditional methods. In addition, the adoption of a DR program leads to a peak load decrease by 7.5% and a valley load filling increase by 9.6%. Through simulation results, we find that the proposed framework improves the forecast accuracy up to 15%, resulting in a more resilient and economic microgrid management strategy. The impacts of matching machine learning forecasting with enhanced optimization methods and solving methodologies may turn into a breakthrough in microgrid operational optimizations and economic advantages, as the work emphasizes.