Metaheuristic Optimization and Machine Learning-Based Unit Commitment Strategies in Smart Grids

This study introduces Machine Learning and Metaheuristic-Based Unit Commitment (ML-MetaUC), which is a hybrid method for intelligent UC in smart grids. The ML component uses supervised models such as linear regression and random forest on historical and real-time environmental data to predict the upcoming demands for electricity. The Metaheuristic Optimization (MO) layer uses several methods, including Differential Evolution (DE) and Ant Colony Optimization (ACO), to ascertain the appropriate time to activate the generators. Simulation on the IEEE 14-bus test system showed that ME-MetaUC reduced operating expenses by 10.7% and boosted convergence speed compared to traditional approaches. In addition, the proposed framework has a high degree of flexibility to various load scenarios, contributing to increased system dependability. Under conditions of uncertainty, the ML-MetaUC framework, a data-driven and scalable solution for smart grid energy management, makes it possible to schedule producing units robustly and more effectively.