Optimizing residential energy usage patterns in smart grids using hybrid metaheuristic techniques

Introduction. This study applies hybrid metaheuristic optimization techniques to intelligently schedule household loads, ensuring a balance between cost reduction, comfort and grid stability in smart homes. Problem. The growing gap between energy demand and supply leads to high electricity costs, increased appliance waiting times, a higher peak-to-average ratio (PAR) and reduced user comfort. Efficient management of residential energy consumption remains a major challenge for sustainable smart grid operation. Goal. This study aims to minimize electricity costs, reduce PAR and enhance user comfort by optimally scheduling household appliances and shifting loads from peak hours to off-peak hours. Methodology. A demand-side management approach is implemented using 5 metaheuristic optimization algorithms: harmony search algorithm (HSA), flower pollination algorithm (FPA), hybrid harmony flower pollination algorithm (HFPA), multiverse optimization algorithm (MVO) and cuckoo search algorithm (CSA). Real-time pricing is employed as the pricing model. MATLAB simulations were conducted for 10, 30 and 50 smart homes, each comprising 15 residential loads categorized as controllable or base appliances. Results. Simulation results demonstrate that the proposed HFPA consistently outperforms HSA, FPA, MVO and CSA across all tested scenarios, achieving notable reductions in electricity cost and PAR while minimizing appliance waiting times. Scientific novelty. The hybrid HFPA effectively combines the strengths of HSA and FPA, balancing exploration and exploitation to deliver superior performance in multi-objective optimization for home energy management systems. Practical value. The proposed HFPA achieved up to 19.86 % reduction in electricity cost and 81.03 % minimization in PAR, significantly enhancing user comfort and operational efficiency. The method can be further extended for integration with renewable energy sources and machine learning-based predictive control systems. References 32, tables 6, figures 5.