Prosumer-centric stochastic home energy management system with quantile scenario modeling

This paper presents a quantile-scenario-based stochastic Home Energy Management System (HEMS) integrating rooftop photovoltaic generation, wind turbines, battery energy storage, and electric vehicles within a unified multi-objective optimization framework. The proposed model simultaneously minimizes electricity costs and the peak-to-average ratio (PAR) while maximizing user comfort by optimising the scheduling of household appliances, battery charging/discharging, and EV charging/discharging. A novel Scenario-Selective Decoding (SSD) operator is introduced to address the over-hedging issue in conventional stochastic HEMS by decoupling appliance scheduling from battery and EV dispatch, thereby reflecting their distinct sensitivities to renewable uncertainty. Joint solar-wind scenarios are generated directly from probabilistic day-ahead quantile forecasts, avoiding computationally intensive Monte Carlo simulations. The framework further incorporates Conditional Value-at-Risk (CVaR)-based chance-constrained optimization, where grid import and ramp-rate violations are embedded as risk-aware penalties within a Tchebycheff scalarisation scheme. The Walrus Optimization Algorithm is employed and benchmarked against metaheuristic algorithms using Wilcoxon rank-sum tests. Results under cloudy conditions demonstrate enhanced robustness, achieving a 0.61% reduction in CVaR5% and a 28.3% reduction in expected peak-to-average ratio, and improved operational reliability compared with a deterministic baseline. The findings demonstrate that the proposed SSD-based stochastic HEMS provides a computationally efficient and risk-aware scheduling framework for uncertainty-resilient prosumer energy management.