Multi-Objective Load Scheduling for Home Energy Management Considering the Thermal Behavior of the Building and User Comfort

Thermostatically controlled appliances are major electricity consumers in residential buildings, yet conventional home energy management systems (HEMSs) often model them as fixed-power loads, neglecting thermal dynamics. This paper presents a multi-objective HEMS that integrates the NSGA-III evolutionary algorithm with physics-based thermal models, data-driven meteorological forecasting, and electric vehicle smart charging (unidirectional, grid-to-vehicle) considering battery degradation. Air-conditioning demand is estimated using a resistance—the capacitance thermal model of the room, while tankless water-heater demand is calculated from the required temperature rise. Both models are driven by weather forecasts generated using a stacking ensemble of three deep learning models. User preferences are represented through a comfort matrix, while rooftop photovoltaic (PV) generation and hourly grid carbon emission factors are incorporated. The HEMS simultaneously minimizes daily electricity cost and ${\mathrm{CO}}_2$ emissions while maximizing user comfort over a 24 h horizon under two time-of-use tariff schemes. The results show that under the Brazilian White Tariff, the knee-point solution reduces the daily electricity cost by 27.1% and the emissions by 28.3% with respect to a non-optimized household while keeping 90.2% of the theoretical maximum comfort; within the Pareto front, the minimum-cost solution costs 22.2% less than the maximum-comfort schedule. The minimum-emission solution reduces emissions to about 1.09 kgCO₂eq/day, and PV self-consumption rises from 50.7% to 94.0%, while grid import in the peak tariff block falls by 67%.