Multi-objective home energy management system with realistic load classification and multi-source uncertainty modeling: A robust optimization approach

Abstract

Home Energy Management Systems (HEMS) play a vital role in improving residential electricity consumption by coordinating appliances with distributed energy resources and the grid. However, effective HEMS scheduling is challenged by uncertainties in photovoltaic (PV) generation, electricity pricing, and appliance operation durations, all of which influence cost and comfort. This paper presents a day-ahead robust multi-objective optimization framework for HEMS that jointly minimizes energy cost and user discomfort while addressing multi-source uncertainties. Shiftable appliances are modeled through a realistic classification distinguishing interruptible loads—with deterministic and uncertain durations—and non-interruptible loads—with fixed or variable power profiles. Robust optimization is applied to handle PV, tariff, and operation-duration uncertainties, an aspect rarely explored in prior work. The proposed mixed-integer linear programming framework integrates PV systems, energy storage systems, and electric vehicles, all supporting bidirectional power exchange. User comfort is represented via thermal and time-based components, with a novel normalization method ensuring consistent evaluation across appliances. Validation through eight simulation scenarios confirms the framework’s ability to manage uncertainties and align with user preferences, providing robust scheduling plans that effectively balance economic efficiency and comfort. • Developed a day-ahead robust multi-objective HEMS optimization model for residential energy use. • Proposed realistic load classification covering duration uncertainty and power profiles. • Addressed uncertainties in PV output, electricity prices, and appliance operation times. • Integrated PV, ESS, and EV systems into a unified MILP framework with bidirectional flow. • Introduced a novel normalization method for evaluating time-based user discomfort.