Home Energy Management System Optimization with Stochastic Programming

Abstract

This thesis explores the optimization of Home Energy Management Systems (HEMS) with a focus on battery electric vehicle (BEV) charging, using two different methods: Linear Programming (LP) and Stochastic Programming. The BEV is modeled with vehicle-to-everything (V2X) functionality, allowing it to both charge and discharge energy to support the household or grid. The primary aim is to incorporate uncertainty into the optimization process, acknowledging that in real-world scenarios, exact information about the vehicle’s arrival time and current state of energy (SOE) is rarely known in advance. We begin by examining uncertainty in both arrival time and SOE, then narrow the focus to uncertainty in SOE alone. In the final part of the thesis, we investigate the impact of power tariffs to determine whether incorporating them into the optimization yields meaningful benefits for households. Our results show that stochastic optimization introduces a small increase in computational cost while significantly reducing the risk of unwanted outcomes caused by uncertainty. While the cost difference per vehicle is fractional, the impact becomes substantial when scaled across many BEVs. The benefits extend beyond cost savings, by optimizing charging to occur during low spot price hours, we also reduce the environmental footprint, as electricity is typically greener during these periods. In addition, the ability to sell back energy to the grid when it is not needed helps balance demand and supports grid stability. Therefore, assuming perfect foresight and relying solely on LP is not representative of real-world conditions. We conclude that accounting for uncertainty through stochastic methods is a cost-effective and scalable improvement for future energy management systems.