An approximate dynamic programming based home energy management system and its impact on the distribution system

Abstract

The residential sector consumes more than 20 percent of electricity in the United States. In order to shift peak load, a utility company provides time-of-use rates or even real-time pricing to its customers. Typically, peak electricity rates can be up to twice as much as the off-peak rates. Therefore, electricity costs can be different even when the customer is using the same amount of energy. A home energy management system (HEMS) can provide autonomous control of residential appliances to minimize the electricity cost while maintaining the customer's thermal comfort in the presence of uncertain weather and electricity consumption. In this thesis, a HEMS is developed to optimize forward-looking schedules for a residential home's heating, ventilation, air conditioning (HVAC), water heater (WH), electric vehicle (EV), and battery storage (ES). An approximate dynamic programming (ADP) approach is used to define, model, and solve the HEMS problem. Moreover, a model predictive control (MPC) based simulation framework is developed to study the impact of the HEMS on the distribution system and compare the ADP with another prevailing algorithm - mixed integer linear programming (MILP). Extensive simulation has been carried out and simulation results show that the proposed ADP-based HEMS can effectively schedule appliances to minimize the electricity cost and maintain customer thermal comfort. Furthermore, a coordinated two-stage real-time market mechanism in an unbalanced distribution system is proposed to fully utilize flexibility services provided by the HEMS for alleviating line congestion, voltage violation, and substation level power imbalance.