Modeling Demand Response Using Utility Theory and Model Predictive Control

Today’s power generation and distribution systems are designed to accommodate peak and not average demand. Due to uncertainty of weather and users’ convenience-driven consumption behavior, electricity demand varies significantly over a daily cycle, thus the waste of a significant portion of the system’s capacity. On the other hand, demand response has long been proposed to incentivizing consumers to change their energy consumption behavior in achieving load leveling. This research is motivated by two facts. First, thermostatically controlled loads (TCL) are the primary contributors of residential energy consumption. Second, residents’ decisions on energy consumption are based on not only cost minimization but convenience maximization. Therefore, we develop an agent-based simulation model, in which each agent (e.g., household) makes periodical decisions over a daily cycle on selecting optimal set points for thermostatically controlled appliances to maximize his/her utility. The latter consists of electricity cost and thermal comfort, and is developed using multi-attribute utility functions. Further, model predictive control (MPC) is used to calculate the actual thermal dynamics over times during a day. This framework allows for consumers to have varying tradeoffs between ‘comfort’ and ‘cost.’ Two pricing structures, time-of-use and dynamic pricing, are studied.