Using agent based simulation and model predictive control to study energy consumption behavior under dynamic pricing

Abstract

In the interest of increasing energy efficiency and avoiding higher generation costs during peak periods, utility companies adopt various demand response (DR) methods to achieve load leveling or peak reduction. DR techniques influence consumer behavior via incentives and cause them to shift peak loads to off-peak periods. In this paper we study the energy consumption behavior of residents in response to a variable real-time pricing function. We consider thermostatic loads, specifically air conditioning, as the primary load and apply the model predictive control (MPC) method to study the behavior of consumers who make consumption decisions based on a trade-off between energy cost and thermal comfort. An agent-based simulation is used to model a population where each household is an agent embedded with the MPC algorithm. Each household is associated with a multi-attribute utility function, and is uniquely defined via the use of stochastic parameters in the utility function. INTRODUCTIONResidential electricity consumption behavior is by nature unpredictable and thus raises much interest in demand response (DR). In power economics literature, DR has long been proposed for incentivizing consumers to change their energy consumption behavior in achieving load leveling. Energy efficiency in a grid network can be achieved if the system load can be accurately predicted and balanced. DR tries to change the energy consumption behavior of consumers by providing them with financial incentives and education, encouraging them to use less energy during peak hours and more energy during off-peak hours in an attempt to level the system load. However, most DR programs provide financial incentives and assume that consumer behavior is driven primarily by cost. Fahrioglu and Alvarado (2000) applied game theoretical principles to study the interaction between the utility company and its customers. They obtained load relief during peak times by designing incentive compatible contracts that used nonlinear cost functions. Mohsenian-Rad et al. (2010) discussed the use of a distributed algorithm on smart meters to find optimal consumption schedules for subscribers. They achieved peak load reduction by using a pricing scheme based on non-linear cost functions and game theory analysis. Samadi et al. (2010) proposed a real-time pricing algorithm based on utility maximization. In our research, we study consumer behavior as a function of their comfort as well as cost incentives, because it is unrealistic to assume that all people value these incentives equally. The Annual Energy Outlook 2012 report from the US Energy Information Administration (EIA 2012) indicated that residential customers contributed about 37% of the total energy used in 2011. Thermostatically controlled loads (TCL) make up about 45% of the total residential energy use; 23% is attributed to air conditioning alone. Therefore, in this paper we focus mainly on the consumer behavior of 288 978-1-4799-7486-3/14/$31.00 ©2014 IEEE