Probabilistic estimation of the potentials of intervention-based demand side energy management

Successful enrollment of customers in intervention-based demand side energy management (DSM) programs, such as energy efficiency and installation of PV panels, depends on having accurate estimates of the benefits of these programs available and communicated to the customers. The program benefits may include long-Term financial savings, and their contribution to managing supply and demand in transition to a sustainable grid. Among them, the most needed measure of benefit is the estimated energy saving for each individual after an intervention. Accurate estimates of energy savings for each individual customer are thus crucial for ensuring high program enrollment. In this paper, we formulate the problem of estimating energy savings to understand the potential benefits of enrolling customer in an intervention-based DSM program. Due to highly uncertain customer load and estimation of long-Term (infinite time horizon) savings, traditional deterministic analysis approaches, such as load forecasting, will yield poor results. We propose a Gaussian Process (GP)-based approach capable of capturing uncertainty and adapting arbitrary data length for infinite time horizon estimation. This allows utilization of probability estimation to show customers their potential energy savings and resultant revenues for participating in an intervention program for a certain period of time. Such property is verified by highly accurate estimation results running on a set of customer AMI data obtained from Pacific Gas and Electric Company. The simulation results not only highlight the feasibility of the intervention concept, but also provide benefit potentials that could be used to persuade customers to enroll in energy efficiency programs.