Forecasting the power consumption of a single domestic electric water heater for a direct load control program

Thermal storage units such as residential electric hot waters are feasible candidates to be used in a direct load control programs. They can store electric power in form of heat for later use. PowerShift Atlantic (PSA) is a leading project in developing a demand-side management program to provide up to 32MW reserve capacity through electric water heaters. To control water heaters properly, the controller needs to have an estimation of the on/off status of each individual water heaters in advance. This paper presents a monte-carlo statistical based method to create a short-term load forecast of the individual loads. This method was compared with a traditional neural network based load forecast. This paper also presents a model to estimate the relative error of the aggregated load forecast. The proposed methods were deployed on the PSA pilot and the experimental results are discussed in this paper.