Distribution transformers short-term load forecasting models

Abstract

The increasing penetration of variable energy sources, plug-in electric vehicles (PEV) and storage in the distribution system generates the need for load forecasting at a more granular level than at the substations. Load forecasting at the distribution transformer level can provide an estimate of how the load is distributed in the distribution network rather than a substation total estimate. It opens possibilities for various applications and use cases including better demand-response management, resource scheduling, enhanced losses minimization, and more accurate management of distribution transformer loading. In order to achieve higher accuracy of transformer loading forecasting, distribution transformer meters or equivalent aggregation of the transformer served load based on smart meters is required. This paper presents two linear models for computationally efficient short-term load forecasting on distribution transformers. The first model uses an aggregate load and the second uses an average load approach. The average load method exhibits prediction accuracy superior to the aggregate load approach. The proposed models are tested on smart grid community data provided by Pecan Street.