A neural network based very short term load forecaster for the interim ISO New England electricity market system

This paper presents the development and implementation of an artificial neural network (ANN) based very short-term load forecasting (VSTLF) model for the interim electricity market of ISO New England (ISO-NE). The main outcome of the forecaster is the 5-minute forecast of New England internal system demand that will be used directly by the 5-min real-time resource dispatch function in the existing spot market. The design of the ANN structure, the selection of the training sets, raw data pre-processing, the training process itself as well as validation and testing are discussed in detail. The ANN model has been tested under a wide variety of conditions and the results of the study demonstrate a high forecast accuracy. An off-line training system based on back-propagation algorithm is developed to support the training and retraining of the ANN of the VSTLF model. A real-time VSTLF application is developed and integrated into ISO-NE's energy management system (EMS).