Implementation of neural networks for forecasting cardinal points on the electricity demand curve

One of the tasks of the UK National Grid Company plc (NGC) is to forecast demand for electricity. In recent years, neural networks (NNs) have been increasingly used as a tool to perform short term forecasting of half-hourly demand in megawatts (MW). This paper describes the forecasting requirements of NGC, and an implementation of NNs for forecasting cardinal points of demand, rather than half-hourly values. It discusses the process for creating an NN model and the problems encountered. Areas of potential improvement are suggested