Very short-term load forecast for demand side management in absence of historical data

Abstract

Load forecasting plays an important role in today’s electricity grid, due to the presence of distributed generation. In this paper, various methods for short-term load forecasting are presented with the purpose to serve as tool to operate within a demand side management system. The need for large amount of historical data is avoided, opening the possibility for immediate use by customers without recorded load history. Widely distributed, such system can increase the reliability of the modern grid. The results showed that the Holt-Winters forecasting procedure give better results in comparison to neural networks with gradient descent, principally if the load rises fast in the early morning hours.