A Simplified Short Term Load Forecasting Method Based on Sequential Patterns

Load forecasting is an essential part of a power system both for planning and daily operation purposes. As far as the latter is concerned, short term load forecasting has been broadly used at the transmission level. However, recent technological advancements and legislation have facilitated the installation of smart meters in many utilities. Thus, historical residential consumption data are now present which were not available until recently. Moreover, the gradual incorporation of new loads, such as Heat Pumps and Electric Vehicles, will necessitate congestion prognosis at various distribution locations so as to activate demand side management. As a result, it is expected that load forecasting will be implemented at more disaggregated levels than the transmission, possibly reaching the secondary distribution transformer point. Traditional forecasting techniques, although being rather accurate, require considerable expertise for model construction and re-construction. Consequently, they might be impractical to use in case multiple regional forecasts are to be conducted. In this perspective, a simplified hour-ahead load forecasting algorithm was created so as to provide an automated approach to the problem as an alternative to other established forecasting techniques. This algorithm is based on sequential patterns and, hence, the continuous data are discretized in order to compare recent to past patterns. Although some error due to discretization is introduced, the method performs adequately well in comparison with an ARIMA model