Short term load forecasting with a hybrid clustering algorithm and pattern recognition

Load forecasting is an important part of the operational procedure of a power system and enormous research effort has been placed on the development of accurate prediction methodologies. The focus is on the hybrid models developed with a clustering algorithm to group data with common characteristics, and a function approximation to capture the underlying characteristics of each cluster of data. The developed clustering scheme generates clusters whose common property is that they are described by the same linear model. A demonstration of the proposed methodology is performed for the one step ahead forecasting of load data from the Californian and the New England power systems, and exhibited a reduction on the forecasting error by more than 7% compared to a conventional neural network methodology.