Short-Term Load Forecasting using a Cluster of Neural Networks for the Greek Energy Market

In the context of the liberalization of the Greek Energy Market, load forecasting is essential in various system programming procedures. Short-term load forecasting extends from one to seven days, although in this paper a model is proposed for the next calendar day in step of sixty minutes. The objective is to design and implement a software-based short-term load forecasting model for the Greek interconnected transmission system that will show improved performance compared to previous methods. The proposed model introduces a categorization of the forecasted days along with a dedicated artificial neural network for each category. Appropriate input vectors are selected at the training process for each custom-built network.