Bus load forecasting via a combination of machine learning algorithms

Aim of this work is to develop a novel hybrid model for the day-ahead prediction of a distribution substation feeder load. The model comprises an unsupervised machine learning stage, where a clustering of daily load curves takes place, and an Artificial Neural Network (ANN), based on the clustering output, which performs the prediction. The model is tested on the day-ahead prediction of a complete year, where the load corresponds to a bus that refers to a sub-urban area in Northern Greece. The proposed model is compared with the one that has been developed for the Greek interconnected system. Experimental results indicate that the model leads to increased accuracy and is able to simulate the high nonlinearities of the bus load.