PREDICTION PERFORMANCES OF ELECTRICITY DEMAND: MODIFYING INPUTS THROUGH CLUSTERING

Accurate prediction of electricity demand can bring extensive benefits to any country as the forecast values help the relevant authorities to take decisions regarding electricity generation, transmission and distribution much appropriately. The literature reveals that, when compared to conventional time series techniques, the improved artificial intelligent approaches provide better prediction accuracies. However, the accuracy of predictions using intelligent approaches like neural networks are strongly influenced by the correct selection of inputs and the number of neuro-forecasters used for prediction. This research shows how a cluster analysis performed to group similar day types, could contribute towards selecting a better set of neuro-forecasters in neural networks. Daily total electricity demands for five years were considered for the analysis and each date was assigned to one of the thirteen day-types, in a Sri Lankan context. As a stochastic trend could be seen over the years, prior to performing the k-means clustering, the trend was removed by taking the first difference of the series. Three different clusters were found using Silhouette plots, and thus three neuro-forecasters were used for predictions. This paper illustrates the proposed modified neural network procedure using electricity demand data