Change Detection in Electricity Consumption Patterns Utilizing Adaptive Information Theoretic Algorithms

The high-resolution data on electricity consumption, recorded by smart meters at customers' premises, are valuable sources of operational information and consumption patterns. In addition, customers' characterization plays an undeniable role in the implementation of demand response (DR) programs, as any changes in the consumption patterns could affect DR programs. Therefore, an accurate algorithm for detecting changes in the consumption patterns is very useful in not only the implementation of DR but also in other fields, such as load forecasting and peak shaving. This article proposes a reliable procedure for detecting changes in customers' consumption patterns. For this reason, an adaptive algorithm is introduced to improve the clustering quality of customers' consumption patterns by determining the optimum number of clusters, using a locally weighted entropy-based segmentation. Moreover, considering the customers' consumption records in different time slots as the features, another adaptive algorithm is introduced for feature selection based on the mutual information concept. The proposed method is evaluated by applying a real dataset provided by the Irish Social Science Data Archive. The results corroborate the efficiency of the proposed procedure.