A Weekly Load Data Mining Approach Based on Hidden Markov Model

With the development of advanced metering infrastructure, massive smart meter readings are generated and stored in smart grids, which makes it possible for detecting of tremendous social value embedded in load data. The majority of the existing load data mining works are performed on the daily time scale without adequate consideration of load information between the days. To better describe the power consumption characteristics of users, a data mining approach based on the weekly load curves is proposed in this paper. First, the piecewise aggregate approximation technique is utilized to reduce the dimensions of the raw weekly load data. Then, a Davies-Bouldin index-based adaptive k-means algorithm is proposed to cluster the studied users into several groups. Finally, a hidden Markov model describing the probabilistic transitions of different load levels is established for each cluster to extract the representative dynamic weekly load features. A feasible tool based on dynamic characteristics of load patterns is invented to evaluate the short-term load forecasting methods, which realizes the pre-check for the forecasting results without future real measurements in the forecasting horizon. Case studies on a real dataset demonstrate that the proposed method is capable of extracting weekly load characteristics of users.