A Fine-gained Prediction Algorithm Based on the Feature Matching for Electricity Usage Demand Forecast

Abstract

We study the prediction problem for electricity usage demand, which is crucial for the stable operation of smart grid and the improvement of the users’ experience. Many prediction algorithms for electricity usage demand have been proposed to improve the prediction accuracy of the total electricity consumption in a region. However, most existing methods focus on the long- and medium-term electricity usage demand, and fail to predict the short-term fine-gained electricity usage demand. In this paper, we propose a fine-gained prediction algorithm based on the feature matching for electricity usage demand on smart meters. This paper extracts the electricity usage feature of each user from electricity usage data recorded by the smart meter, and conducts feature matching on the real-time electricity usage data. Finally, the matched features are used to predict electricity consumption. Experiment shows that our model achieves well performance on the fine-gained electricity usage data.