Feature Extraction Method of Electric Power User-side Metering Data Based on Text Classification

Nowadays, many companies begin to take measures such as rational and planned electricity consumption and off-peak electricity consumption. How to make targeted load forecasting for users is the most concerned issue. In this paper, the load characteristic curves of different types of users are obtained by text classification, and the similarity between the curves and the load characteristic curves is compared and analyzed. In this paper, the load shape index of typical daily load curves of users is calculated and counted by using five indicators, such as load rate, daily peak-valley difference, peak-period load rate, average load rate and valley load rate. Compared with different algorithms, the accuracy of text classification algorithm is 93%, and the running time is 0.19s s. The accuracy of decision tree algorithm is 90%, and the running time is 0.26s s. In this paper, the feature extraction method of power user-side metering data based on text classification is helpful to the monitoring of power theft or abnormal power load.