A SHORT-TERM ELECTRICITY FORECASTING APPROACH BASED ON FEATURE CONSTRUCTION AND CAE-LSTM

The line loss rate can reflect the management level and economic benefits of the enterprise. The supply and sale of electricity in different periods will cause errors in the line loss statistics, so short-term electricity forecasting is needed. To solve the problems that the existing approaches cannot fully mine the factors affecting the electricity, a short-term electricity forecast approach based on feature construction and CAE-LSTM is proposed. Features were constructed by data analysis, and MIC was employed for screening. ARIMA was employed to forecast the electricity value and new data was reconstructed by features. CAE-LSTM was applied to extract the features of the data and get the predicted result. Experimental results show that the proposed approach can extract data features more effectively and achieve higher prediction accuracy.