Improved Power System Load Forecasting Using Information on Climate and Social Events

Abstract

Power system load forecasting is crucial for power system planning, operation, and control, which reduces operational costs and improves economic efficiency. However, the current forecasting techniques, including LSTM and ARIMA models, ignore the influence of important factors like weather conditions, public holidays, and social events on power system load, which may give rise to inaccurate prediction results. To mitigate this issue, the present work makes use of the Mann-Kendall mutation detection algorithm to detect abrupt changes in power system load caused by the factors mentioned above. A correction function is then developed to improve the prediction accuracy of a conventional prediction model like ARIMA. The experimental results validate the effectiveness of the proposed approach.