Load Forecasting Based on Weather Forecast Correction and Cumulative Temperature Index

Traditional load forecasting methods have often overlooked inaccuracies in weather forecasts, especially those caused by sudden weather changes and the cumulative effects of prolonged high temperatures, leading to insufficient prediction accuracy under extreme climate conditions. To address this, the paper proposes a load forecasting method that incorporates weather forecast corrections and temperature accumulation indicators. First, weather forecast accuracy is improved by analyzing historical and actual weather data using an adjustment error method. Second, a temperature correction model is developed to account for the cumulative impact of sustained high temperatures on load growth. Finally, the load is predicted using a BP neural network algorithm, which integrates historical load data, historical meteorological data, and corrected weather forecasts. Test results show a prediction error of only 1.61%, demonstrating the robustness and accuracy of the proposed method.