A Weather-Induced Power Load Forecasting Method Based on Asymmetric Single Valley Curves

During the same time period, temperature is the most important weather factor affecting power load. However, there is not a simple linear relationship between temperature and power load. Furthermore, power load changes are associated with weather conditions such as humidity, wind speed, and radiation, and the influence of these factors varies across different regions. Taking these characteristics into consideration, seven climate comfort indices are introduced that are temperature-driven while also considering humidity, radiation, and wind speed factors. A four-parameter asymmetric single valley curve with continuous first derivatives is constructed to quantify the relationship between power load and weather conditions. Nonlinear curve fitting is performed using the improved Levenberg-Marquardt-Fletcher (LMF) algorithm, and initial values are set based on the locally estimated scatterplot smoothing method (LOESS) to enhance computational efficiency. Experimental results in Guangdong Province demonstrate the stability of this method in predicting daily power load variations during different seasons. Moreover, the model performs well in predicting peak load under extreme weather conditions.