Research on short-term power load machine learning prediction model based on hyperparameter optimization

With the smart grid becoming an increasingly important part of the power grid, accurate prediction of short-term power loads has become indispensable. This paper aims to improve the accuracy of the short-term power load forecasting model through hyperparameter optimization to accurately predict short-term power loads and maintain the efficient operation of a power system. This paper proposes a power load prediction model based on a variety of common machine learning algorithms (such as KNN, Multiple linear regression (MLR), Decision Tree (DT), and Random Forest(RF)). At the same time, the differential evolution algorithm, whale optimization algorithm, ant colony optimization algorithm, and bat optimization algorithm are used to optimize the model's hyperparameters, and the prediction results are compared to improve the prediction accuracy. The research results show that the prediction results of the data set of region 2 selected in this paper are superior to those of region 1. Compared with the unoptimized random forest model, the random forest model prediction performance based on bat optimization has significantly improved prediction accuracy. The decision coefficients (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>) of average2, max2 and min2 were 0.8172, 0.8043 and 0.8341, respectively. This study expands the short-term power load forecasting method and provides references for similar research. Future studies can further consider other external factors or explore more possible improvement directions.