PROJECT TITLE :

Industrial Power Load Forecasting Method Based on Reinforcement Learning and PSO-LSSVM

ABSTRACT:

It is very difficult to obtain high-performance industrial power load forecasting as a result of the many different complex factors that influence it. In-depth research into industrial power load forecasting using a combination of machine learning and other methods is being conducted for industrial enterprise power consumers. As a consequence of this, a fresh approach to power load forecasting is suggested by taking into account the different load characteristics that are typical of various geographical areas, types of businesses, and patterns of production. In the first step of this process, the improved K-means clustering analysis is used to sort the historical load data according to the production patterns to which they belong. After that, the prediction algorithm is proposed, which combines reinforcement learning with particle swarm optimization and the least-squares support vector machine. After all of that has been processed, the load data are split up into their respective patterns, and finally, the improved algorithm that was presented in this article is used for short-term load forecasting. The method of forecasting that is presented in this piece is data-driven and makes use of actual datasets. The findings of the experiment with the simulation show that the improved prediction algorithm is able to recognize changes in various production patterns and identify the load characteristics of various regions and industries with a high degree of prediction accuracy, which has value for practical application.