Research on Power Trading Decision Support System based on GMM and Light GBM

With the continuous growth of renewable energy in the power system, accurate prediction of wind power becomes an important issue in the auxiliary decision-making system of power trading, but the uncertainty of wind power generation brings great challenges to the stable operation of the power system. Accurate wind power prediction is helpful for the efficient development of wind power energy. Aiming at the problem of insufficient data mining in the face of complex wind power data, an ultra-short-term wind power prediction algorithm based on Gaussian mixture model (GMM), ridge regression algorithm and light gradient lift machine (LightGBM) is proposed. The proposed algorithm is specifically designed for application in the auxiliary decision-making system for power trading. The method uses GMM to cluster the features, and then uses the ridge regression algorithm to predict each cluster after clustering. Considering the limitations of a single prediction algorithm in the face of highly volatile wind power data, the prediction value after the ridge regression algorithm and the data after feature importance screening together constitute the final data set and transmit it to the LightGBM algorithm to obtain the final prediction result. Compared with widely used prediction algorithms such as CNN, SVM, and LightGBM, this method has higher stability and prediction accuracy. By applying this methodology in the auxiliary decision-making system for power trading, the proposed algorithm enables effective short-term wind power generation prediction and provides reliable decision-making support for the electricity market. Its application contributes to the improvement of operational efficiency in power systems and facilitates the sustainable utilization of renewable energy resources.