Power Big Data–Based Combination Forecast Method for Medium-Long Term Power

Abstract

The power grid will generate massive data in the process of power generation, transmission, transformation and power consumption. It is directly related to the development of the power grid to make use of power big data to forecast power system in the medium and long term. In the past, many researchers used a single model to predict power. However, in the context of big data, power grid data is generated in many different environments and times, and no single prediction model can guarantee satisfactory prediction results for the big data generated at any time and in any environment. To improve the accuracy of power prediction, a medium and long-term power prediction method based on LSTM, gray scale prediction model and LSSVM is proposed in this paper. The proposed combination method first calculates the weights of the prediction results of different models by means of the variance-covariance method, and then fusions the final prediction results by weighting. To verify the effectiveness of the combined model, a medium- and long-term power prediction experiment is carried out. The experimental result shows that the prediction accuracy of the proposed combined prediction model is higher than the prediction of each single prediction model in accuracy.