Short-term Day-to-day Maximum Load Forecasting based on Data Mining and Deep Learning

Abstract

With the rapid development of energy Internet renewable energy, more uncertainty and randomness are added to power system. Accurate load forecasting will improve power generation plans and the utilization of renewable energy. A dayto-day maximum load forecasting method is proposed which uses Gaussian mixture model (GMM) clustering based on multidimensional scaling (MDS) and support vector machine (SVM) after empirical mode decomposition (EMD). MDS is used to reduce the daily load data from 24 to 2 dimensions. Different daily load data obtained by GMM clustering are formed into daily maximum load curves. EMD method is used to decompose load curves into a series of relatively stable component sequences, and then the SVM model establishes prediction models for the decomposed components. Finally, the predicted components are summed to obtain the predicted value of the maximum load. The calculation results show that the prediction accuracy of the prediction model proposed is about 3% higher than that of the traditional prediction model.