关键词: 短期电力负荷预测, 预测精度, 经验模态分解, Hurst指数, 支持向量回归机, 贝叶斯优化算法, 组合预测模型, 误差评价

Abstract:

Short-term power load is the key to the balance of power supply and demand, in view of the short-term power load prediction accuracy problem, the EMD(Empirical Mode Decomposition)-Bayes-SVR(Support Vector Regression) combination prediction model is proposed, that is the original power load sequence is decomposed into several IMF(Intrinsic Mode Function) and a Res(Residual) by EMD method, and each IMF is reconstructed into high frequency components, low frequency components and residual components according to the Hurst index, and the parameters optimization of SVR are optimized by Bayesian optimization algorithm. The optimal parameters obtained by the optimization are brought into the SVR and the reconstructed three components are predicted separately, and the predicted values of the three components are added together to obtain the final prediction result. Taking the historical power load data of Nebraska in the United States as an example, eight single prediction models and seven combined prediction models are established as reference models to predict the power load series in this area. Experimental results show that the combined EMD-Bayes-SVR prediction model can effectively predict the change trend of short-term power load, and the error evaluation indexes of MAE(Mean Absolute Error), RMSE(Root Mean Square Error) and MAPE(Mean Absolute Percentage Error) are decreased by 29.84%, 32.05% and 22%, respectively when compared with the SVR model, which are significantly lower than other reference models.

Key words: short-term load forecasting, prediction accuracy, empirical mode decomposition, Hurst index, support vector regression, Bayesian optimization algorithm, combined forecasting model, error evaluation