基于灰色关联分析与随机森林回归模型的短期负荷预测

Short-Term Load Forecasting Based on Grey Correlation Analysis and Random Forest Regression Model

针对人工神经网络和支持向量机存在的泛化误差大、具有局部最优以及参数选取困难等缺点,将随机森林回归模型引入电力系统短期负荷预测,提出了一种基于相似日与随机森林回归模型的短期负荷预测方法。利用灰色关联分析法计算原始训练样本与预测日各影响因素间的关联系数,选取相似度较高的历史样本构成相似日样本集,对随机森林回归模型进行训练。将预测日的特征向量输入训练好的模型中,取所有回归树输出结果的平均值作为最终的负荷预测结果。实际算例表明,与常规支持向量机法和常规随机森林回归法相比,该组合方法可以有效地提高短期负荷预测的精度。

In view of the defects in artificial neural network and support vector machine, such as large generalization error, the existence of local optimization and the difficulty of selecting optimal parameter, a short-term load forecasting method based similar days and random forest regression models is proposed. Grey correlation analysis method was used to calculate the correlation coefficient of each influencing factor between the original training samples and forecasting day factor. And then the historical samples with high similarity were chosen to form the sample set of the similar days and train the random forest regression model. The eigenvectors of the forecasting day were put into the trained model, and the average outputs of all regression trees were taken as the final load forecasting. Compared with the conventional support vector machines and conventional random forest regression, the result shows that the combined method can effectively improve the accuracy of the short-term load forecasting.