基于EEMD和随机森林的月度负荷预测

Monthly Load Forecasting Based on EEMD and Random Forest

准确的负荷预测是电力市场稳定运行的关键。2017年实施偏差电量考核给售电公司带来了极大的挑战。可再生能源和新能源的接入需要高精度的负荷预测。利用集合经验模态分解(EEMD)算法将全社会用电增速序列分解为6个子序列,将子序列组合成高、中、低频分量序列,再对中、低频分量用随机森林(RF)法选取最优参数构建模型,将分量预测结果相加重构成最终预测结果。并与RF、支持向量机(SVM)和EEMD-SVM的实验误差进行了对比,结果表明本文所构建模型的预测精度要优于对比模型,同时验证了该方法在月度负荷预测方面的有效性和可行性。

Accurate load forecasting is the key to the stable operation of power market. The implementation of deviation power assessment in 2017 has brought great challenges to the power selling company. Renewable energy and new energy access requires load forecasting with high accuracy. A set of ensemble empirical mode decomposition （EEMD） algorithm is used to decompose the whole social electricity growth sequence into 6 subsequences, and then the subsequences are combined into high, medium and low frequency components. Then the optimal parameters of the medium and low frequency components are selected by the optimal parameters of the random forest （RF）. Finally, the final prediction results are formed by the aggravation of the component prediction results. The experimental errors of random forest （RF）, support vector machine （SVM） and ensemble empirical mode decomposition support vector machine （EEMD-SVM）are compared. The example results show that the prediction accuracy of the model is higher than that of the contrast model, and the validity and feasibility of the method in monthly load forecasting is verified.