考虑频域分解后数据特征的空调负荷预测模型

Cooling load forecasting model considering data characteristics after frequency domain decomposition

中文关键词  负荷预测  变分模态分解  最小二乘支持向量机  极端梯度提升  随机误差分析

英文关键词  load forecasting  variational mode decomposition  least-square support vector machine  eXtremeGradient boosting  random error analysis

中文摘要

      针对空调负荷预测实际应用中容易存在数据散杂且可用信息匮乏的问题, 从负荷序列的非线性、非平稳性 和随机性出发, 提出了一种基于变分模态分解(VMD)的负荷预测方法. 对不同数据特征序列考虑不同算法的数据观 测与训练原理差异, 充分发挥各个模型优势. 首先采用随机森林(RF)进行特征选择, 利用VMD将负荷序列按趋势分 量、平稳分量和噪声分量进行分类重构, 并分别对非线性序列建立最小二乘支持向量机(LSSVM)预测模型, 时序平 稳序列建立极端梯度提升(XGBoost)预测模型, 采用正态分布拟合随机误差, 得到各子序列预测结果并进行叠加输 出最终负荷预测结果. 实验结果表明, 所提方法能准确反映负荷的特性并具有更好的预测精度, 能有效预测空调负 荷, 为空调节能优化控制策略提供依据.

英文摘要

      Aimed at the problems of scattered data and lack of available information in practical engineering application of air conditioning cooling load forecasting, a load forecasting method based on Variational Mode Decomposition (VMD) is proposed according to the nonlinear, non-stationary and random characteristics of the cooling load sequence. Considering the differences of data observation and training principles of different algorithms for sequences showing different data characteristics, the advantages of each model are fully utilized. Firstly, Random Forest (RF) is used for feature selection. Then, VMD is applied to decompose cooling load series into the trend part, the stationary part and the noise part. Least- Square Support Vector Machine (LSSVM) model is built for the nonlinear part, eXtremeGradient Boosting (XGBoost) model is built for the linear part, and probability distribution is fitted for the noise part, following which the ultimate cooling load forecast could be obtained by accumulating prediction values from each sub-series. The experimental results show that the proposed method can comprehensively capture the characteristics of the original cooling load series and displays superior capacity for cooling load forecasting, providing a basis for the optimal control strategy of air conditioning energy saving.