基于Copula理论与KPCA-GRNN结合的区域综合能源系统多元负荷短期预测模型

Short-term multiple load prediction model for regional integrated energy system based on Copula theory and KPCA-GRNN

通过Copula理论对多元负荷之间以及多元负荷与天气因素之间的非线性相关性进行分析。基于分析结果构建模型样本集,通过核主成分分析(KPCA)对样本集进行降维解耦,采用广义回归神经网络(GRNN)模型对多元负荷进行预测,模型参数采用遗传算法进行优化。以北方某建筑级区域综合能源系统夏季多元负荷中的电负荷和冷负荷算例为例,与其他两种预测模型进行对比,验证结果表明,基于Copula理论与KPCA-GRNN结合的区域综合能源系统多元负荷短期预测模型具有较高的预测精度,该模型具有一定的应用价值。

The nonlinear correlation between multiple loads and between multiple loads and weather factors is analyzed by Copula theory. Based on the analysis results,the model sample set is constructed. The sample set is decomposed by KPCA. The GRNN model is used to predict the multiple load and the model parameters are optimized by genetic algorithm. Taking the calculation example of electrical load and cooling load in a building-grade regional integrated energy system in north China as an example,compared with the other two prediction models,the verification results show that the model in this paper has higher prediction accuracy and better application value.