为增强综合能源系统负荷精细化分解水平,充分利用误差信息以进一步提升预测性能,提出一种基于聚合混合模态分解和时序卷积神经网络(temporal convolutional network,TCN)的综合能源系统负荷修正预测框架。首先,采用改进完全集合经验模...为增强综合能源系统负荷精细化分解水平,充分利用误差信息以进一步提升预测性能,提出一种基于聚合混合模态分解和时序卷积神经网络(temporal convolutional network,TCN)的综合能源系统负荷修正预测框架。首先,采用改进完全集合经验模态分解对电、冷和热负荷初步分解处理,随后利用变分模态分解对具有强复杂性的子序列进一步分解。然后,依据最大信息系数(maximum information coefficient,MIC)分析多元负荷的耦合特性并通过多元相空间重构(multivariate phase space reconstruction,MPSR)丰富特征信息。最后,构建基于TCN的修正预测模型。以校园综合能源系统算例对比不同预测模型,结果显示所提修正预测框架的电、冷和热负荷预测均具有较低的平均绝对百分比误差,有效解决了预测中模态分解的模态混叠以及模态高频分量问题,实现预测误差修正。展开更多
Under the support of National Natural Science Foundation of China including international cooperative research project, key project and other project, professor Chen Xikang from Academy of Mathematics and Systems Scie...Under the support of National Natural Science Foundation of China including international cooperative research project, key project and other project, professor Chen Xikang from Academy of Mathematics and Systems Science under the Chinese Academy of Sciences, together with his colleagues, put forward in-put-occupancy-output technique and then used it in national grain output prediction approach. The main achievements are as follows:展开更多
文摘为增强综合能源系统负荷精细化分解水平,充分利用误差信息以进一步提升预测性能,提出一种基于聚合混合模态分解和时序卷积神经网络(temporal convolutional network,TCN)的综合能源系统负荷修正预测框架。首先,采用改进完全集合经验模态分解对电、冷和热负荷初步分解处理,随后利用变分模态分解对具有强复杂性的子序列进一步分解。然后,依据最大信息系数(maximum information coefficient,MIC)分析多元负荷的耦合特性并通过多元相空间重构(multivariate phase space reconstruction,MPSR)丰富特征信息。最后,构建基于TCN的修正预测模型。以校园综合能源系统算例对比不同预测模型,结果显示所提修正预测框架的电、冷和热负荷预测均具有较低的平均绝对百分比误差,有效解决了预测中模态分解的模态混叠以及模态高频分量问题,实现预测误差修正。
文摘Under the support of National Natural Science Foundation of China including international cooperative research project, key project and other project, professor Chen Xikang from Academy of Mathematics and Systems Science under the Chinese Academy of Sciences, together with his colleagues, put forward in-put-occupancy-output technique and then used it in national grain output prediction approach. The main achievements are as follows:
