基于CEEMDAN-GRU的电网工程主材价格多步预测  

Multi-step Forecasting for Main Material Prices of Power Grid Projects Based on CEEMDAN-GRU

作　　者：张继钢 吴良峥 乔慧婷 陈雯 ZHANG Jigang;WU Liangzheng;QIAO Huiting;CHEN Wen(Energy Development Research Institute,China Southern Power Grid,Guangzhou 510663,China)

机构地区：[1]南方电网能源发展研究院有限责任公司,广东广州510663

出　　处：《控制工程》2023年第11期2134-2142,共9页Control Engineering of China

基　　金：中国南方电网有限责任公司科技项目(ZBKJXM20220003)。

摘　　要：电网工程主要原材料价格变动对电网工程的造价控制有着重要影响。为提高主材价格的预测精度,提出了一种基于自适应噪声完备集合经验模态分解和门控循环单元的电网工程主材价格多步预测方法。首先对原始价格序列进行分解,随后根据分解所得的各子序列的模糊熵值进行聚类。对模糊熵值较大的聚合序列进行变分模态分解,分解所得的各子序列利用GRU模型进行多步预测;对模糊熵值较小的各聚合序列直接进行多步预测。基于真实数据对所提预测方法的性能进行了实验,结果表明所提方法在预测精度上有明显提升,对电网工程材料价格预测具有较大的参考价值。The price change of main raw materials in power grid projects has an important influence on the cost control of power grid projects.In order to improve the forecasting accuracy of main material prices,a multi-step forecasting method for main materials prices of power grid projects based on the complete ensemble empirical mode decomposition with adaptive noise and gated recurrent unit is proposed.Firstly,the original price sequence is decomposed;then clustering is carried out according to the fuzzy entropy value of each subsequence obtained from the decomposition.The aggregation sequence with large fuzzy entropy is decomposed by variational mode decomposition,and the decomposed subsequences are multi-step predicted using the GRU model.The aggregation sequences with small fuzzy entropy are multi-step predicted directly.Based on real data,the performance of the proposed prediction method is tested.Compared with the comparison methods,the prediction accuracy is significantly improved,which demonstrated the reference value of the proposed method for the material price prediction of power grid projects.

关 键 词：价格预测 自适应噪声完备集合经验模态分解 门控循环单元网络 模糊熵 变分模态分解 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]