基于VMD-CNN-BiLSTM-CBAM的配变短期负荷预测方法

Short-term Load Forecasting Method for Distribution Transformers Based on VMD-CNN-BiLSTM-CBAM

随着智能电网的发展,配变重过载预警的准确性对于维持电网稳定运行至关重要。本文提出了一种新的基于VMD-CNN-BiLSTM-CBAM模型的配变短期负荷预测方法,旨在提高预警准确性。该方法首先运用K均值聚类算法筛选出相似日,并利用变分模态分解(variational mode decomposition,VMD)对相似日的负荷数据进行分解,得到一系列内在模态函数(intrinsic mode function,IMF)分量。随后,各IMF分量通过结合卷积神经网络(convolutional neural network,CNN)、双向长短时记忆网络(bidirectional long short-term memory,BiLSTM)和卷积块注意力机制(convolutional block attention module,CBAM)的混合模型进行特征学习和预测。最终,使用样本熵理论将预测得到的各IMF分量重构合成,获得配变的预测日负荷曲线。实验结果表明,本文提出的方法在预测精度上有明显提升,为配变重过载预警提供了有效的技术支撑。

With the development of smart grid,the accuracy of heavy load early warning of distribution transformer is very important to maintain the stable operation of power grid.A new short-term load forecasting method for distribution transformer based on VMD-CNN-BiLSTM-CBAM model is proposed in this paper.Firstly,k-means clustering algorithm is used to filter out similar days,and variational mode decomposition(VMD)is used to decompose the load data of similar days,and a series of intrinsic mode function(IMF)components are obtained.Then,the IMF components were Convolutional neural network by a hybrid model of Convolutional Block Attention Module(CBAM),Bidirectional Long-Short-Term Memory(BILSTM)and Convolutional neural network(CNN).Finally,the predicted IMF components are reconstructed by using the sample entropy theory,and the predicted daily load curves of distribution variables are obtained.The experimental results show that the method proposed in this paper has a significant improvement in the prediction accuracy,and provides an effective technical support for the overload warning of distribution transformer.