基于矩阵特征重构的深度学习负荷监测方法

为解决负荷特征学习和训练模型难以求解、识别准确率不高的问题,文章提出了一种基于奇异值特征矩阵重构的深度学习非侵入式负荷监测方法,首先利用奇异值分解算法对采集到的混合信号进行负荷分离,并设定奇异值的门限,保留通过门限的左右奇异值矩阵,然后获取处理后的左右奇异值矩阵的克罗内克积,实现信号的特征矩阵重构;将大量典型家电的运行电流数据转换成重构特征矩阵的形式,并使用卷积神经网络模型进行训练,从重构的特征矩阵中提取独立负荷特征,进而建立能够处理重构特征矩阵数据的卷积神经网络模型,并基于该模型提取数据特征,从而达到辨识负荷特征的目的。

In order to solve the problem that the load feature learning and training model is difficult to solve and the recognition accuracy is not high, this paper proposes a deep learning n on-invasive load monitoring method based on the reconstruction of singular value eigenmatrix. Firstly, the singular value decomposition algorithm is used to separate the mixed signals, and the threshold of singular value is set, and the left and right singular value matrix passing through the threshold is reserved. Then, the Kronecker product of the left and right singular value matrices is obtained to reconstruct the characteristic matrix of the signal;a large number of running current data of typical household appliances are converted into the form of reconstructed characteristic matrix, and the convolution neural network model is used for training, and the independent load characteristics are extracted from the reconstructed characteristic matrix. Furthermore, a convolution neural network model which can deal with the reconstructed feature matrix data is established, and the data features are extracted based on the model, so as to achieve the purpose of load feature identification.