基于卷积神经网络的非侵入式负荷识别方法

Non-intrusive load identification method based on convolutional neural network

针对非侵入式负荷识别中识别准确率不高的问题,提出基于卷积神经网络的非侵入式负荷识别方法。首先对电阻类负荷、电容类负荷、电感类负荷进行数学建模,并结合tracebase master数据集构建样本库;然后建立卷积神经网络模型,损失函数为交叉熵函数,优化算法采用自适应矩估计优化算法;最后对网络进行训练和测试。仿真结果表明,与循环神经网络等方法相比,本文的方法能够有效识别非侵入式负荷,并具有很好的抗噪性能,具有良好的应用前景。

Aiming at the problem of low recognition accuracy in non-intrusive load identification, a non-intrusive load identification method based on convolutional neural network is proposed. Firstly, mathematical modeling of resistance load, capacitance load and inductance load is carried out, and the sample library is constructed by combining tracebase master data set;then the convolutional neural network model is established, the loss function is cross entropy function, and the optimization algorithm uses adaptive moment estimation. Optimize the algorithm;finally train and test the network. The simulation results show that compared with the methods of cyclic neural network, the proposed method can effectively identify non-intrusive loads and has good anti-noise performance, and has a good application prospect.