基于分组空洞残差网络的非侵入式负荷分解

NON-INTRUSIVE LOAD DISAGGREGATION BASED ON GROUP DILATED RESIDUAL NETWORK

针对当前深度学习在非侵入式负荷分解应用中准确率低、易梯度消失、对使用频率较低的电器分解误差大等问题,提出一种分组空洞残差网络。进行滑动处理增加样本数量后,一方面基于残差网络提取深层负荷特征,降低网络优化难度,解决梯度消失问题;另一方面通过空洞卷积增大感受野,捕获更多数据,解决长时序数据较难学习的问题。实验结果表明,该模型比现有方法分解准确率更高,对使用频率较低的电器分解鲁棒性更好,对实现准确非侵入负荷分解有重要意义。

In view of the fact that the current deep learning has low accuracy,vanishing gradient and high disaggregation error of appliances with low frequency in non-intrusive load disaggregration,a non-intrusive load disaggregation method based on deep group dilated convolution residual network is proposed.The sliding window was used to increase the number of samples.This algorithm extracted deep load characteristics based on residual network,reduced the difficulty of network optimization and solved the problem of vanishing gradient.On the other hand,it used dilated convolution to increase the receptive field,captured more data,and solved the problem that long-term data is difficult to learn.The experimental results show this model can get better disaggregration results than existing studies,especially for the disaggregation of electrical appliances with low usage.