基于门控循环单元神经网络及负荷激活提取的非侵入式负荷监测算法

Non-intrusive Load Monitoring Algorithm Based on Gated Recurrent Unit Neural Network and Load Activation Extraction

非侵入式负荷监测技术对于电力公司准确预测电力负荷、科学制定电网调度方案以及为用户提供良好的用能方案策略有着重要意义。提出一种以门控循环单元为神经网络架构,结合了数据去噪以及负荷激活提取的方式,通过大量的功率数据完成了对神经网络模型的训练。通过测试实现了对于开源数据集UK-DALE中的四种典型电器负荷的功率分解,以平均绝对误差以及能量分解准确率作为分解的评价指标,取得了较为优异的效果。实现了对于家庭负荷的准确分解,证明了所提方法的有效性。

Non-intrusive load monitoring technology is of great significance for power companies to accurately predict power loads,scientifically formulate power grid scheduling plans,and provide users with good energy consumption plan strategies.A gated recurrent unit was proposed as the neural network architecture,which combined data denoising and load activation extraction,and completed the training of the neural network model through a large amount of power data.The power decomposition of the four typical electrical loads in the open source data set UK-DALE was realized through testing,and the mean absolute error and the energy decomposition accuracy were used as the evaluation indicators of decomposition,and excellent results were achieved.The accurate decomposition of household loads is realized and the effectiveness of the proposed method is proved.