RPROP神经网络在非侵入式负荷分解中的应用

为了解决常用家电设备投切状态辨识问题，提出一种以神经网络为辨识模型的方法，增强其快速辨识能力。首先，从负荷印记出发，针对各用电设备的稳态电流谐波特性，建立用电设备特征标签。然后，采用弹性BP(Resilient back propagation，RPROP)神经网络，将输入数据特征向输出层非线性映射，实现快速收敛至全局最优点。训练中采用多种设备组合方式，进行用电设备特征辨识。最终，以五类常用用电设备进行实验，实验结果表明该算法能够有效地识别家用设备的工作状态组合，且对功率相近、谐波具有较小差异的用电设备工作状态也具有很好的辨识能力。

Application of RPROP neural network in nonintrusive load decomposition

In order to identify the common-used household appliances, this paper proposes a kind of neural network which is effective to enhance the identification ability. First of all, based on load signature, aiming at harmonic characteristics of steady-state current in each electrical equipment, the feature tag is thereby established. Then, the RPROP neural network is adopted, which makes the input data feature nonlinearly map to output layer, and guides the neural network to converge to global optimal point rapidly. When training the neural network, the combined features are used to decompose the characteristics of electrical equipment. Finally, the experimental results of five common electrical appliances demonstrate that the proposed algorithm can effectively identify combined working states of household appliances, and it also can decompose the working states of electric appliances with similar power and little different harmonics.