基于改进遗传算法的非侵入式电器负荷识别

Non-invasive electrical load recognition based on improved genetic algorithm

针对传统非侵入式负荷识别算法在电器负荷接近或较小时并不能得到较理想识别效果的问题,提出了一种基于谐波特征和遗传算法的非侵入式电器负荷识别算法.通过提取电流数据的非活性电流及其谐波特征,来增大不同用电器间的差异性,从而提高识别精度;并使用遗传算法优化的神经网络的权重、阈值和隐含层神经元个数来提高分类识别精度,达到细粒度用电分析的目的;使用包含5种家用电器的用电场景测试所提出算法的识别精度,并将其与反向传播神经网络算法相比较.仿真测试结果表明,提出的特征和算法具有更高的负荷识别准确率及更快的识别速度.

Aiming at the problem that the traditional non-intrusive load identification algorithm can not obtain the ideal recognition effect when the load of electrical equipment is close or smaller, a non-intrusive electrical load recognition algorithm based on the harmonic characteristics and genetic algorithm was proposed . The non-active current of current data and its harmonic characteristics were extracted to increase the difference between different electric appliances so as to improve the recognition accuracy. The weights, thresholds and neuron number in the hidden layer of neural network optimized with the genetic algorithm were used to improve the classification recognition accuracy so as to achieve the purpose of fine-grained power analysis. The recognition accuracy of proposed method was tested with a power scenario containing five household electric appliances, and was compared with that obtained with the BP neural network algorithm . The results of simulation test show that the proposed characteristics and algorithm have higher load recognition accuracy and faster recognition speed.