: . 激光与光电子 VI 轨迹像素化图像作为 CNN 输入；其次，分析 CNN 超参数对模型性能影响，并使用 PSO 算法寻求最优解以提升模型识别效果；最后， 基于 PLAID、WHITED 公开数据集对 PSO-CNN 模型进行对比验证。实验结果表明， 该模型的识别准确率、F-measures 平均值皆优于其他模型，有效降低了设备之间的 混淆，具有良好的识别能力与泛化能力。 关键词 非侵入式电力负荷识别；深度学习；卷积神经网络；粒子群优化算法 中图分类号 TM714 文献标志码 A Non-intrusive Electric Load Identification Algorithm for Optimizing CNN Hyper-parameters Zhao Anjun1, Zhao Xiao1, Jing Jing2*, Xi Jiangtao1, Cui Pufang1 1School of Information and Control Engineering, Xi 'an University of Architecture and Technology, Xi’an Shanxi 710055, China; 2Northwest China Architecture Design and Research Institute, Xi’an Shanxi 710018, China; Abstract Aiming at the problems of low recognition rate and hyper-parameters setting of deep learning model in electric load recognition, a non-intrusive electric load recognition model(PSO-CNN) combining particle swarm optimization algorithm(PSO) and convolutional neural network (CNN) was proposed. Firstly, the pixelated image of VI trajectory of each appliance is used as the CNN input feature. Secondly, the influence of CNN hyper-parameter on model performance was analyzed, and PSO algorithm is used to find the optimal solution to improve model recognition effect. Finally, the PLAID and WHITED public data sets were used to compare and verify the PSO-CNN model. The experimental results show that the recognition accuracy and average F-measures of this model are better than other models. The
