Key Technology of Non-intrusive Load Feature Recognition for Industrial Users Based on Edge Computing

The objective of this article is to study a key technology of non-intrusive load feature recognition for industrial users based on edge computing to improve energy management and production efficiency. In the study, the method steps of data collection and preprocessing, power load feature extraction and selection, model training, and load feature recognition effectiveness experiments are adopted. First, the data collection and preliminary processing are carried out by deploying edge computing nodes in the industrial field and using the efficient embedded system. Second, the features of time series signals are extracted using the fast Fourier transform (FFT), and the principal components analysis (PCA) is applied for feature dimensionality reduction. Then, the convolutional neural networks (CNN) are selected to classify and recognize the features. Finally, the performance of the method in this article in terms of resource utilization, recognition precision, response time, and model update efficiency is evaluated through experiments. The experimental results show that under high load conditions, the method in this article outperforms the traditional cloud computing non-intrusive load monitoring (NILM) method in terms of central processing unit (CPU) usage, memory usage, and network bandwidth usage. When the load intensity level is 7, the CPU usage rate is 85%; the memory usage rate is 46%; the network bandwidth usage is 7.23MB/s, while those of the traditional method are 99%, 92%, and 12.45MB/s, respectively. In addition, in the recognition precision testing, the method in this article outperforms the traditional method in the recognition accuracy of motors, heaters, compressors, and pumps and has a shorter classification time. In the model update efficiency testing, the highest update efficiency of edge nodes reaches 100%. These data demonstrate the advantages of the method in this article in resource optimization and real-time performance and verify the effectiveness and efficiency of the edge computing technology in the industrial load feature recognition.