In this paper, a method of power system equipment recognition based on image processing is proposed. Firstly, we carry out wavelet transform on the sound signal of power system equipment collected from the site, and obtain the wavelet coefficient–time diagram. Then, the similarity of wavelet coefficients–time images of different equipment and the same equipment in different periods is calculated, which is used as the basis of the feasibility of image recognition. Finally, we select the HOG features of the image, and classify the selected features using SVM classifier. The method proposed in this paper can accurately identify and classify power system equipment through sound signals, and is different from the traditional method of classifying sound signals directly. The advantages of image processing can be effectively utilized through image processing to avoid the limitations of sound signal processing.

Under the general trend of smart grid development in China, it has especially importance to maintain the stability of power generation, the safety of power operation and the reliability of power supply. However, most power plants need to participate in the frequency regulation market and the power spot market, resulting in frequent load fluctuations and often unstable operating conditions of power generation equipment. In this study, a real-time monitoring method based on a hybrid Genetic Algorithm (GA) and Convolutional Neural Networks (CNN) algorithm is utilized to monitor the operation status of power transformers in power plants in real time. The GA-CNN algorithm model is proposed by analyzing the advantages and disadvantages of CNN and GA. It is proved that the accuracy of the GA-CNN is greatly improved compared with the CNN. In the recognition results, the error rate of the GA-CNN is only 1.86%, while that of the CNN is 4%; the random matrix accuracy of the predicted and actual output values of the GA-CNN model is 98.11%, and the three factors affecting the operating status of the equipment, namely temperature and humidity of the external environment and the daily power generation of the power plant, are also acceptable. The model selected for this study is able to detect abnormalities in the operating state of power transformers and provide timely feedback on changes in the external environment of the equipment.

The equipment condition monitoring based on computer hearing is a new pattern recognition approach, and the system formed by it has the advantages of noncontact and strong early warning abilities. Extracting effective features from the sound data of the running power equipment help to improve the equipment monitoring accuracy. However, the sound of running equipment often has the characteristics of serious noise, non-linearity and instationary, which makes it difficult to extract features. To solve this problem, a feature extraction method based on the improved complementary ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) and multiscale improved permutation entropy (MIPE) is proposed. Firstly, the ICEEMDAN is utilized to obtain a group of intrinsic mode functions (IMFs) from the sound of running power equipment. The noise IMFs are then identified and eliminated through mutual information (MI) and mean mutual information (meanMI) of IMFs. Next, the normalized mutual information (norMI) and MIPE are calculated respectively, and norMI is utilized to weigh the corresponding MIPE result. Finally, based on the separability criterion, the weighted MIPE results are feature-dimensionally reduced to obtain the multiscale entropy feature of the sound. The experimental results show that the classification accuracies of the method under the conditions of no noise and 5 dB reach 96.7% and 89.9%, respectively. In practice, the proposed method has higher reliability and stability for the sound feature extraction of the running power equipment.