In household appliances, nonlinear load appliances are gradually increasing. Dong Zhiwen and Su Jingjing from the School of Computer and Control Engineering of Minjiang University and other units wrote an article in the first issue of Electrical Technology in 2024, proposing a random forest fault arc identification method based on the time domain characteristics of the signal combined with the variational mode decomposition intrinsic mode energy entropy. Experiments show that, compared with other methods, the accuracy of fault arc identification in this method can reach 99%, and it is suitable for low-voltage distribution fault arc identification under a variety of typical loads and nonlinear loads
Electric energy is becoming more and more important in the rapid development of society, and the development of electric power industry provides convenience for people, but poor management or improper use may lead to electric shock casualties and fires, and bring heavy losses to the lives and property safety of the people. One of the main causes of residential electrical fires is the occurrence of faulty arcs.
In the home power distribution network, due to the complex lines and complex environment, it is very easy to cause fire accidents if the fault arc is not cut off in time, so it is very important to study the detection method of the fault arc. With the increase in the type and quantity of electrical products in industrial and commercial areas, nonlinear load electrical appliances are gradually popularized.
In the case of nonlinear load and compound load, it is difficult for traditional detection methods to achieve accurate identification. Solving the defects of traditional methods with the help of artificial intelligence algorithms is the main research direction in the field of fault detection. At present, the focus of fault arc research has expanded from the time-domain, frequency-domain and time-frequency domain analysis of current signals to the extraction of fault feature values and the optimization of machine learning algorithms, but the diversity of load types in low-voltage power distribution systems has a great impact on feature extraction and fault detection.
In order to solve the problem that the current is greatly affected by the load type when the low-voltage series fault arc is generated, and it is difficult to accurately identify and classify the fault arc through a single characteristic parameter, researchers from Minjiang University and other units obtained the IMF by combining the variational mode decomposition (VMD) on the basis of extracting the time-domain characteristics of the current waveform and obtaining its energy entropy as the characteristic quantity of fault judgment. Combined with the time-domain characteristics and energy entropy of the current waveform, the eigenvector is constructed and input into the random forest classifier to realize the detection and classification of arc faults of different load types.
Fig.1. Schematic diagram of fault arc test circuit
Fig.2 Flow of arc fault detection algorithm based on random forest
They found that in nonlinear load fault arc detection, the recognition accuracy is relatively low when the time domain feature is applied alone, especially when the normal current and fault current waveforms of the load are similar. Combined with the energy entropy feature, the identification rate of fault arc is effectively improved. The results show that the combination of time domain features and energy entropy features can significantly improve the detection accuracy of nonlinear load fault arcs.
Table Fault arc identification results of different classification models
In addition, compared with other loads, the fault characteristic information of the vacuum cleaner is not obvious, and the recognition accuracy at this time is low, and there is a large room for improvement.
According to the researchers, by comparing the accuracy of support vector machine, naive Bayesian, multilayer perceptron, K-nearest neighbor classification and the random forest model in this paper for fault arc recognition, the results show that the fault arc recognition accuracy of this model is the highest, reaching 99%.
The results of this work were published in the first issue of Electrical Technology in 2024, and the title of the paper is "Fault arc detection method based on variational mode decomposition energy entropy mixed time-domain features and random forest", and the authors are Dong Zhiwen and Su Jingjing. This project is supported by the Natural Science Foundation of Fujian Province, the Industry-University Cooperation Project of Fujian Provincial Universities, and the Scientific Research Project of Minjiang University.