A fundamental parameter of polluted insulator online monitoring is the leakage current, which has already been shown to be well-related to the pollution discharge of insulators. In this article, in an effort to quantitatively reflect the discharge intensity and the discharge status by the leakage current, we carried out an experimental study on artificial pollution discharge of insulators. A high-speed photographic apparatus was utilized to capture the entire process of local arcs on a porcelain insulator surface, including the arc generation, the arc development, and the flashover, for which the associated leakage current of insulators was synchronously digitized. A comparative analysis of the relation between the two-dimensional discharge image and the leakage current waveform in the process of arc generation and development shows that if the arc area on the insulator surface is relatively small and the leakage current passes through zero, the arc might completely become extinct, whereas this phenomena will not occur if the arc area is larger. In addition, the amplitude of the discharge arc area is found to be roughly proportional to the square of leakage current over the range of leakage current amplitude from 0 to 150 mA. Our results can provide an important guidance for judgment of the discharge status and the discharge intensity on insulator surfaces using the leakage current of insulators.

The surface discharge development processes of polluted porcelain insulators for power transmission lines are tightly related with the development of leakage current (LC), the characteristics of LC, the insulating condition, and discharge intensity of the insulator surface have an important significance for revealing the contamination discharge state of insulators. In order to analyze the LC characteristics of porcelain insulators in the process of pollution flashover, artificial pollution flashover tests on porcelain insulators were conducted in the artificial fog cabinet, and the characteristics of LC waveforms in time-domain and frequency-domain were simultaneously measured and analyzed during the tests. The results indicated that the amplitude of LC, fundamental harmonic, the third harmonic, and fifth harmonic had a strong correlation, the maximum of LC(I_m), the rate of total harmonics (THD), and the phase difference of fundamental harmonic (θ) were used for the representation of the characteristics of the LC waveform. The LC has the characteristics of high amplitude, low proportion harmonic, and small phase difference between the fundamental harmonic and voltage before the flashover occurrence. The test results provide effective reference for porcelain insulators in pollution flashover forecasting.

In some fire classification task samples, it is especially important to learn and select limited features. Therefore, enhancing shallow characteristic learning and accurately reserving deep characteristics play a decisive role in the final fire classification task. In this paper, we propose an integrated algorithm based on bidirectional characteristics and feature selection for fire image classification called BCFS-Net. This algorithm is integrated from two modules, a bidirectional characteristics module and feature selection module; hence, it is called an integrated algorithm. The main process of this algorithm is as follows: First, we construct a bidirectional convolution module to obtain multiple sets of bidirectional traditional convolutions and dilated convolutions for the feature mining and learning shallow features. Then, we improve the Inception V3 module. By utilizing the bidirectional attention mechanism and Euclidean distance, feature points with greater correlation between the feature maps generated by convolutions in the Inception V3 module are selected. Next, we comprehensively consider and integrate feature points with richer semantic information from multiple dimensions. Finally, we use convolution to further learn the deep features and complete the final fire classification task. We validated the feasibility of our proposed algorithm in three sets of public fire datasets, and the overall accuracy value in the BoWFire dataset reached 88.9%. The overall accuracy in the outdoor fire dataset reached 96.96%. The overall accuracy value in the Fire Smoke dataset reached 81.66%.