Haixin Tong

The struggling and twitching of an electrically shocked body in low-voltage distribution grids can complicate and diversify the electric shock current, posing challenges to traditional electric shock detection methods. To tackle this issue, we propose a fault identification method for animal electric shocks considering unstable contact situations. First, a low-voltage distribution network grounding fault experimental platform is established in a real system, creating a sample library that encompasses various operational scenarios. Second, based on the dynamic characteristics of contact resistance and the principles of arc breakdown generation, an electrical equivalent modeling analysis is conducted on the unstable contact region. The underlying mechanisms that increase the complexity and diversity of electric shock current are revealed. Then, innovative features such as “the sum of periodic point differences,” “energy fluctuation arrays,” “pulse regularity,” and “the count of reversals in the direction of changes” are proposed. By integrating several traditional features and automatically extracted encoded features, a feature vector for animal electric shock faults is established from the perspectives of dynamic amplitude, high-frequency components, energy, and encoded features. In addition, a light gradient boosting machine optimized by multiattraction mechanism zebra optimization has been developed. This model achieves accuracy in pattern recognition tasks through its unique ensemble computation and one-sided sampling mechanism, while also reducing time and space costs. Finally, the proposed method is compared with three existing electric shock identification methods using our experimental platform's testing set. The results indicate that the proposed method demonstrates superiority in terms of accuracy for identifying animal electric shocks.

This study aims to tackle the challenge of explaining the underlying mechanisms behind the time-varying impedance phenomenon in animals experiencing low-voltage electric shocks. A dynamic impedance model that considers the effect of water electrolysis (WEDI) has been developed. First, we conducted root cause analyses through progressive validation experiments, identifying water within the shocked body as the key factor influencing impedance variation. Monitoring hydrogen concentration above the electrodes and measuring mass changes in the shocked body revealed that the time-varying impedance is closely related to internal water electrolysis. Second, we quantitatively analyzed the impact of water molecule decomposition during electrolysis on the salt concentration and conductivity within the electrically shocked body. A mathematical relationship between the variable resistance within the body and time was derived. A dynamic impedance model for animal electric shock that considered the effects of water electrolysis was subsequently established, explaining the underlying mechanism behind the time-varying impedance phenomenon. Finally, the Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE), and Root Mean Square Error (RMSE) between the electric shock current predicted by the WEDI model and actual measurements were 0.00357, 0.00350, and 0.00446. Compared to existing models, the WEDI model demonstrates superior accuracy and interpretability.

This paper aims to tackle the difficulties in identifying series arc faults with feature aliasing in low-voltage electricity scenarios. Supported by feature sensitivity analysis, a method for identifying low-voltage series arc faults using a time variance-accuracy coefficient optimized random forest (TVARF) model is proposed. Firstly, in the quantitative analysis of real sample feature sensitivity, it is discovered that the main reason for the above issue is that the sensitivity of the sample features decreases in certain scenarios, thereby interfering with the judgment of traditional arc detection methods. Secondly, the random forest model, with its anti-interference capabilities derived from its ensemble operation mechanism, is utilized as the core model of proposed method. Thirdly, this paper constructs a TVA coefficient to achieve multi-objective optimization of random forest hyper-parameters and institutes a program module that use slow-voltage load state samples for training the random forest model.Additionally,aprogramfordenoisingandextractingtimefrequency features is embedded into the input of the trained random forest model. Finally, the TVA-RF method hit 99.978% accuracy in laboratory tests, outperforming traditional methods in accuracy, training efficiency, and calculation speed. In real low-voltage systems, it achieved 98.096% accuracy. Overcoming feature overlap interference, the TVA-RF method accurately identified low-voltage series arc faults.