Research on Intelligent Diagnosis Algorithm of Leakage Current in Low Voltage Distribution Network Based on Machine Learning

With the development of smart grids, the security of power systems has become a key issue. This paper discusses a smart leakage current diagnosis method for low-voltage distribution networks. This study mainly uses the BP neural network model, the LightGBM algorithm, and the LightGBM model combined with the IWOA (Imperialist Competitive Algorithm) optimization strategy to conduct leakage current diagnosis. The performance differences between the three are compared and analyzed. The BP neural network simulates the working principle of neurons in the human brain to perform nonlinear mapping and can capture complex data features. The LightGBM is an efficient gradient boosting framework that uses the leaf-wise growth strategy and histogram cache acceleration technology to make the model training more efficient. Furthermore, by introducing the IWOA optimizer to optimize the parameters of the LightGBM model, we have achieved the best results in leakage current diagnosis, improving the accuracy of diagnosis and enhancing the model's generalization ability. The experimental results show that, on the same test set, the IWOA-based LightGBM model has a significant advantage in computational efficiency while maintaining a high prediction accuracy. Although the BP neural network can achieve a higher accuracy in some cases, it has some shortcomings in training time and resource consumption compared to LightGBM. Taking into account the diagnosis accuracy and system operational efficiency, the IWOA-based LightGBM is more suitable for real-time monitoring of low-voltage distribution network leakage fault diagnosis. This study not only provides more reliable leakage fault diagnosis methods for low-voltage distribution networks, but also lays a foundation for further improving the intelligentization level of the grid. Future research directions will focus on improving the algorithm's real-time performance and adaptability to complex grid environments.