Abstract:

Objectives Timely and accurate detection of line-break faults is crucial for ensuring the normal and safe operation of distribution networks. However, the application of traditional criteria for single-phase line-break fault detection in active distribution networks has certain limitations. Therefore, a single-phase line-break fault detection method based on an improved support vector machine and integrating multiple electrical characteristic quantities is proposed. Methods First, an electrical characteristic index system incorporating startup criteria, traditional criteria, and active criteria is established. Second, the startup criteria are processed using the switching quantization method. Subsequently, kernel principal component analysis is employed to eliminate low-contribution feature indicators from the characteristic index system excluding the startup criteria. Finally, the dimensionality-reduced data are input into the support vector machine, and the sparrow search algorithm is utilized to optimize the parameters of the support vector machine, thereby obtaining the line-break fault detection model. Results Simulation case studies conducted on a modified IEEE 15-node model demonstrate that the proposed method effectively realizes dimensionality reduction of the characteristic quantities and improves the detection accuracy by 8.87% compared to using a single criterion. Conclusions The proposed method solves the problem that traditional criteria for single-phase line-break fault detection are prone to failure, and can effectively achieve fault detection under different scenarios.

Key words: active distribution network, single-phase line-break, support vector machine, sparrow search algorithm, kernel principal component analysis