改進SSA優化BP神經網絡的變壓器故障診斷

摘" 要： 變壓器故障類型的準確診斷對保障電網的安全與穩定至關重要。針對BP神經網絡與麻雀搜索算法（SSA）存在收斂緩慢和易陷入局部極值導致無法準確診斷的問題，提出將改進的麻雀搜索算法（ISSA）優化BP神經網絡應用于變壓器故障診斷。首先，引入非線性慣性權重和縱橫交叉策略，從而提高算法的收斂速度和全局尋優能力；其次，將ISSA與傳統SSA在收斂函數上進行對比分析，得到ISSA算法在迭代12次后以52%的準確率收斂，而SSA算法迭代23次后才達到25%的準確率，證明了ISSA在收斂速度和精度方面有明顯提高；最后，將ISSA?BP、SSA?BP和BP診斷模型進行對比。實驗結果表明，ISSA?BP模型準確率達到了97%，比SSA?BP、BP神經網絡模型分別提高了4%和11%，可以認為提出的算法模型在變壓器故障診斷領域具有更高的精度與良好的發展前景。

關鍵詞： 麻雀搜索算法； BP神經網絡； 變壓器； 故障診斷； 非線性慣性權重； 縱橫交叉策略

中圖分類號： TN919?34； TM41" " " " " " " " " " 文獻標識碼： A" " " " " " " " " " " 文章編號： 1004?373X（2025）04?0145?06

Improved SSA optimized BP neural network for transformer fault diagnosis

WANG Fanrong1， 2， WANG Junhan1， JIANG Junjie1

（1. School of Electrical and Electronic Engineering， Hubei University of Technology， Wuhan 430068， China;

2. Xiangyang Industry Research Institute， Hubei University of Technology， Xiangyang 441100， China）

Abstract： Accurate diagnosis of transformer fault types is crucial to ensure the safety and stability of the power grid. In allusion to the problems that BP neural network and Sparrow search algorithm （SSA） have slow convergence speed and easy to fall into local extreme， which cannot make accurate diagnosis， an improved Sparrow search algorithm （ISSA） is proposed to optimize BP neural network for transformer fault diagnosis. The nonlinear inertia weights and vertical and horizontal crossover strategies are introduced to improve the convergence speed and global optimization ability of the algorithm. A comparative analysis of the convergence function between ISSA and traditional SSA shows that ISSA algorithm converges with an accuracy of 52% after 12 iterations， while SSA algorithm converges with an accuracy of 25% after 23 iterations， which proves that ISSA has significantly improved convergence speed and accuracy. The diagnostic models of ISA?BP， SSA?BP and BP were compared. The experimental results show that the accuracy of ISSA?BP model can reach 97%， which is 4% and 11% higher than that of SSA?BP and BP models， respectively. The proposed algorithmic model has higher accuracy and good development prospect in the field of transformer fault diagnosis.

Keywords： Sparrow search algorithm; BP neural network; transformer; fault diagnosis; nonlinear inertia weight; vertical and horizontal crossover strategy

0" 引" 言

油浸式變壓器因其工作環境復雜多變、超負荷運行和易絕緣受潮老化等問題容易發生故障，當變壓器出現故障時，只依靠實地員工的經驗來判斷故障的確切原因是非常困難的，這會造成非常嚴重的后果[1]。因此，對變壓器故障性質進行檢測，并對其故障類型進行準確和高效的診斷，對保障電網的安全與穩定具有重要意義[2]。

當變壓器發生故障時，會導致H2、C2H4等氣體溶解于絕緣油中，有相關研究表明，對變壓器油中溶解氣體進行檢測能夠從側面反映出變壓器故障類型[3]。……