Fault diagnosis method for power grid transformers by integrating improved AFSA and radial basis function network

Power grid transformers are vital for stable power supply, and their failure can disrupt grid operations. Accurate fault diagnosis is essential to ensure reliability. This study proposes an improved artificial fish swarm algorithm (AFSA) for transformer fault diagnosis, integrating radial basis function networks (RBF) and kernel limit learning models to enhance accuracy. The method processes transformer data more effectively, reducing diagnosis errors by 0.014-0.029 compared to standalone RBF. Fusion models achieved 4.9%-5.9% higher accuracy than RBF alone. Notably, it excelled in gas concentration prediction, achieving zero deviation for C2H2. The results demonstrate superior performance over traditional methods, significantly improving fault diagnosis precision. This approach offers valuable guidance for maintaining transformer reliability in power grids.