Aiming at the shortcomings in application of BP neural network to power transformer fault diagnosis

a neural network evolved by adaptive chaotic particle swarm optimization is proposed. In this algorithm

the inertial weight is adjusted by evolution speed factor and concentration factor

the learning factors are improved

and chaotic system is introduced to eliminate the conventional BP neural network’s weaknesses such as slow convergence speed and the trend to local minimum value by optimizing network parameters. Subsequently

a transformer fault example based on DGA is simulated

and the simulation results are compared with that of conventional transformer diagnosis methods. It is concluded that the proposed algorithm can improve the efficiency of diagnosis and the accuracy of fault pattern recognition.