In order to improve the online monitoring capability of electromagnetic transformers, particle swarm optimization algorithm is combined with BP neural network, and L2 regularization term is introduced to prevent overfitting, in order to enhance the online monitoring and fault detection capability of electromagnetic transformers. The newly designed system includes data acquisition, data processing, improved neural network models, decision support, and user interface. The particle swarm optimization algorithm is used to optimize the weights and thresholds of the BP neural network to improve its predictive performance. The BP neural network adopts a multi-layer feedforward neural network structure and trains network parameters through backpropagation algorithm. The study also established a mathematical model for electromagnetic transformers, and trained a neural network model by simulating fault signals. The experimental results show that the proposed improved algorithm can effectively identify various types of faults with an accuracy rate of up to 100%, and has high diagnostic accuracy and practical value.