In the process of power grid construction

power grid infrastructure files are important process data. However

its complicated quantity and types bring a great challenges to the green digital management and early warning of the files. To address this challenge

a kind of electronic management and early warning method for electric grid infrastructure files based on the improved fusion of prior knowledge convolutional neural network (PKCNN) is proposed in this paper. Firstly

the PKCNN convolutional layer is used to extract features from the input electric grid infrastructure files

and prior knowledge is introduced to assist in the training of the network parameters. Then

a nonlinear SVM optimized by the firefly algorithm (FA) is used to improve the classification function in the PKCNN to enhance the accuracy of the electronic management and early warning of electric grid infrastructure files. Finally

the electronic management and early warning method of electric grid infrastructure file based on improved PKCNN model is set up for simulation and verification. The results show that the PKCNN network has stronger feature learning ability and faster convergence speed than traditional CNN

and the use of SVM to improve the classification function in PKCNN can significantly improve the recognition accuracy of PKCNN. Compared to the traditional CNN and CNN-SVM model-based warning methods

the method proposed in this paper has higher accuracy and stronger generalization in the identification and early warning of electric grid infrastructure file quality management.