Research On Classification Method Of Spare Parts Of Power Grid Based On Convolutional Neural Network

It has become an urgent task for the current power grid industry and academia to explore the scientific classification method of power grid spare parts,improve the quality and reliability of power supply,and thus improve the power grid’s response ability to cope with accidents and save the cost of power grid enterprises.How to combine the characteristics of power grid spare parts to determine the multi-attribute indicators that affect their classification and scientifically classify large quantities of power grid spare parts according to the indicators has become the premise and key problem of power grid spare parts management.Aiming at the problem of power grid spare parts classification,this paper studies the power grid spare parts classification method based on convolutional neural network.The specific research contents include:(1)Analyze the current situation and shortcomings of the classification management of spare parts in power grid companies,and put forward the improvement ideas.The current multi-level inventory structure,spare parts classification and management of power grid companies are combed and analyzed in detail,and the existing problems and shortcomings in the management of power grid spare parts are summarized.Combined with the actual operation situation of power grid company,the improvement idea of the classification method of power grid spare parts is put forward.(2)A classification model of power grid spare parts based on convolutional neural network is constructed.Firstly,according to the characteristics and historical consumption of power grid spare parts,the qualitative indexes and quantitative indexes affecting their classification were analyzed,and a multi-attribute classification index system with the characteristics of power grid spare parts was constructed.Secondly,the influence index is taken as the input of the convolutional neural network,and the ShuffleAdapter algorithm is used to transform the data of one-dimensional power grid spare parts into two-dimensional ones,and the number and size of the convolutional neural network’s convolutional layer,pooling layer and fully connected layer are adjusted.Batch Normalization layer(BN)is added to improve the training efficiency and classification performance of convolutional neural network models.Finally,the Softmax classifier is used to obtain the classification results.(3)According to the classification results of power grid spare parts,the corresponding classification management strategy is put forward.In view of the features of different kinds of spare parts in the power grid,respectively using partial order,continuous review inventory control model of economic order quantity model,cycle counting,inventory control model for different categories grid spare parts inventory level,safety stock,such as maximum inventory index to carry on the quantitative calculation,get the grid of spare parts classification management strategy.(4)To carry out case analysis.According to the data of some spare parts of power supply Company A,data preprocessing of sample data was carried out first,and the index data was standardized and the sample data volume was expanded.Then,the training set data of power grid spare parts is input into the convolutional neural network model constructed in this paper by using the method of 5-fold cross validation for training.The model structure is optimized and the hyperparameters are adjusted through the model training performance to improve the classification performance of the model,which verifies the effectiveness of the classification model constructed in this paper.Finally,the inventory management strategy of the classified power grid spare parts is analyzed,and the order point,order quantity and maximum storage water equality index of different kinds of power grid spare parts are calculated,so as to provide decision support for enterprise managers.