Semantic Analysis Model And System Design Of Defect Evaluation For Unstructured Data Of Power Equipment

The power defect description text is an important component of power unstructured data,which contains rich information about the health status of power equipment.Therefore,the rapid and accurate grade evaluation based on the severity of defects is not only helpful for the timely maintenance of electric power equipment,but also can ensure the stable and healthy operation of the whole power supply system.Generally,according to the main components of the power equipment and the corresponding phenomenon standards,the operation and maintenance personnel classify the defect description into four categories:other,general,emergency and major.But diverse results which may lead to a low guarantee of the defect grades may be acquired by different engineers for same defect descriptions.Therefore,how to realize the accurate evaluation of defect description,guide the timely implementation of equipment maintenance work is a problem that needs to be solved right now.This paper plans to carry out research based on traditional machine learning and deep learning.A semantic analysis model based on traditional machine learning and deep learning for power equipment defect level evaluation is proposed to solve the problems in power equipment defect evaluation,such as the difficulty in the fusion of discriminant results,the lack of feature extraction of defect description text and the need to further improve the accuracy of the model.The main achievements are as follows.（1）A power defect level determination method based on multiple optimization algorithm and DS evidence theory is proposed.In this method,the experimental data is vectorized according to the word scales and the weight table of light colors at first.Then MOA algorithm is used to find the optimal search radius of vectorization training set data.Finally,the grade results of the test set obtained by MOA during 50 experiments within the better search radius interval are fused by DS evidence theory.Through corresponding comparative experiments,it can be seen that the method proposed in this chapter is superior to the traditional k-means model,which illustrates that the method proposed in this paper is effective and feasible.（2）A power defect level determination method based on attention mechanism optimized combination neural network is proposed.Firstly,the distributed character vector is used for vector representation of power defect descriptions.Then,the local features and sequence features of power defect descriptions are extracted by using the convolutional recurrent neural network.Finally,using the attention mechanism to assign weights of the semantic features obtained by the combination neural network,so as to reduce the loss of key features and further enhance the influence of key information on the classification results.Taking 110000 defect description data of a power grid company in Southwest China from 2014 to 2019 as experimental objects,the Acc,MF₁and WF₁values of the method proposed in this paper are 0.9275,0.9112 and 0.9275.It illustrates that the method is effective and feasible to determine the defect grades of power defect descriptions and it can provides help for the intelligent operation of power grid.（3）Python was used to design the power equipment defect evaluation system to facilitate the power operation and maintenance personnel to determine the level of power equipment defect description.