| With the continuous improvement of power equipment technology,the differentiation between national power industry standards and international common technical standards has gradually become an important factor restricting China’s power equipment exports and the participation of power companies in international power project construction.The power Internet of Things(IoT)is the application of the IoT in the power field and in smart grids.Any technological development cannot be separated from the support of standards.Real-time monitoring of the latest developments in the formulation and publication of power IoT standards can help grasp the trend of the core technology development of the power IoT,study and analyze the standardization status of the power IoT,and master its development trends,which is of strategic significance.However,in the current standard retrieval process,there are problems such as inaccurate retrieval and slow speed,and there is no complete evaluation system for standard evaluation.Therefore,by studying methods such as retrieval,classification,and evaluation of power IoT standards,efficient and reliable standard retrieval models,power IoT standard layout,and power IoT standard evaluation models can be obtained to provide direction and technical support for the development of power IoT technology,promote power IoT standardization,and conduct specific research as follows:(1)To solve the problem of low matching accuracy and efficiency in the standard retrieval process,an improved text matching model combining machine reading is proposed.The model introduces the n-gram and TF-IDF algorithm,which not only obtains the word frequency and inverse document frequency of the standard text,but also fully considers the word order problem.On this basis,a machine reading comprehension model is added to solve the problem that the original model does not have understanding ability.Finally,the effectiveness and reliability of the standard retrieval model is verified through experiments.(2)To improve the classification and display of power IoT standards retrieved,a method combining word vectors and graph neural network models is proposed to classify power IoT standards.This method trains on the constructed power IoT standard dataset and can achieve the classification of multi-target standards,providing technical support for subsequent classification research and further improving the classification efficiency and speed of the model.(3)The evaluation index system of power IoT standards is divided into three categories:status,quality,and economy.The three categories of indicators are preliminarily selected to complete the preliminary construction of the key evaluation index system for power IoT standards.A structure entropy-factor analysis-based index selection optimization model is built to optimize and adjust the preliminary selected indicators,and the final key evaluation index system for power IoT standards is constructed.Due to the fuzziness of expert evaluation of power IoT standards,a power IoT standard comprehensive evaluation model based on interval numbers is constructed,and the practicality and effectiveness of the evaluation system are verified through experiments.The analysis results show that the improved text matching model combining machine reading and the classification model combining word vectors and graph neural networks proposed in this paper can improve the efficiency,accuracy,and reliability of power IoT standard retrieval and classification,providing technical and theoretical support for power IoT standard layout analysis and standardization promotion.The power IoT standard comprehensive evaluation model constructed in this paper can effectively evaluate power IoT standards and improve the analysis of power IoT planning layout,providing theoretical material support for power IoT standard applicatio. |