| Adequate power supply provides the prerequisites for China’s economic development.Based on the particularity of China’s energy structure,thermal power plants are still the country’s largest power generation model.In recent years,research on intelligent classification algorithms has been continuously deepened,effectively solving the problem of obtaining effective information under massive data.Aiming at the problems that the managers of thermal power generation enterprises have in the safety training management,this paper combines the intelligent classification algorithm and considers the engineering implementation of the "safety two-vote training assessment system".A power plant production event based on the convolutional neural network algorithm Report single classification algorithm.This paper studies common classification algorithms.By analyzing the characteristics of production event report sheets of thermal power plants,a convolutional neural network algorithm is selected as the basis of the classification algorithm and optimized.Use the Ik-Analyzer tokenizer to perform Chinese word segmentation on the text of the event report to obtain the constituent phrases of the report.Text-Rank algorithm was used to filter out the keywords of each report.The report key word text is selected as the corpus to input into the word2 vec model,and the space vector of the key word object is obtained initially.Considering the corresponding engineering applications and subsequent upgrade work,languages and frameworks under the same platform should be used for development as much as possible.Therefore,Google’s open source project deeplearning4 j is used to conduct training on convolutional neural network classification."To categorize.Finally,iterative calculation is used to obtain the neural network classification model.After completing the modeling work,according to the data classification model,you can obtain the recommended classification types of unclassified event reports,and you can also use the current hot events to perform event report learning and recommendation.The actual measurement proves that for the unclassified report form used for reference,the algorithm in this paper can effectively filter out the classification types that match it and recommend them to the user,and statistics the matching degree of various classification types.For hot events,it can also be classified by the classification model It is recommended to perform event report learning to facilitate users to perform reinforcement learning on relevant knowledge points. |
