Design And Implementation Of Multi Round Dialogue System For Power Grid Equipment Troubleshooting

In the face of the rapid growth of new knowledge and a large number of references in the power grid equipment system,the maintenance staff are limited by the reaction speed and thinking ability in the actual maintenance process,and can not timely and accurately identify the fault and make maintenance decisions from these complex fault information.With the continuous development of communication technology and artificial intelligence technology,dialogue system based on artificial intelligence technology is gradually rising,which provides great convenience and maintenance reference for staffs maintenance process.Therefore,the research on how to apply artificial intelligence technology to the dialogue system for power grid equipment troubleshooting is of great significance and help to reduce maintenance costs and work errors,and promote the intellectualization and informatization of maintenance work.Firstly,this paper makes an in-depth study on the current dialogue system for power grid equipment troubleshooting,and summarizes three shortcomings of the current dialogue system for power grid equipment troubleshooting:first,the current dialogue system for power grid equipment troubleshooting is mainly single round,although simple tasks can be realized through one interaction,However,due to the ambiguity of the problems provided by the staff and the lack of information,it is still unable to complete some specific dialogue tasks,such as the solutions described for the maintenance problems,and can not accurately locate the problems and give answers.Second,the current power grid system can not realize intelligent dialogue because of the lack of a large number of Q&a tagging training data.Although the community dialogue and corpus can be searched by means of crawlers,it is far from being the bottom support of the dialogue system in terms of professionalism and quantity.Third,the power grid system has accumulated a large number of unstructured text data,such as case reports,maintenance standards,etc.most of the knowledge in this part of the data depends on the traditional knowledge management mode that experts in specific fields manually extract,sort out and store the data in the database in the form of Graphs and tables,because the structure of storing knowledge is relatively single,Reasoning is difficult and can not be used as the underlying knowledge base of dialogue system.Therefore,this paper proposes a multi round dialogue system for power grid equipment troubleshooting,which takes knowledge atlas as the bottom support.On the one hand,there is no need for dialogue training data,and dialogue interaction can be realized through knowledge map.On the other hand,compared with the traditional data organization form,the knowledge map connects the originally unconnected data,which is convenient for the system to reason and use.The effectiveness of the system is verified by testing on the power grid maintenance work order and multiple rounds of user satisfaction evaluation.In addition,this paper optimizes the construction of knowledge map,especially the knowledge extraction model based on remote supervision.Through a series of experiments,it not only improves the accuracy of the model,but also mines more effective data.Finally,based on Rasa framework,this paper designs and implements a multi round dialogue system for power grid equipment troubleshooting.Firstly,this paper makes a detailed demand analysis of the system,and divides the system into original data module,knowledge extraction layer,data persistence layer,maintenance decision interaction module and interaction and data management visualization module.Through a series of system tests,it is verified that the system can meet the use requirements of the scene and meet the design expectations.