基于NL2SQL实现电力数据智能交互

Realization of Power Data Intelligent Interaction Based on NL2SQL

为解决电网业务数据的智能交互查询,降低数据查询成本,提高电网数据利用率等问题,文章提出基于BERT(bidirectional encoder representations from transformer)模型的NL2SQL(natural language to SQL)算法,拟实现电力数据的智能交互、提升服务质量、加快电网数据资产变现。该研究可以实现用户通过输入拟查询的问句,经BERT模型词嵌入后,利用Text2Sql技术将BERT模型的处理结果转化为SQL语句,实现从数据库中查询到所需要的信息。文章构建了NL2SQL模型和算法训练,实验结果表明该方法的可行性和易用性。该方法能够帮助移动作业人员便捷查询数据,快速找到结构化的电力数据信息,辅助经营管理者决策分析,为各层级业务研究人员节省查询成本,加快电力数据资产变现。

In order to solve the problem of intelligent interactive query of power grid business data,reduce the costs of the data query and improve the utilization rate of power grid data,this paper proposes an nl2sql(natural language to SQL)algorithm based on the Bert(bidirectional encoder representations from transformers)model to realize the intelligent interaction of the power grid data,improve the quality of the services and accelerate the realization of the power grid data assets.The research is intended to help the users input the questions they want to query,embed the words through the Bert model,and transform the results of the Bert model into the SQL statements by using the text2sql technology so as to query the required information from the database.This paper constructs the nl2sql model and its algorithm training.Experimental results show feasibility of the method.It helps the mobile operators to query data conveniently,find out the structured power data information quickly,and assist the managers to make decision analysis,saving the query costs for the business researchers at all levels and speeding up the realization of power data assets.