Research On Intelligent Diagnosis Of Water Supply System Based On PI Database

The safe operation of large coal-fired power plants can not be separated from the analysis and monitoring of a large number of production data.Plant Information System(PI)is a set of software modules for monitoring and analysis of the whole plant.In 2016,the total installed capacity of thermal power generation in China has reached 990 million kilowatts,accounting for 74.6%of the total installed capacity of power in China,while the coverage of plant information system is close to 100%.Such a large market share has brought tremendous impetus to the development and innovation of factory information system in China.More and more thermal power plants begin to invest human and financial resources in the secondary development and application of PI system.Taking the feed water system of Huaneng Dalian Power Plant as the research object,this paper analyses the function,structure and working principle of the feed water system,and introduces the operation principle and equipment parameters of its main core equipment:steam-driven feed water pump,electric feed water pump and regenerative heater.Through investigation and research,it gives the common difficulties in the operation and maintenance of the feed water system.Question.This paper uses VB as development platform and SQL Server as PI database.Through PI-SDK interface,the platform applies for industrial real-time parameters of water supply system of unit to PI database as diagnostic basis,uses SQL Server database instruction set to apply historical data batch to PI database through PI-ODBC mode,as a typical fault database and establishes corresponding fault mathematical model,and as training data set of neural network regression and classification model,and through adjustment.The whole network makes it the best expression.In this paper,three models are established respectively by using real-time data of industrial parameters of water supply system obtained from PI database:using least square method and variance analysis method to obtain abnormal fluctuation of motor current of electric feed water pump;using neural network regression model to model outlet flow of steam feed water pump,real-time detecting whether the outlet flow of feed water pump is in the prediction interval or not,after the evaluation difference appears.Real-time alarm is given to the front desk,and neural network classification model is used to model the leakage state of the water side of the heating system,and real-time detection of potential safety hazards in the heating system is carried out.The alarm helps operators to find abnormalities early,and advance the time node of human intervention,effectively preventing major accidents caused by minor failures,greatly guaranteed the safe and stable operation of the unit.