Development Of Unit On-Line Performance Monitoring System And Investigation On Knowledge Discovery In Power Plant Monitoring Data

To meet the increasing demand of enterprise informatization, more and more power plant set up their own Supervisory Information System (SIS). Real time/History Database is the common real time information foundation for SIS, and it can provide some useful applications such as the unit on-line performance monitoring. In addition that, information sent to the database is permanently archived with no loss of data providing a full audit trail. So how to make full use of these data is an urgent task for us. Our aim is to find"Nuggets"hidden in these data by using Data Mining technical.Our work is jointly supported by National Natural Science Foundation of China (NO.50376011) and Erdos Power Plant SIS Project. The main contents and achievements of this thesis are as follows:1,Models for the on-line performance monitoring system for Erdos Power Plant are introduced, including boiler arithmetic models, turbine arithmetic models, and economy diagnose models for each of them.2,Based on the theory of equivalent enthalpy drop, we derived the Loss Equation for the distribution loss which occur in the turbine heater cycle, including the terminal difference loss,the pressure drop loss and so on.3,Partial algorithm theory of equivalent enthalpy drop is developed in this thesis. The superposition,accuracy and calibration of the partial formulas are discussed. The reasonable explanation and verification are also presented.4,The technology of software engineering is used to create and maintain this software. UML (Unified Modeling Language) is used in all the process of this software development, which consist of four stages: use case capture, analysis, design and implementation.5,According to the requirements of operating analysis of thermal power units, data mining algorithm models are introduced. An application framework for Knowledge Discovery in real time/history Database has been proposed.6,The k-means clustering algorithm and its application such as to determine the target value of some main parameters of the unit are introduced in this thesis. Another soft clustering algorithm—Fuzzy c-means and its application are also presented.7,Algorithms of Association Rule are discussed in detail. The quantitative Association Rule can be used to search for the best operate pattern of the equipment in the power plant such as the condenser and so on. Moreover, it can be used to inspect the operation situations of different working groups in power plant. The Boolean Association Rule can be used to find out the connections of the alarm signals in the power plant.8,Some classification Algorithms including ID3 algorithm,C4.5 algorithm as well as Naive- Bayes Classifiers are introduced. These classification Algorithms can be used to diagnose the fault happened in the power plant.