Research And Application Of Data Validation And Optimization Value For Unit Performance Optimization System

With the rapid development of the power industry in our country, the integration level of information in power plants will be higher and higher. Large quantity of information that reflects the condition of unit performance, is provided by mass sensor data and the experience of experts. Making use of the method of data mining to treat and apply these precious information resource, will have the important and realistic meaning to operation optimization and economical efficiency, stability, reliability...etc. of the unit. Data validation and optimization target value in the Unit Performance Optimization System (UPOS) are studied in this thesis, using the technology of data mining. The essential achievements can be described as follows:Firstly, the thesis introduces the basic theory of data mining and emphasizes the theory of data validation, association pattern and regression pattern in data mining. Secondly, the using of Principal Component Analysis (PCA) and novel Robust Auto Associative Network (RAAN) for data validation is expatiated and improved in this thesis. The results show that the two methods are effective in data validation. And at the same time, six kinds of sensor failure modes are summed up. The novel RAAN can successfully detect sensor failure and recover signals for many reasons. Then, the method of applying association rules and regression analysis into the decision of optimization value is introduced in details. And this method is effective in forecasting optimization values of parameters under the unit's current condition. Finally, data validation and optimization value finding software is made and integrated into UPOS according to the theory of data mining and knowledge engineering. And many new ideas and algorithms are presented. This software can support UPOS to realize the unit performance optimization.