Research On Online Identification Technology In Intelligent Power Generation

With the development of China’s power industry,in response to the "Made in China 2025" policy,major power plants are gradually upgrading their traditional Distributed Control System(DCS)to smart power generation.Smart power generation is closely linked with artificial intelligence,using high-performance computers and high-speed communication networks to achieve a deep integration of industrialization and informationization of power generation.Accurate modeling under variable operating conditions is the basis for optimization and control of various complex production processes in smart power generation.The complex systems in power generation processes are multivariable,nonlinear and highly coupled,and it is difficult to realize the modeling of nonlinear systems by traditional mechanism modeling,which can be effectively solved by using intelligent algorithms for system identification.Therefore,model identification and corresponding control strategies have become a hot and difficult research area in the field of power production process control.Offline identification has a high accuracy of parameter estimation,but has the disadvantages of large amount of stored data,large amount of operation and high complexity of algorithm.In addition,off-line identification cannot reflect the system working condition change adaptively.The research of online identification of the system is a necessary path for the development of intelligent power generation.In order to realize the application of online identification in intelligent power generation,this paper firstly investigates the online identification of Takagi-Sugeno(T-S)fuzzy model based on entropy clustering,which is an online clustering algorithm that can realize the online identification of the system in combination with the recursive least squares method.However,this algorithm also relies on a priori knowledge and expert experience for the selection of input variables of the system.In this paper,we use the algorithm of selecting significant input variables to perform offline machine optimization of possible input variables before online identification of flue gas oxygen content in thermal power units,to improve the accuracy and reduce the time complexity of the online identification algorithm,and to better apply the system online identification to intelligent power generation.