| In order to improve the whole economic benefits and the level of produce management for power plant in our country, the appearance of Supervisory Information System (SIS ) in plant level is inevitable, and the problems correlated with SIS system have already become a kind of hot subject for research. Because complexity of study objects in power plant, nonlinear, coupling relation among parameters, during the process of establishing and developing of SIS functional models exists the problems such as inaccurately, sometime impossible to set up a model. In this thesis, Using the development of "The on-line system of unit performance monitoring and energy loss analysis in power plant " and establishment of optimization problem models in SIS as examples , conventional methods of modeling for solving the problems such as the calculation of unit performance, determination of target value of operation parameters, energy loss analysis as the operation parameters deviates from target value were introduced , and advantages and disadvantages of those methods were also discussed . In order to address the above problems of the conventional modeling methods, the Artificial Intelligence (AI) modeling methods such as Artificial Neural Network (ANN) and Genetic Algorithms (GA) were adopted in the SIS functional model establishment, which has advantage in the modeling of complex objects. The ANN of Self Organization Feature Map (SOM) and BP network were adopted to establish the model of steam turbine power output, with the aid of SOM's ability in clustering, the limitation of conventional collection of training samples were addressed. With the model after trained, the difference between the predicted power output calculated by our model and the actual value generated by steam turbine located in 1.5 percentage, most of them located in 1 percentage. With the established model and the data of power needed in cycle cooling water system, the best vacuum value of condenser can be determined, and energy loss result by deviation of some main operation parameter can also be determined, compared with conventional models , the model established by us has the advantage of pertinence to the actual equipment since the sample data were collected from the operating equipment. To address the optimization problem of unit commitment problem (UCP) in power plant with GA, A refined Float Genetic Algorithms (FGA) with the constrained conditions partially solved combined with punishing function (FGA-PPF) were introduced in this thesis, FGA-PPF refined in the dealing with its constrained conditions, the strategy of mutation , initialization of population of FGA with respect to the features of UCP. FGA-PPF resolved the problem of pre-mature in FGA , in the application to a five units power plant UCP, its succeed rate arrived 100%. The optimization effect of FGA influenced by the parameters such as generations , probability of crossover (Pc), probability of mutation (Pm), coefficient of punishing function and selection pressure were analysed , the analysis result offered a reference in the selection of those parameters for FGA's application . Development method of COM(Component Object Model), which is suit to the configuration of SIS (Supervisory Information System) in power plant ,was introduced to establish the general model of thermodynamic system and its components 。 With those models, a thermodynamic system instance can be created with the configuration of an actual system, By means of communication between the interfaces of COM, performance calculation and energy-loss analysis can be carried out on-line. Finally, method of COM development was introduced with the example of Heater and integrated thermodynamic system. The adoption of COM technique in SIS is conducive to the models' generalization, flexibility, re-usability and functional expansibility.
|
PDF Full Text Request