Data Driven Modeling For Ultra-supercritical Unit Coordinated Control System

At present,ultra-supercritical coal-fired units with its high efficiency and energy-saving advantages have gradually become the mainstream of thermal power unit construction in China.Ultra-supercritical unit is a typical nonlinear multivariable system with high parameters,strong coupling,large delay and large inertia.Establishing a mathematical model suitable for ultra-supercritical unit is helpful for staff to understand the dynamic characteristics of the unit more deeply.The modeling problem has attracted many scholars to study the mathematical model of ultra supercritical units.On the basis of studying the modeling method of a 1000MW ultra-supercritical unit,this paper collects part of the historical operation data from the unit's DCS and makes a deep research on the intelligent modeling method of its coordinated control system from the point of view of linearity and nonlinearity.The main contents of this thesis include:1)Research on modeling methods of unit.Three basic modeling methods such as mechanism modeling and system identification and compound modeling,are introduced in detail,and two intelligent modeling methods such as particle swarm optimization algorithm and neural network,are studied.System modeling is to establish an estimation model equivalent to the system,and the corresponding modeling method should be selected according to the internal mechanism and dynamic characteristics of the syste.2)The simplified transfer function linear model of coordinated control system for ultra-supercritical unit is established by using particle swarm optimization algorithm.Firstly,the transfer function matrix structure model of coordinated control system is obtained by mechanism analysis.Then,the parameters of each sub-model are identified by particle swarm optimization algorithm according to the local operating data of unit DCS system.The transfer function model of the system is obtained and the validity of the model is verified.3)The nonlinear model of coordinated control system for ultra-supercritical units is established by using BP neural network,RBF neural network and improved Elman neural network respectively.A large number of actual operation data of the unit are used to train the structural parameters of each neural network model.The test results show that all the three neural network models have achieved good modeling results,and the improved Elman neural network model proposed in this paper has the best precision,which proves that the neural network has a high engineering application prospect in unit modeling.