Neural Network Inverse Modelling And Steam Temperature Optimal Control For The Superheated Water-spray Desuperheating System

With the increasing proportion of new energy generation in power grid and the regional grid-centric load management mode,large-capacity thermal power units more and more frequently participate in the deep peak load regulation of power grid,thus a power unit often needs to change load in large scope and work in dynamic load conditions,which means higher requirements for its overall control quality.The water-spray desuperheating system of a supercritical power unit has the characteristics of non-linearity,large time delay and strong coupling,the traditional control method is difficult to guarantee the effect when the unit is in deep peak load regulation,which often leads to the problems of over-temperature and over-regulation time.It may lead to the creep of the boiler heating metal material surface or the creep of the steam pipe speeding,which not only affects the safety of the unit,but also reduces the service life of the unit.The application of intelligent optimization control strategy to improve the control quality of superheated steam temperature is an important research direction of steam temperature control.Based on the analysis of the operation characteristics of supercritical units,the superheated steam temperature control scheme based on neural network inverse control is proposed in this paper.The structure and training methods of several kinds of neural networks are introduced briefly,next the concept of neural network inverse control and the design principle of inverse controller are introduced.By analyzing the influencing factors and the water-spray desuperheating system characteristics of supercritical units,appropriate neural network input and output variables are selected to establish the inverse models with BP?RBF and Elman neural networks.By using the dynamic data of the 600 MW unit,the inverse model offline training is carried out.The most reasonable neural network inverse model is determined,based on which neural network compensation inverse control scheme is studied.Detailed on-line control simulation experiments in large scope dynamic work conditions are carried out by applying the established model.The simulation results show that the neural network inverse control method proposed in this paper effectively improves the control quality of superheated steam temperature,and enhance the flexibility of unit operation with well engineering application prospects.