Establishment Of Intermediate Point Temperature Prediction Model And Research On Predictive Control For Supercritical Units

Large supercritical units have developed rapidly in China because of their high thermal cycle utilization rate and low pollution.As one of the main temperature parameters of supercritical units,the middle point steam temperature can indirectly reflect the change trend of main steam temperature.By adjusting the middle point steam temperature,the water coal ratio can be effectively adjusted.Therefore,it is of great significance to study the prediction and control of middle point steam temperature for the operation safety and economic benefits of the unit.In this paper,the temperature of a power plant is taken as the research object,and two aspects of work are done on the controlled object: first,the prediction of the middle point steam temperature is studied,and a model based on one-dimensional convolutional neural network(1D-CNN)and multi-layer bidirectional long short memory neural network(Bi-LSTM)is used to effectively predict the middle point steam temperature.Then,the control strategy of the middle point steam temperature is studied.Taking the feedwater flow as the control quantity,the predictive control of the middle point steam temperature is studied theoretically and experimentally.In order to study the predictive control of the middle point steam temperature,this paper uses the improved particle swarm optimization algorithm to identify the model parameters,and uses the generalized predictive control(SSA-GPC)based on sparrow optimization algorithm as the control strategy.Compared with the traditional PID control,the control performance of the system has been greatly improved.The main research work of this paper is as follows:(1)This paper adopts a network model based on 1d-cnn network and multi-layer Bi-LSTM to solve the problem of difficult prediction of steam temperature at the middle point.First,1D-CNN convolution network is used to process the data collected on site,and then the processed data is sent to the built network model for training.During the training process,adma optimizer algorithm is used to optimize the model parameters.In addition,in this model,in order to prevent over fitting of the network model,Variable learning rate(lr)is added in the model training stage.The simulation results show that the model has achieved satisfactory results in prediction speed and accuracy.This experiment has important reference value for studying the combination of middle point steam temperature prediction and depth network.(2)Taking the feedwater flow as the control quantity,the predictive control of the middle point steam temperature is studied.Firstly,the improved particle swarm optimization algorithm is used to identify the parameters of the transfer function model from the feedwater flow to the middle point steam temperature.Then,in view of the shortcomings of the traditional PID control of the middle point steam temperature in thermal power plants,such as large overshoot,long adjustment time and poor antiinterference ability,based on the identified transfer function model,this paper adopts a generalized predictive control strategy based on sparrow optimization algorithm(SSAGPC)to control the middle point steam temperature.Based on the traditional generalized predictive control,the control strategy adopts sparrow optimization algorithm to optimize the weight coefficient and softening factor of the control quantity.Compared with the traditional PID control,the overshoot,adjustment time and antiinterference ability of the system using this control strategy have been greatly improved.