Superheated steam temperature prediction method based on multi-innovation stochastic gradient optimization

Abstract

The invention relates to a superheated steam temperature prediction method based on multi-innovation stochastic gradient optimization, and the method comprises the following steps: 1), dividing working conditions according to the load of a unit, and enabling the inlet steam temperature and outlet steam temperature data of a final-stage superheater to serve as a training sample TX during the normaloperation of the unit under each working condition; 2) preprocessing the training sample TX to enable the mean value of each variable to be 0 to obtain an input matrix; 3) constructing a superheatedsteam temperature prediction Hammerstein nonlinear identification model and determining model parameters which need to be identified; 4) identifying model parameters to be identified by adopting a multi-innovation stochastic gradient identification method; and 5) inputting to-be-predicted final-stage superheater inlet steam temperature data into the superheated steam temperature prediction Hammerstein nonlinear identification model after parameter identification to obtain final-stage superheater outlet steam temperature prediction data, and completing prediction of the superheated steam temperature. Compared with the prior art, the method has the advantages of being high in prediction precision, suitable for large inertia and highly nonlinear data and the like.

Description

technical field [0001] The invention relates to the technical field of generator set operation control, in particular to a superheated steam temperature prediction method based on multi-innovation stochastic gradient optimization. Background technique [0002] The superheated steam temperature is a very important control parameter during the operation of the generating set, and its stability has a great influence on the safe and economical operation of the generating set. At present, coal-fired units are developing towards high parameters and large capacity. The characteristics of large inertia and high nonlinearity of superheated steam temperature are becoming more and more obvious, which increases the difficulty of superheated steam temperature control. Traditional cascade PID often cannot control steam temperature well. Therefore, a superheated steam temperature prediction method considering the large inertia and highly nonlinear characteristics of superheated steam tempe...

AI Technical Summary

Problems solved by technology

At present, coal-fired units are developing towards high parameters and large capacity. The characteristics of large inertia and high nonlinearity of superheated steam temperature are becoming more and more obvious, which increases the difficulty of superheated steam temperature control. Traditional cascade PID often cannot control steam temperature well. Therefore, a superheated steam temperature prediction method considering the large inertia and highly nonlinear characteristics of the superheated steam temperature is needed

[0003] Moreover, the current main parameter estimation methods include identification methods based on intelligent optimization algorithms, identification methods based on auxiliary models, and identification methods based on stochastic gradients. However, in superheated steam temperature prediction, intelligent optimization algorithms converge slowly and are prone to fall into local The optimal solution leads to inaccurate identification of the parameters. The stochastic gradient algorithm uses the gradient search principle to estimate the parameters, but the calculation amount is small, but the convergence speed is slow. To solve the above problems, it is also necessary to improve the identification speed and accuracy of the prediction model

Images