Superheated Steam Temperature Model Parameter Identification Based On SIS Data

Analysis of real industrial model can not be separated effectively, and the vast majority of industrial processes is a big inertia, wide range of operating conditions , non-linear and time variability of the complex system, asked the very high-precision model . Combination rule based on decomposition of multi-modeling strategy can be largely overcome these deficiencies.Model identification for the overheating problem, the proposed scheme more than the specific task of modeling can be summarized as follows:first, in order to ensure the accuracy and validity of the experimental data for real-time data pre-processing, use of B-Gray Correlation Analysis of selected auxiliary variable thermal system identification with the data treated;second, classification criteria under the conditions, this article put forward an improved double-K means clustering algorithm, this method embodies the original K-means algorithm the advantages of simple and effective, also solve it for a given value and the initial cluster centers K dependent large and not suitable for analysis of large amounts of data issues;third, local models in different conditions were built using the least squares support vector machines (LS-SVM) modeling;last, This paper presents a particle swarm to find the optimal synthesis strategy for the global weights,to solve the problem of local model for the synthesis. This method can be adaptive tradeoff model of global and local characteristics, fast optimal weight vector.In this paper, to Zhejiang Changxing Power Plant Superheated System, selected the SIS day 24 hours of continuous data modeling method used to establish a multi-system model. Experiments conducted in matlab7.1 platform, simulation results have good fitting, verification of the validity of the method. The program for the modeling of complex working conditions provide a viable idea, and has some promotional value.