Study on Optimization of Prediction Model of Flue Gas Oxygen Content in Thermal Power Plant Based on Least Squares Support Vector Machine

Abstract

The oxygen content in flue gas is an important monitoring parameter of boiler operation, and also an important basis that reflects the perfection of combustion equipment and boiler operation. Quickly and accurately measuring the oxygen content in flue gas according to operating conditions has important guiding significance for optimizing boiler combustion process. Based on the operation data of a 1 000 MW ultra-supercritical boiler in a power station, 31 factors affecting flue gas emission were selected, and cross validation (CV), particle swarm optimization (PSO) algorithm, and genetic algorithm (GA) were used to find the optimal parameters of the least squares support vector machine (LSSVM) model, and the prediction model of flue gas oxygen content was established. The research results show that, compared with PSO-LSSVM and CV-LSSVM models, GA-LSSVM prediction model has better prediction ability for flue gas oxygen content, and has the advantages of high prediction accuracy, good generalization ability, and strong robustness. The relative error and mean square error of fitting prediction are 0.54% and 0.23% respectively, and the relative error and mean square error of generalization prediction are 1.66% and 2.13% respectively. It can accurately measure the oxygen content of boiler flue gas in thermal power plants, which lays a foundation for further optimal operation of boiler combustion system.

Keywords

• thermal power plant
• least squares support vector machine (lssvm)
• particle swarm optimization (pso) algorithm
• genetic algorithm (ga)
• cross validation (cv)