摘 要：锅炉烟气含氧量是机组运行最重要的参数之一，为了准确测量氧量，在支持向量机（SVM）的基础上，提出最小二乘支持向量机（LSSVM），并结合粒子群算法（PSO）对模型参数（C，g）进行寻优,从而建立锅炉输入和输出变量之间的关系模型。将该方法应用到某电厂600 MW燃煤机组中，用训练后的模型进行预测，并与SVM模型预测结果进行比较。结果表明：采用LSSVM方法，能够辨识出多个变量与氧量之间的复杂关系，对锅炉氧量的预测误差为±0.03；并且PSO-LSSVM预测精度比PSO-SVM模型高，PSO-LSSVM模型具有预测精度高、泛化能力好、鲁棒性强和训练时间较短等优点。

Abstract: The oxygen content in flue gas of power plant is one of the most important parameters.In order to accurately predict the oxygen content，a prediction model was proposed for the boiler using Least Squares Support Vector Machine (LSSVM) and Particle Swarm Optimization (PSO) ,with which the complex relation between input variables and output variable was successfully established.The proposed method was applied to a 600 MW pulverized coal-fired power plant.And the results of LSSVM were compared to those of SVM model.Results showed that the LSSVM method can recognize the complex relationship among many variables and the oxygen content，achieving the accurate prediction of the boiler oxygen content.The prediction accuracy of PSO-LSSVM is higher than that of PSO-SVM model.In summary,the PSOLSSVM model has high prediction accuracy,and excellent generalization and stability,and requires short training time.

TP274.2