基于自适应遗传算法的锅炉低NO_x燃烧建模及其优化

Adaptive genetic algorithm based low NO_x combustion modeling and optimization for boilers

为了对锅炉NOx排放进行优化控制,分析现场运行数据,建立了基于支持向量机(SVM)的NOx排放特性模型,采用改进的自适应遗传算法对模型参数进行寻优,并比较了基于Sigmoid核函数和RBF核函数的SVM模型的性能。结果表明,RBF核函数更具有优势,且SVM算法有良好的泛化能力和预测精度。结合SVM建模和自适应遗传算法寻优,对锅炉可调参数进行优化,使该炉的NOx排放量从423mg/m3降低至219.4mg/m3,下降幅度达到了48.13%。通过上述方法可得到低NOx排放的最佳运行参数组合,为电站锅炉的运行优化指导和NOx排放控制提供参考和依据。

In order to optimize the NOxemission from boilers,field operation data were analyzed and the NOxemission model based on support vector machine（SVM）was established.The modified adaptive genetic algorithm was applied to optimize the model parameters.Moreover,the performance of the Sigmoid kernel function based SVM model and that of the RBF kernel function based SVM model was compared.The results indicate that,the RBF kernel function based SVM model has more advantages.The SVM algorithm has good generalization ability and prediction precision.The adjustable parameters of the boiler were optimized through combining the SVM modeling and the adaptive genetic algorithm,which reduced the NOxemissions from 423mg/m3 to 219.4mg/m3,by about 48.13%.