基于GA-SVM的电站锅炉烟气含氧量软测量模型

GA-SVM-based soft-sensor model for oxygen content in flue gas of utility boilers

针对电厂烟气含氧量测量存在的投资大、精度低等问题,在烟气含氧量理论研究的基础上,选择合理的二次变量,引入支持向量机(SVM)建立二次变量与烟气含氧量的软测量模型。利用遗传算法(GA)对模型中的惩罚系数和核函数参数进行寻优,进而利用最优值构建了GA-SVM烟气含氧量软测量模型,并对比利用粒子群算法及网格搜索法对参数的寻优结果,对模型的准确性、泛化性进行测试。仿真结果表明:遗传算法比粒子群算法和网格搜索法更易找到全局最优解;GA-SVM软测量模型误差在±0.2%以内,相对误差在±4%以内,能满足不同负荷、不同时间段锅炉烟气含氧量的预测要求,其对烟气含氧量的测量更准确。

To solve the problems of great investment and low accuracy in measuring the oxygen content in flue gas of utility boilers, the soft-sensor model of secondary variables and oxygen content in flue gas was established by selecting reasonable secondary variables and introducing the support vector machine （SVM）, on the basis of studying the flue gas oxygen content theory. The genetic algorithm （GA） was used to optimize the penalty coefficient and kernel function parameter, and then the GA-SVM soft measurement model of oxygen content in flue gas was built up by applying the optimal value. Compared with the parameter optimization results of particle swarm optimization algorithm and grid method, the accuracy and generalization of the model was tested. The results show that, the genetic algorithm was easier to find the global optimal solution than the particle swarm optimization algorithm and the grid search method. The soft-sensor GA-SVM model had an error within ±0.2% and relative error within ±4%, indicating its measurement result is more accurate and it can meet the prediction requirements for oxygen content in the flue gas under conditions with different loads and operation periods.