基于微分进化和SMO算法的烟气含氧量软测量

火电厂中烟气含氧量是一个难以测量的量,采用支持向量机中的序列最小优化(SMO)算法对其进行软测量,并采用改进的SMO算法提高建模速度.SMO算法的性能很大程度上依赖于其学习参数,选择合适的SMO参数是一个亟待解决的问题.而微分进化算法(DE)具有很强的全局搜索能力,在多峰函数的寻优问题上已表现出优异的性能.为此,采用DE算法选择SMO的参数,提出了基于DE算法的SMO参数选择方法.仿真表明,该方法能够准确预测烟气含氧量的变化,比用遗传(GA)算法和粒子群(PSO)算法优化SMO参数具有更高的精度和更快的速度.

Soft Measurements of Flue-gas Oxygen Content Based on Differential Evolution and SMO Algorithm

At present,the flue-gas Oxygen Content is difficult to measure.The SMO(Sequence Minimum Optimization) algorithm,one method of support vector machine(SMO),is used for the soft measurement of power plant flue-gas oxygen content in this paper.The performance of the SMO algorithm relies heavily on their learning parameters,choose appropriate SMO parameters is a problem to be solved.Differential evolution algorithm(differential evolution,DE) is a population evolution optimization algorithm based on real number coding,with strong global searching capability,in the multi-modal function optimization problem has shown the outstanding performance.So the DE algorithm is used for SMO parameter selection,putting forward the SMO parameter selection method based on DE algorithm.The simulation results show that this method under discussion can forecast accurately the flue-gas oxygen content,and has a higher accuracy than GA and PSO algorithms,which has major significance for realizing economic combustion of the thermal power plant.