基于LS-SVM和SPEA2的电站锅炉燃烧多目标优化研究

利用最小二乘支持向量机(LS-SVM)对锅炉燃烧特性建模,构造了以锅炉效率与NOx排放为组合的锅炉燃烧多目标优化模型,并与BP神经网络建模比较,分析表明模型在泛化能力、收敛速度和最优性均优于神经网络模型;针对锅炉高效低污染燃烧多目标问题,提出利用多目标进化算法SPEA2(强度Pareto进化算法)实现运行工况寻优,然后根据模糊集理论在Pareto解集中求得满意解,获得锅炉燃烧优化调整方式.通过某600 MW机组的仿真计算,并与加权遗传算法比较,结果表明本文算法在Pareto前沿具有更好的多样化,克服了将多目标函数加权求和转化为单目标优化问题只能找到凸Pareto最优域及需要多次运行得到Pareto解集的缺陷,计算结果可指导运行人员进行参数优化调整,提高燃烧经济性.

Multiobjective optimization of coal-fired boiler combustion based on LS-SVM and SPEA2

Research on multiobjective optimization of boiler combustion to lower emission and increase its efficiency is presented.Strength Pareto Evolutionary Algorithm(SPEA2) was employed to solve the multiple and conflicting objectives and perform a search to determine the optimum solution of the Least Square Support Vector Machines model(LS-SVM),which was used to set up a boiler combustion response property model for NO_x emission and efficiency,so as to obtain currently optimum combustion adjustment mode of boilers.Comparison with the artificial neural network model shows the superiority of the proposed LS-SVM approach,and confirms that the Multiobjective Evolutionary Algorithm(MOEA) approach can find multiple Pareto-optimal solutions in one single run and this ability makes it attractive for solving problems of multiobjective optimization for boiler combustion.