论文部分内容阅读
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。
The combustion characteristics of the boiler were modeled by least square support vector machine (LS-SVM), a boiler combustion multi-objective optimization model based on boiler efficiency and NOx emission was constructed and compared with the BP neural network model. The analysis shows that the model The generalization ability, the convergence speed and the optimality are all better than the neural network model. Aiming at the multi-objective problem of high efficiency and low pollution combustion in boiler, the multi-objective evolutionary algorithm SPEA2 (strength Pareto evolutionary algorithm) is proposed to optimize the operating conditions. Set theory in the Pareto solution set to find the satisfactory solution, get the boiler combustion optimization adjustment. The simulation results of a 600 MW unit are compared with those of the weighted genetic algorithm. The results show that the proposed algorithm has better diversification in the Pareto front, which overcomes the problem that the weighted summation of multiple objective functions can be transformed into a single objective optimization problem, Optimal domain and the defects that Pareto solution set needs to be run multiple times. The calculation results can guide the operators to optimize the parameters and improve the combustion economy.