基于人工神经网络和模拟进化算法的分级燃烧优化

Staged combustion optimization based on artificial neural network and simulation evolvement algorithm

为控制锅炉燃烧向环境排放NOx造成的污染,提出了分级燃烧技术的综合优化方案。建立了基于人工神经网络及模拟进化算法的100MW火电机组锅炉分级燃烧优化模型,选取16个影响因子进行了分级燃烧的7个可调节参数优化,以达到机组的性能优化目标。锅炉负荷为100%、90%、80%及70%,相应神经网络训练次数分别为11523、14810、13410及19732时满足均方差要求。该神经网络模型优化时采用的种群数为80,交叉概率为0.8,变异概率为0.15。结果表明:锅炉效率和NOx排放量优化计算值同实测值相对误差低于1%;NOx平均排放量由原来的812mg/m3降为645mg/m3。

The staged-combustion burner design was optimized to reduce the environmental pollution due to NO_x emitted by the boiler combustion. The staged-combustion model was based on artificial neural network and simulation evolvement algorithm for a boiler with a 100 MW turbine-generation unit. The boiler performance was optimized for 16 design parameters and 7 regulating parameters that affect the combustion. The analysis considered boiler loads of 100%, 90%, 80% and 70% which required 11 523, 14 810, 13 410 and 19 732 neural network training steps for the training values to meet the mean square deviation requirement. The optimized design gave a species amount of 80, a cross probability of 0.8, and a variation probability of 0.15. The results show that the relative errors in the boiler efficiencies and NO_x emissions between the calculated and measured results are less than 1%, and that the average NO_x output of the boiler decreases from 812 mg/m^3 to 645 mg/m^3.