计及先验知识及烟温偏差约束的四角切圆锅炉燃烧优化

Combustion Optimization for Tangentially Fired Boiler Considering Priori Knowledge and Gas Temperature Deviation

对某660 MW切圆燃烧锅炉进行了单因素调整试验。基于试验数据,建立强化趋势学习的先验支持向量机(先验SVR)模型,实现了CO、飞灰含碳量、NO_x浓度、烟温偏差等参数的软测量。结果表明,先验SVR模型比基于交叉验证法的PSO-SVR模型具有更好的泛化能力,建模平均相对误差降低2%左右,能正确地描述燃烧特性。在此基础上,采用多目标遗传优化算法,计算不同烟温偏差限值条件下的锅炉效率最高、NO_x最低的Pareto前沿线,推荐运行方式的烟温偏差较传统燃烧优化结果降低25℃,电站锅炉在安全范围内能够更加经济、环保地运行。

Single-factor experiments were performed on a 660 MW tangentially fired boiler.Based on the experimental data,the priori support vector machine models with strengthening trend learning were established for the soft sensor of CO,the carbon content of fly ash,NO_x emission,gas temperature deviation,and so on.Compared with PSO-SVR model based on cross verification,it was concluded that the priori support vector machine model has better generalization ability.The average relative error of modeling was reduced by 2%,and the model can describe the combustion characteristics correctly.Then,multi-objective genetic optimization algorithm was chosen to calculate the Pareto front line with the highest furnace efficiency and the lowest NO_x under different gas temperature deviation.It shows that gas temperature deviation of the recommended operating mode was 25℃ lower than the traditional combustion optimization result,providing instructions for utility boiler to be operated more economically and environmentally friendly.