基于声波测温和LSSVM_GA的电厂NO_x排放建模与优化

Modeling and optimization for NOx emission of power stations based on acoustic temperature measurement and LSSVM_GA

因传统燃烧优化实验控制电厂NOx排放的方法很难满足复杂多变的燃烧工况,为更智能地对NOx排放进行监测和更方便地对其进行优化,对某电厂2#炉300 MW工况下NOx排放优化实验时的DCS内运行数据和声波测温系统内的温度分布数据进行采集。利用最小二乘支持向量机,以炉膛温度信息和其他影响NOx排放的主要因素为输入,以NOx排放浓度为输出建立NOx排放预测模型,在预测模型的基础上利用遗传算法对顶部4层分离燃尽风开度进行直接寻优,达到降低NOx排放的目的。结果表明:加入炉膛温度信息后的NOx排放模型准确度更高,遗传算法优化之后的NOx排放浓度显著降低,优化后参数更符合工程实际。

Traditional methods to control NOx emissions through combustion optimization experiments can hardly meet complicated and changeable combustion conditions now. For more intelligently monitoring and better optimizing NOx emissions, data within DCS and temperature profile within the acoustic measurement system of furnace 2#in a power station under 300 MW working condition are collected during the optimizing experiment of NOx emissions. That is,a least squares support vector machine is used to create a prediction model for NOx emission based on furnace temperature information and other factors that affect NOx emissions as input value and NOx emission concentrations as output value. Apart from the model, a genetic algorithm is applied to optimize the opening of over fire air of four-layer separation on top so as to reduce NOx emissions. The results show that the NOx emission model is more accurate when furnace temperature information is added,and the NOx emission concentration is significantly reduced and the parameters are more suitable for engineering practice after the opening is optimized through the genetic algorithm.