基于DPPSO-FNN算法的再生烟气氮氧化物预测方法

Prediction Method of Nitrogen Oxide Concentration in Regeneration Flue Gas Based on DPPSO-FNN Algorithm

为了解决催化裂化装置再生烟气氮氧化物浓度难以实时预测的问题,提出了一种基于动态参数的粒子群优化-模糊神经网络算法的再生烟气氮氧化物浓度的智能预测方法。基于数据补偿方法实现对催化裂化装置缺失数据段的补遗,弥补了不同参量之间数据尺度不匹配的缺点;建立了基于模糊神经网络算法的氮氧化物预测模型,提取了再生烟气产排过程中的动态特性,实现了输入输出数据的准确表达;设计了一种基于动态参数的粒子群优化算法,提高了算法对模型的优化能力,获得了再生烟气氮氧化物浓度值;最终将氮氧化物预测模型应用于再生烟气的产排过程。实验结果表明该预测方法具有较好的预测精度以及可接受的预测误差,可以满足催化再生器出口氮氧化物浓度的预测需求。

In order to solve the problem that it is difficult to predict the nitrogen oxide(NO) concentration in regeneration flue gas of fluid catalytic cracking(FCC) unit in real time, an intelligent prediction method of NO_(x) concentration in regeneration flue gas based on dynamic parameters particle swarm optimization-fuzzy neural network(DPPSO-FNN) algorithm is proposed. The data compensation method is used to supplement the missing data segment of FCC unit, which makes up for the mismatch of data scales between different parameters. The NO_(x) concentration prediction model based on fuzzy neural network(FNN) algorithm is established, the dynamic characteristics of regeneration flue gas in the process of production and emission are extracted, and the accurate expression of input and output data is realized. A particle swarm optimization algorithm based on dynamic parameters is designed to improve the optimization ability of the algorithm to the model and obtain the NO_(x) concentration in regeneration flue gas. Finally, the NO_(x) concentration prediction model is applied to the production and emission process of regeneration flue gas. The experimental results show that the prediction method has good prediction accuracy and acceptable prediction error, and can meet the prediction requirements of NOconcentration at the outlet of catalytic regenerator.