基于改进云自适应粒子群优化算法的NOx含量测量

摘要

脱硝反应器入口NOx浓度的及时、准确测量,对精确调节喷氨量、控制氮氧化物的排放至关重要.针对NOx气体分析仪测量存在的精度差、滞后性等问题,基于传统云理论,并结合径向基函数(RBF)神经网络,提出了改进的云自适应粒子算法(CPSO)-RBF神经网络的测量模型.利用云模型理论中云滴具有随机性、稳定倾向性等特点,提出了一种新型分段式自适应调整粒子群惯性权重算法.利用此优化算法,对神经网络参数进行优化,提高了测量模型的精度.将该模型应用于SCR反应器入口的NOx含量测量中,实例仿真表明,改进算法优化的神经网络模型具有较高的精度,为反应器入口NOx含量的实时、准确测量提供了一定的理论依据,也为实际生产过程中NOx的测量与控制提供了一定的参考.%The timely and accurate measurement of NOx content at inlet of denitrification reactor is very important to accurately adjust the amount of ammonia spray and the NOx emission control.Aiming at the problems of the serious delay and poor precision of the NOx gas analyzer,based on traditional cloud theory and combining with the radial basis function(RBF) neural network,the measurement model based on CPSO-RBF neural network is proposed.By using the features of cloud droplets,i.e.,randomness and stable tendency,the new type of segmented adaptive adjustment particle swarm inertia weight algorithm is proposed.The parameters of neural network are optimized using this optimization algorithm,thus the accuracy of the measurement model is enhanced.The model is applied in the NOx measurement at the inlet of SCR reactor,the simulation of practical example indicates that the neural network model optimized by the improved algorithm features high accuracy,it provides certain theoretical basis for real time and precise measurement of NOx at inlet of the reactor;and certain reference for NOx measurement and control in practical production process.