摘 要：混合的Box-Jenkins模型能够较好地表征含噪声干扰系统的特性，针对该模型提出了基于粒子群优化算法的过程模型与噪声模型交替估计辨识方法，避免了常规方法直接对过程模型和噪声模型共同辨识容易陷入局部最优的缺陷。仿真试验表明：交替估计算法可以达到对混合Box-Jenkins模型的精确辨识，其误差约为常规辨识的1/10。为验证方法的实用性，以某电厂机组烟气脱硝系统为对象，建立混合Box-Jenkins模型，利用粒子群算法进行交替辨识，所得过程模型与实际输出基本一致，取得了较好的参数辨识效果，该方法可应用到工业过程中这类系统的参数辨识。

Abstract:The hybrid Box-Jenkins model can represent relatively well the characteristics of a noise disturbance system.For the model under discussion,an alternative estimation and identification method for the process model and noise model was proposed based on the particle swarm optimization algorithm and such a defect that it was easy to fall into a local optimum for the conventional method to jointly identify the process model and noise model was avoided.The simulation test results show that the alternative estimation method can attain a precise identification of the Box-Jenkins model.To verify the practicability of the method in question,with the flue gas denitrification system in a power plant serving as an object,a hybrid Box-Jenkins model was established and a particle swarm optimization algorithm was utilized to conduct an alternative identification.It has been found that the process model thus obtained will be basically identical to those of the actual output,thus achieving a relatively good parameter identification result.The method in question can be used for identification of parameters of systems of the same kind in industrial processes.

TP273

上海市"科技创新行动计划"高新技术领域项目（16111106300）