针对热工温度对象具有非线性、大惯性、时变性从而导致常规热工模型辨识方法精度不高、适应性差以及标准粒子群算法辨识法易早熟、无法得到全局最优等问题，将细菌趋化、混沌搜索和模拟退火思想引入标准粒子群算法中，对比分析了细菌趋化粒子群算法、混沌模拟退火粒子群算法和模拟退火粒子群算法的辨识结果，探讨了改进算法的工程应用性问题，并将改进算法应用于单入单出的实验室锅炉温度和多入单出的超超临界锅炉主蒸汽温度模型的辨识中。辨识结果表明，细菌趋化粒子群算法可以很好地反映对象的动态特性，提高了辨识精度，缩短了辨识时间，改进了辨识效果。

Aiming at the problems that the characteristics of nonlinear, large inertia, time variation of thermal temperature objects lead to the identification accuracy of conventional model not high and its adaptability poor, traditional particle swarm optimization algorithm easy to be premature and the global optimal solution not be able to obtained, bacterial chemotaxis ,chaotic search and simulated annealing algorithm based on particle swarm optimization algorithm were used. By comparing and analyzing the identification results of bacterial chemotaxis particle swarm algorithm, chaotic simulated annealing particle swarm algorithm and simulated annealing particle swarm algorithm, some engineering application problems were discussed. Furthermore, this method was applied to model identification for the laboratory boiler temperature of the single input and single output system and Ultra Supercritical boiler of the multi input and single output system. The identification results show that the bacterial chemotaxis particle swarm algorithm can reflect the dynamic characteristics of the object，improve the identification accuracy, shorten the identification time and improve the identification effect.

N945.14

上海市电站自动化技术重点实验室课题(13DZ2273800) ；上海发电过程智能管控工程技术研究中心课题(14DZ2251100)