物理化学学报 >> 2016, Vol. 32 >> Issue (9): 2223-2231.doi: 10.3866/PKU.WHXB201607152

论文 上一篇    下一篇

基于混合型遗传算法的森林可燃物热解动力学参数优化方法

牛慧昌1,2,*(),姬丹2,刘乃安1   

  1. 1 中国科学技术大学,火灾科学国家重点实验室,合肥230026
    2 广州中国科学院工业技术研究院,城市公共安全技术研究中心,广州511458
  • 收稿日期:2016-03-23 发布日期:2016-09-08
  • 通讯作者: 牛慧昌 E-mail:niuhc@mail.ustc.edu.cn
  • 基金资助:
    国家自然科学基金(51476156);广东省科技计划项目(B010118001);广东省科技计划项目(2014B010125003);牛慧昌受中国科学技术大学火灾科学国家重点实验室开放课题(HZ2015-KF10)

Method for Optimizing the Kinetic Parameters for the Thermal Degradation of Forest Fuels Based on a Hybrid Genetic Algorithm

Hui-Chang NIU1,2,*(),Dan JI2,Nai-An LIU1   

  1. 1 State Key Laboratory of Fire Science, University of Science and Technology of China, Hefei 230026, P. R. China
    2 Research Center of Urban Public Safety Technology, Institute of Industrial Technology Guangzhou & Chinese Academy of Sciences, Guangzhou 511458, P. R. China
  • Received:2016-03-23 Published:2016-09-08
  • Contact: Hui-Chang NIU E-mail:niuhc@mail.ustc.edu.cn
  • Supported by:
    the National Natural Science Foundation of China(51476156);Project of Science and Technology Planning of Guangdong Province, China(B010118001);Project of Science and Technology Planning of Guangdong Province, China(2014B010125003);the Open Project of State Key Laboratory of Fire Science, University of Science and Technology of China(HZ2015-KF10)

摘要:

森林可燃物热解动力学参数的优化计算是构建综合热解模型的关键步骤。传统的基于梯度的优化方法收敛速度快但全局寻优能力不足,基于“生物进化理论”的遗传算法具有全局寻优能力但收敛速度慢。本研究首先探讨了单纯的遗传算法对初始值设置的依赖,发现设定合适的初始值能够稳定计算结果,加快算法的收敛速度。针对初始值未知的情况,本文提出了将单纯的遗传算法与迭代算法相结合构建混合型遗传算法的流程。然后以樟子松松枝为例,采用热重分析仪开展了森林可燃物热解实验。假设可燃物热解失重过程遵循三步一级平行反应模型,通过对比单纯遗传算法和混合型遗传算法的收敛过程,发现混合型遗传算法能够快速地获取全局最优的动力学参数,显著地提高遗传算法的优化性能。

关键词: 混合型遗传算法, 非线性拟合, 森林可燃物, 热解, 动力学

Abstract:

For thermal degradation of forest fuels, the optimization of kinetic parameters is a crucial step for the construction of comprehensive pyrolysis model. Traditional gradient-based optimization methods are characterized by strong converging speed, but with weak global optimization capability. The Darwinian survivalof-the-fittest theory based genetic algorithm (GA) is a good tool for global optimization, but with weak converging speed because of the general principles of this algorithm. In this study we evaluated the dependence of the pure GA on the setting of the initial values (IVs), and found that the use of the correct initial values accelerated the converging speed and stabilized the results of the GA. A hybrid genetic algorithm (HGA) was used when the IVs were unknown. This algorithm shares the merits of iterative algorithms and GA. Thermogravimetric experiments were performed using the branches of Pinus Sylvestris and the results were used to compare the converging performances of GA and HGA under the assumption of a three-step, first-order pyrolysis model. The results of these analyses verified the validity and reliability of the HGA.

Key words: Hybrid genetic algorithm, Nonlinear fitting, Forest fuel, Thermal degradation, Kinetics

MSC2000: 

  • O643