基于多族群PSO算法的森林空间动态多目标优化研究

发布时间：2018-11-14 08:27

【摘要】：在实际工程领域存在许多复杂的动态多目标优化问题,其目标函数、约束条件或定义域中至少有一个是随时间变化的,所以,其最优解也会随着时间而变化。多族群PSO算法是采用多个子群并行搜索且子群是动态变化的一种求解多目标优化问题的算法。该算法在求解静态多目标优化问题时是可行的,但是在求解动态多目标优化问题时缺少对环境变化的检测。因此,本文在多族群PSO算法的基础上引入一种环境检测算子,提出了一种改进型多族群PSO算法,并利用改进算法求解森林空间动态多目标优化模型。本文主要研究内容如下:1、提出了一种改进型多族群PSO算法,并对算法性能进行了测试。根据实际问题预设阈值θ,随机选取当前粒子种群中20%个体,计算新适应度值与原适应度值的欧氏距离差的平均值作为环境检测算子ε,通过比较ε和θ的大小判断环境的变化。同时,研究分析和比较五种惯性权重下PSO算法的性能,选择性能最好的一种惯性权重作为改进算法的惯性权重。然后利用四个典型动态多目标优化问题测试改进算法的性能,并与基于改进的距离预测机制及自适应差分算子的动态多目标免疫优化算法PDMIOA和基于动态粒子群算法的动态多目标优化算法dDMS-MOPSO进行对比。试验结果表明文中提出的改进算法在收敛性、分布性以及跟踪性等性能上都优于PDMIOA和dDMS-MOPSO算法。2、构建了森林空间结构动态多目标优化模型。结合单木生长模型中林木生长随时间的变化规律,从林分传统结构、林分空间结构和林分垂直结构三个方面选取林木混交度、竞争指数、角尺度、林层指数、空间密度指数、开阔比数、大小比数、单株材积和健康指数为目标函数,建立森林空间结构动态多目标优化模型,并利用本文提出的改进型多族群PSO算法对该模型进行求解。3、森林空间结构动态多目标优化模型的验证。针对现实林分,以三年为一个林分调整周期,设置三个环境变量,以是否考虑林分调整对林分结构的影响为标准进行两次实验。首先对两次实验下模型的输出结果进行分析,提出了相应的林分调整策略。然后通过对比调整前后林分的均质性指数、混交度、竞争指数、健康度、空间密度指数以及开阔比等指标,发现两次实验下调整后的林分各方面的指标都优于调整前的指标,林分结构明显得到了改善。因此,本文所提出的森林空间结构动态多目标优化模型是可靠的,可为日后智能林业的发展提供相应的技术支持。
[Abstract]:There are many complex dynamic multi-objective optimization problems in the field of practical engineering. At least one of the objective functions, constraints or domain is time-dependent, so the optimal solution will change with time. Multi-population PSO algorithm is a multi-objective optimization algorithm which uses multiple subgroups to search in parallel and the subgroup is dynamic. The algorithm is feasible in solving static multi-objective optimization problems, but it lacks the detection of environment change in solving dynamic multi-objective optimization problems. Therefore, based on the multi-population PSO algorithm, an environment detection operator is introduced, and an improved multi-population PSO algorithm is proposed, and the improved algorithm is used to solve the forest spatial dynamic multi-objective optimization model. The main contents of this paper are as follows: 1. An improved multi-population PSO algorithm is proposed and the performance of the algorithm is tested. According to the preset threshold 胃 of the actual problem, 20% individuals in the current particle population are randomly selected, and the average Euclidean distance difference between the new fitness value and the original fitness value is calculated as the environmental detection operator 蔚. The change of environment is judged by comparing the magnitude of 蔚 and 胃. At the same time, the paper analyzes and compares the performance of PSO algorithm under five inertial weights, and selects one of the best inertial weights as the inertial weight of the improved algorithm. Then four typical dynamic multi-objective optimization problems are used to test the performance of the improved algorithm. It is compared with the dynamic multi-objective immune optimization algorithm (PDMIOA) based on the improved distance prediction mechanism and the adaptive difference operator and the dynamic multi-objective optimization algorithm (dDMS-MOPSO) based on the dynamic particle swarm optimization (DPSO). The experimental results show that the proposed improved algorithm is superior to PDMIOA and dDMS-MOPSO in terms of convergence, distribution and tracking performance. 2. The dynamic multi-objective optimization model of forest spatial structure is constructed. According to the variation law of tree growth with time in single tree growth model, the mixed degree, competition index, angle scale, stand layer index and spatial density index were selected from three aspects: stand traditional structure, stand spatial structure and stand vertical structure. The open ratio, size ratio, volume per plant and health index are taken as objective functions to establish the dynamic multi-objective optimization model of forest spatial structure. The improved multi-population PSO algorithm proposed in this paper is used to solve the model. Verification of dynamic multiobjective optimization model of forest spatial structure. Three environmental variables were set up for a stand adjustment cycle of three years, and two experiments were carried out according to the criterion of whether to consider the influence of stand adjustment on stand structure. Firstly, the output results of the model are analyzed, and the corresponding stand adjustment strategies are proposed. Then, by comparing the indexes of evenness index, mixing degree, competition index, health degree, spatial density index and open ratio before and after adjustment, it was found that the indexes of all aspects of the adjusted stand after two experiments were better than the indexes before and after adjustment. The stand structure is obviously improved. Therefore, the dynamic multi-objective optimization model of forest spatial structure proposed in this paper is reliable and can provide corresponding technical support for the development of intelligent forestry in the future.
【学位授予单位】：中南林业科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18

【相似文献】