王海军, 曾浩然, 张文婷, 张彬, 周玲慧. 基于改进的GPU并行NSGA-Ⅲ的土地利用优化配置[J]. 农业工程学报, 2020, 36(21): 283-291. DOI: 10.11975/j.issn.1002-6819.2020.21.034
    引用本文: 王海军, 曾浩然, 张文婷, 张彬, 周玲慧. 基于改进的GPU并行NSGA-Ⅲ的土地利用优化配置[J]. 农业工程学报, 2020, 36(21): 283-291. DOI: 10.11975/j.issn.1002-6819.2020.21.034
    Wang Haijun, Zeng Haoran, Zhang Wenting, Zhang Bin, Zhou Linghui. Land use optimization allocation based on improved NSGA-Ⅲ by GPU parallel computing[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2020, 36(21): 283-291. DOI: 10.11975/j.issn.1002-6819.2020.21.034
    Citation: Wang Haijun, Zeng Haoran, Zhang Wenting, Zhang Bin, Zhou Linghui. Land use optimization allocation based on improved NSGA-Ⅲ by GPU parallel computing[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2020, 36(21): 283-291. DOI: 10.11975/j.issn.1002-6819.2020.21.034

    基于改进的GPU并行NSGA-Ⅲ的土地利用优化配置

    Land use optimization allocation based on improved NSGA-Ⅲ by GPU parallel computing

    • 摘要: 土地利用优化配置是实现土地资源可持续利用的重要途径。该研究根据地理单元发展演变特点对基于参考点的非支配排序的遗传算法进行针对性改进,并耦合多目标优化方法,构建了土地利用空间优化模型。针对目前空间优化模型耗时过长,效率低的问题,该研究将GPU(Graphics Processing Unit, GPU)并行计算和土地利用优化配置模型有机结合,提升模型的优化效率。选取武汉市东西湖区进行实证研究,对比了模型在CPU(Central Processing Unit, CPU)串行计算和GPU并行计算2种方式下的运行耗时,并从最终优化结果中选取生态保护优先和经济发展优先2种典型方案进行分析。结果表明:1)GPU并行计算能够显著提升模型的优化效率,模型运行耗时由原来的158.08 h缩短到了1.68 h;2)模型能够统筹协调多个目标,对研究区域土地的数量结构和空间布局进行合理配置,为规划决策者提供多个可行方案。生态保护优先方案中,生态效益降低了6.16%,经济效益增长了13.64%;经济发展优先方案中,生态效益降低了6.19%,经济效益增长了15.86%。

       

      Abstract: The contradiction between supply and demand of land resources has become increasingly prominent, as the rapid development of urbanization. This problem has hindered the improvement of urbanization and development quality. The optimization of land-use allocation can bean important approach to coordinate the limited land resources, and thereby to ensure the high-quality development of a city. This study aims to establish a spatial optimization model of land use via a multi-objective optimization model with NSGA-III. A multi-objective model consists of the main and the constraint objectives. The main objectives include the maximization of GDP value, the maximization of ESV, the minimization of changing cost from the status que, and the minimized incompatibility of land use types. Besides, the constraint objectives are comprised of 5 quantitative constraints and 4 spatial constraints dataset according to policy planning. The NSGA-III can be well used to solve the multi-objective space optimization of land use, due to its excellent ability of global optimization and spatial search. The recombination and mutation operator were improved, based specifically on the characteristics and developments of geographical units. The efficiency of modified model was improved remarkably via integrating the GPU parallel computing. The Dongxihu District of Wuhan, China, was taken as the study area to test the model. Two typical schemes, including ecological and economic priority, were analyzed to compare the time-consuming of model in the serial computing of CPU and parallel computing of GPU. Consequently, the results demonstrated that: 1) A better optimization efficiency of modified model can be obtained using the GPU parallel computing, where the computing time reduced from 158.08 hours to 1.68 hours. 2) The modified model can be used to coordinate multiple objectives, and thereby to reasonably optimizing land use in terms of quantity structure and spatial pattern, providing for the multiple selections indecision making. In the scheme of ecological priority, the ecological benefits of study area reduced by 6.16%, and the economic benefits increased by 13.64%. In the scheme of economic priority, the ecological benefits reduced by 6.19%, and the economic benefits increased by 15.86%.

       

    /

    返回文章
    返回