【作者】 刘晓路；

【导师】 陈英武；

【作者基本信息】 国防科学技术大学 ， 管理科学与工程， 2011， 博士

【摘要】 对地观测卫星系统( Earth Observation Satellite System, EOSS )是进行地面信息收集的主要太空平台,具有覆盖区域广、安全性能高、不受空域国界、地理条件限制等优点。过去几十年中,我国对地观测卫星无论数量、还是质量都有极大的提升,但系统的整体性能并不高,尤其在应急灾害响应、重点区域覆盖等方面能力较差。本文提出了EOSS顶层设计参数优化问题,通过优化系统的关键性能参数配置,实现对EOSS整体性能的改进和提高,尤其是系统的快速响应能力和重点区域观测能力,围绕此文章的主要工作可概括为:首先,提出了对地观测卫星系统顶层设计参数优化问题,并从应用角度分析了问题的来源和研究意义。同时从EOSS的物理组成和性能计算两个角度分析了问题的复杂性,在此基础上,结合试验优化与仿真优化的思想,分析了求解中所存在的问题,提出了基于试验设计和代理模型的求解框架。EOSS通过对特定目标成像以实现对用户需求的满足,本文将EOSS性能归结为覆盖性能。在点覆盖数字仿真方法基础上,从空间和时间两个方面构建了EOSS的性能指标体系,给出了每个指标的计算方法,研究了影响这些指标的系统因素,并通过定量分析验证了各个因素对不同指标的影响。在求解EOSS顶层设计参数优化问题的过程中,系统的性能指标需要借助仿真进行计算。为了尽可能控制仿真次数,提出了综合拉丁方试验设计方法(CLHD)生成仿真方案,该方法保证所选择的采样点均匀充满整个设计空间,同时保证不同采样点之间相关性较小。论文分析了拉丁方矩阵的优化准则,定义了多目标的收益函数,采用快速模拟退火算法控制方案的生成过程,用正交性较好的拉丁方矩阵作为初始解,算法分别构建了4种变换邻域用以生成改进解和跳出局部最优,采用服从Cauchy分布的降温函数模拟算法的退火过程。通过与其它方法的对比发现,CLHD能够兼顾设计方案的正交性和均匀性,且对试验因素数、因素水平数之间无特殊的要求。针对EOSS仿真所产生的仿真数据,本文提出了基于多点更新的Kriging代理模型对其进行拟合分析;通过代理模型最优与最大化期望提高相结合的机制选择代理模型插值点,提出了基于函数改进的插值点过滤机制,构建了基于改进广义模式搜索算法的Kriging代理模型构建和优化框架。其中,搜索步通过遗传算法和序列二次规划算法实现,筛选步通过不完全动态筛选实现。在此基础上,引入多个标准测试函数对所提方法进行了验证,计算结果证明本文方法具有较好的全局搜索能力,能够有效地跳出局部最优。最后,论述了本文的仿真结构、仿真流程、想定设计和优化过程。根据实际应用设计了两个仿真实例:以最大化区域覆盖率为目标,设计了面向突发事件的小卫星应急发射实例;以最小重访时间为目标,设计了面向全球重点区域覆盖的实例。这两个实例一方面是对本文方法求解流程的细化,另一方面也是对方法有效性的验证,通过与通用商业软件STK/Analyzer优化结果的对比可以发现,相同条件下本文方法所得解的质量更高。更多还原

【Abstract】 As main ground information collecting platforms in space, earth observation satellite systems (EOSS) are charactered by broad coverage, high security performance and unlimited to airspace and national boundaries. As a result, they play outstanding roles in both military demands and civil applications, owning extrusive strategic significance and importance. In the past decades, earth observation satellites in China have been developed greatly, which are incarnated on not only the number but also quality of the satellites. However, the holistic system performance is not high; especially the capabilities of fast response and coverage for important area are bad. Therefore, we propose problem of top design parameters optimization for EOSS.Through configuring system key parameters in a reason way, we can improve performance of EOSS. Based on the research background mentioned above, the main work and innovative points in this paper are as follows.Firstly, bring forward the problem of top design parameters optimization for EOSS, introducing the necessaries of our research from the view of engineering. Also we analyze the complexity of the problem from both physical structure and its performance computation. Grounded on the analysis, a simulation based optimization method is put forward. More precisely, a framework which incorporated design of experiment and surrogate models is constructed.EOSS is designed to image particular targets to fulfill different user requests. Hence, performances of EOSS are mainly defined as coverage performance. By adopting the method of point based numeral simulation we can research performance of EOSS. And a series of measures, which can be divided into two groups: space and time, are proposed to describe the capability of EOSS. Also, we figured out system variables that might affect proposed measures. And many simulation tests are carried out to validate effects of different variables.According to the above discuss, to optimize top design parameters of EOSS we need simulation to calculate its performance. To create simulation plans with fewer test points, we proposed comprehensive latin hyper-cube (CLHD) method. The plans generated by CLHD ensure its sampling points are uniformly distributed in the design space and correlations between points are small. To realize CLHD, we first analyze optimal criterions for Latin hypercube array and define a multi-objective benefit function. The main body of CLHD is carried out by very fast simulated annealing (VFSA) algorithm, in which four transforming neighborhoods are defined and its annealing schedule follows a Cauchy function. Arrays with good orthogonality, which are constructed by Cholesky decomposition, are used as initial solution for VFSA. Finally, many instances are introduced to testify the effectiveness of CLHD. We also analyze the influence of different system variables.EOSS simulation will produce abundant simulation data. And a multi-point updated Kriging surrogate model is constructed to simulate and approximate these data. Points with optimized value or maximal expected improvement are selected to update our surrogate model. And a measure named objective improvement versus distance is defined to filter the selected points. To get optimized solution of the surrogate model, we construct improved generalized pattern search algorithm. In search step, genetic algorithm and sequential quadratic programming are used to find potential update points. In poll step, dynamic incompletion poll is carried out to find points with greater value. And several benchmark functions are used to test the proposed method and the results show that our method own an outstanding global search ability.Finally, we address the simulation structure, flow of EOSS and detail the scenario design and process of system optimization. Then two application examples are designed. One is back grounded on the fast launch of micro-satellite, which is to maximize the coverage percentage of target area. The other is for global coverage and its objective is to minimize revisit time of EOSS. These two instances are used to illustrate the flow of our method. They also testify the validity of the method we put forward. Comparing to the results got by Analyzer, our method is more effective.更多还原