Research On Top Design Of Earth Observation Satellite System Based On Surrogate Model

Top design of satellite system concerns on how to optimize parameters of the satellite orbitand sensor, the number of satellites, which is aimed to maximize the system performance, meanwhile minimize the cost of the system A reasonable system top design can maximize system performance, especially with the highlighting of satellite effect or the exchanging and in-depth expansion of apply mode, which becomes the urgently-solved problem of application departments. This paper analyses the characteristic and complexity of the issue, and adopts simulation optimization technique based on surrogates to solve the problem. Moreover, we research on several surrogates'relative theory and applicability to different conditions, and then put forward the improved general pattern search algorithm. The main contents and conclusion of this paper are outlined as follows:Firstly, we analyses the characteristics of the satellite system and the complexity of the issue from three aspects: entity, coupled relation and system scale, then a simulation-based approach is presented to evaluate the system performance criteria and impact factor set. The set defines a mapping relationship between the design variables and objective functions.Secondly, Response Surface Methodology, radial basis function and Kriging model's principle are described in details. A non-linear sample is used to analyze the applicability of different surrogates. The results show that Kriging model has the best compute performance about the non-linear function. Then the details of the construction of surrogates are studied, especially for the flowchart and implementation method of simulation analyses, and also the sequence iteration approach to construct surrogates.Thirdly, an improved general pattern search is proposed to the time-consuming simulation design. In search step, surrogates is used to select points in the mesh, and the points based on satisfying the principle of"minimizing response surface and maximizing expected improvement"; in poll step, we use surrogates'percentage improvement to dynamic filter, the maximum point must be selected first. Through the added point, the algorithm also achieves the model updating algorithm.Finally, the feasibility of proposed method is verified by two instances. In single satellite system, the orbit and load parameters are optimized; in multi-satellites system, the applicability to mutil-variable problem about the algorithm is validated. At the same time, the performance between the algorithm and the Analyzer is compared. The results show that the proposed method is better than Analyzer on optimization stability and optimization speed.