Affiliation of Author(s):计算机科学与技术学院/人工智能学院/软件学院

Journal:IAENG Int. J. Comput. Sci.

Abstract:Power system is an essential system in satellite, which ensures the security and stability of energy in the whole satellite system. This paper presents a mixed relevance vector machine with modified particle swarm optimization (MPSO-RVM) algorithm to forecast parameters intervals of satellite power system involved the main bus load current and the main bus voltage. First, RVM with radial basis kernel function is established to solve the regression problems of the data in satellite power system. Next, modified PSO algorithm is utilized to find out the optimal parameters of RVM to enhance the generalization capability. In addition, the self-adaptive parameter setting mechanisms is conceived to avoid the MPSO algorithm trapping into the local optima. Moreover, MPSO-RVM model can obtain desirable prediction intervals rather than prediction values. Experimental results demonstrate that MPSO-RVM model can achieve better prediction accuracy, sparser solution and shorter test-time than RVM model and PSO-SVR model. Meanwhile, the majority of samples are located into the prediction interval obtained at higher confidence level. Therefore, the proposed MPSO-RVM model vividly depicts the variation tendency of parameters in satellite power system, which is conducive to adopt available measures for avoiding satellite accidents and faults initiatively.

ISSN No.:1819-656X

Translation or Not:no

Date of Publication:2017-01-01

Co-author:康旭

Correspondence Author:Pi Dechang