Abstract:

According to requirements for accurate prediction and health management of spacecraft, a method of combinational prognosis of parameter values called particle swarm optimization-fuzzy support vector machines is proposed. The method makes respective advantages of particle swarm optimization algorithms, fuzzy mathematics and support vector machines complementary to each other. Incorporating with an example of prognosis of values of output current of the southern solar array of a certain satellite, three evaluation indexes of prognosis, including mean absolute error, mean absolute percentage error and root mean square error, are designed to evaluate prediction results of particle swarm optimization-fuzzy support vector machines at different step-lengths. The result shows that prognosis method of particle swarm optimization-fuzzy support vector machines is effective. The mean absolute percentage errors of particle swarm optimization-fuzzy support vector machines, grey particle swarm optimization neural network model, particle swarm optimization neural network model and grey model are calculated. The result shows that the model of particle swarm optimization-fuzzy support vector machines is most accurate and more efficient in prognosis. It has broad application prospects in the field of prognosis of spacecraft parameters.

Key words: Parameter prediction, Particle swarm optimization, Fuzzy mathematics, Support vector machines