Research On Spacecraft Telemetry Data Prediction Algorithm Based On Time Series

The correct prediction is an important basis for scientific decision-making.Due to the complex space environment,it is necessary to predict the running state of spacecraft in order to manage the spacecraft in orbit more efficiently.The changing trend of telemetry parameters can reflect the running state of the spacecraft effectively in space environment.The detailed information about the device status was contained in the spacecraft telemetry parameters.According to the change rule of data information,we can establish a suitable prediction model for the changes of telemetry parameters.The prediction algorithm based on time series is widely used in the field of spacecraft telemetry data.There is a significant relationship between the parameter values and historical parameters in the time series data.And the change of the historical parameters can affect the trend of the future parameters which reflects the parameters with memory.In this thesis,the current commonly used prediction methods are briefly introduced.And the advantages and disadvantages of the prediction model in dealing with linear data are analyzed and summarized.In order to solve the problem of insufficiency of nonlinear data processing,an artificial neural network with nonlinear mapping function is introduced.At present,the development of BP network is the most mature.It has strong and advantages of solving nonlinear data prediction and the efficient nonlinear mapping capability.It does not have the obvious request for predicting parameter.Meanwhile,so long as carries on the effective study of the historical telemetry parameter,can carry on the prediction to the data future change.However,there are also some shortcomings in the standard BP neural network prediction model itself.Aiming at the shortcomings of the algorithm,the corresponding optimization method is proposed.The telemetry data is often more complex in practice and the nonlinear relationship and linear relationship co-exist in a specific time period.Therefore,the thesis puts forward the method of dividing the telemetry data based on time series into nonlinear module and linear module.Because of the decomposition of time series,it can predict the linear main part of telemetry data by linear time series AR model.The disassembled non-linear sequence part will be handled through the BP algorithm in the next step.The final output is superimposed by nonlinear and linear parts.Meanwhile,aiming at the shortcomings of BP network,the genetic algorithm(GA)which as a global optimization algorithm is used to optimize the initial weight and threshold of BP neural network,thus alleviating the problem of BP network easy to fall into a minimum.In this thesis,the prediction model is applied in examples of predicting the change trend of telemetry data.After several simulation experiments,the results show that the AR-BP-GA comprehensive prediction algorithm is conform with the requirements.And the simulation results are better than those using only one linear AR model.Finally,it shows that the proposed comprehensive prediction algorithm is more practical and effective.