A novel design of prediction interval for telemetry data

A novel design of prediction interval method for telemetry data based on à Trous recursive prediction algorithm model and modified Block Bootstrap Method (MBBM) is proposed in this paper. The satellite's performance is affected by certain unknown factors and thus the telemetry data shows non-linear and non-stationary characteristics. In this situation, traditional linear and non-linear prediction methods (such as AR, ARMA) are not suitable for such kind of data prediction. In this paper, à Trous recursive algorithm is firstly introduced to solve the problem. By this algorithm, telemetry data series from the satellite can be decomposed into wavelet coefficients point-by-point and the predicting parameters can be adjusted on-line in real-time to give a recursive prediction. Furthermore, modified Block Bootstrap Method (MBBM) proposed in this paper is used to obtain the prediction interval of prediction data to get a more precise and applicable result. The modification enables the traditional block bootstrap method to assess the uncertainty of prognostication without any knowledge about the error distribution of the data predicted by à Trous recursive algorithm. At last, an application of the methods in HY-1B satellite's telemetry data processing is given. The simulation result shows that a more precise prediction is approached using à Trous recursive algorithm and the results of modified bootstrap are well consistent with the satellite's final discharge pressure characteristics.