一种用于陀螺随机漂移预测的多尺度混合建模方法

摘要

针对陀螺随机漂移时间序列由于非平稳和非线性造成单一预测模型难以准确跟踪其变化趋势的问题，提出了一种基于集合经验模态分解(EEMD)和灰色极端学习机(GELM)的多尺度混合建模方法。首先，利用集合经验模态分解将随机漂移时间序列按照频率高低分解为多个本征模式分量和一个余量；然后针对不同类型时频特性分量选择合适激活函数和隐层神经元数目的GELM分别进行预测；最后，以等权相加的方式得到最终预测结果。将该方法用于某型激光陀螺随机漂移预测中，仿真结果表明：混合预测模型能够准确预测陀螺随机漂移，预测精度比残差GM（1,1）和GELM预测模型分别提高了33.43%和23.47%，可为激光陀螺的漂移补偿、故障预报和可靠性诊断提供依据。%In view that the time series of gyro random drift can not be precisely predicted by single forecasting model due to its non-linear and non-stationary characteristics, this paper proposes a hybrid multi-scale modeling method based on ensemble empirical mode decomposition (EEMD) and grey extreme learning machine(GELM). Firstly, the drift error data is decomposed into a series of intrinsic mode function and one residue via EEMD; Secondly, GELM predicting models with appropriate activation functions and hidden nodes are constructed to predict each intrinsic mode function and residue respectively;In the end, the outputs of each predicting model are added with equal weight to obtain the final prediction result. By using the proposed method for a laser gyro random drift prediction, the experiment is made which shows that the hybrid prediction method can get more precise result than remanet GM(1,1) and GELM prediction models, whose prediction accuracy increases 33.43% and 23.47% respectively. The hybrid model could provide reliable evidence for drift compensation, fault prediction and reliability diagnoses of laser gyro.