MISO系统辅助模型多新息增广随机梯度算法

Multi-innovation Extended Stochastic Gradient Algorithm Based on an Auxiliary Model for MISO Systems

对于有色噪声干扰的输出误差多输入单输出(MISO)系统,常规的递推最小二乘辨识方法给出的参数估计是有偏的。为了提高随机梯度辨识方法的收敛精度和速度,用辅助模型的输出代替辨识模型信息向量中的未知不可测变量,推导出其辅助模型增广随机梯度辨识算法;再引入新息长度扩展标量新息为新息向量,提出了基于辅助模型的MISO系统多新息增广随机梯度辨识算法。所得算法在每一次的迭代中不仅使用了当前数据和新息,而且使用了过去数据和新息,提高了参数估计精度和收敛速度。仿真例子验证了算法的有效性。

For multiple-input single-output（MISO） output-error systems,estimates from the conventional recursive least squares parameter identification method are biased.In order to enhance accuracy and speed of convergence for stochastic gradient identification algorithm,an auxiliary model based multi-innovation extended stochastic gradient（AM-MI-ESG） algorithm is presented by replacing the unknown unmeasurable variables in the information vector with the outputs of the auxiliary model,and introducing the innovation length and expanding the scalar innovation to an innovation vector.The AM-MI-ESG algorithm uses not only the current data and innovation but also the past data and innovation at each iteration,thus the parameter estimation accuracy and convergence rate can be improved.The simulation results show that the proposed algorithm is effective.