分布式参数系统的随机梯度辨识

Stochastic Gradient Algorithm for Distributed Parameter Systems

针对分布参数系统提出了随机梯度辨识算法。借助有限差分方法将分布参数系统简化为集中参数系统，通过随机梯度算法辨识出集中参数系统的参数。为提高参数收敛速度，提出带遗忘因子的随机梯度算法。最后，根据辨识出的集中参数系统的参数计算出分布参数系统的待辨识参数。仿真结果验证了该方法的有效性。

This paper proposes a stochastic gradient algorithm for distributed parameter systems. By using the finite difference method, the distributed parameter system is simplified as a lumped parameter system. Then a stochastic gradient algorithm is proposed to estimate the parameters of the lumped parameter system. In order to improve the convergence rate, a forgetting factor stochastic gradient algorithm is also derived. Finally, the parameters of the distributed parameter system can be computed by the estimated parameters. The simulation results indicate that the proposed method is effective.