摘要
在传统CARMA模型的基础上,用RLS法拟合高阶CAR模型单独得到白噪声估值,并将之应用于RML算法中。仿真实验结果表明,改进后的RML算法极大提高了算法的精度,具有较强的适应性、稳定性和灵活性,收敛速度快,有效解决了滑动平均模型参数估值收敛较慢的问题,在预报和控制领域中具有一定的实用价值。
In this paper, modified recursive maximum likehood(RML) parameter estimation algorithm is based on the traditional CARMA model, and the RLS method is used to fit the high-order CAR model separately to get the white noise valuation, which is then applied to the RML algorithm. Simulation results show that the improved RML algorithm greatly improves the accuracy of the algorithm, strong adaptability, stability and flexibility, converges quickly, and the problem of slow convergence of sliding average model parameter estimation is solved effectively. It has certain practical value in the field of forecasting and control.
引文
[1] Box.G.E.P,Jenkins.G.M.Time series analysis, Forcasting and Control[M]. San Francisco:Hollen-Day, 1979:152-153.
[2]王秀峰.系统建模与辨识[M].北京:电子工业出版社,2004:96-97.
[3]锂离子电池建模及其参数辨识方法研究[J].中国电机工程学报,2016,36(22):6254-6261.
[4]蓝康孟,洪文学.一种具有特殊滤波器的递推极大似然法[J].计量学报,1995,16(4):290-296.
[5] Zheng Dezhong,He Qun.A New Fast Algorithm of Generalized Predictive Control with Filter[J].Chinese Journal ofElectronics,2006,15(2):242-245.