A DIRECT SPECTRAL METHOD FOR CONTINUOUS-TIME STOCHASTIC SYSTEM ESTIMATION BASED ON MEASURED DATA AND A PRIORI INFORMATION
✍ Scribed by P. Nurprasetio; S.D. Fassois
- Publisher
- Elsevier Science
- Year
- 1997
- Tongue
- English
- Weight
- 462 KB
- Volume
- 11
- Category
- Article
- ISSN
- 0888-3270
No coin nor oath required. For personal study only.
✦ Synopsis
A novel prediction error method for the direct estimation of physically meaningful continuous-time stochastic systems based upon both measured data and a priori information is introduced. The method builds upon a block pulse function spectral formulation postulated by Nurprasetio and Fassois [20], which leads to an induced special-form stochastic difference equation in the discrete spectral domain. The direct estimation method is based upon the identifiability of the stochastic difference equation with respect to the original continuous-time parameters, which are now estimated directly through a prediction error type criterion and a variable metric optimisation scheme. The method eliminates all explicit transformations between the continuous and BPF spectral domains, while allowing for the construction of interval estimates of the continuous-time parameters and the straightforward incorporation of a priori system information. Its effectiveness is demonstrated via an application case study pertaining to the estimation of a stochastic electromechanical system, where the importance of a priori information incorporation is also highlighted.