Model validation and uncertainty

When using an estimation algorithm for system modeling, the model uncertainty is manifested in the order and parameters, and the parameter uncertainty is a major problem for SISO systems. The authors discuss the conventional formula for computing the confidence region of the estimated parameter value, point out in a recursive algorithm the uncertainty of the covariance matrix P(L). Then an updated computation method of that is proposed in the paper. Because of the existence of parameter uncertainty, the evaluation of fit level between measured and simulated responses is also discussed. From a real modeling example of a hydroturbine generating unit, the authors reveal the reason why the influence caused by uncertainty in system behavior of lower frequency band is more serious.