Unbiased estimation in dynamic data reconciliation

Abstract A computationally fast technique accurately estimates process variables when conditions are dynamic due to changes in steady states. The process variable estimators are unbiased and have known distributions. Thus, confidence intervals for true values of process variables are provided. The formulation of this technique was motivated by a recursive, dynamic data reconciliation technique that obtains very accurate estimators. These two techniques are compared in terms of computational speed and accuracy of estimators. The proposed technique is computationally faster, but not as accurate when variances of process measurements are large. However, the accuracy of the proposed estimators is shown to approach that of the recursive technique by iteratively recalculating estimates and when measurement variances decrease.