Auto Tuning of Measurement Weights in WLS State Estimation

This paper describes an approach for choosing and updating measurement weights used in weighted least squares (WLS) state estimation. Since the weights are related to the measurement error variances, sample variances are estimated using historical data from previous measurement scans and the corresponding WLS estimation results. The proposed approach can be implemented as a one-time estimation function for off-line execution or as a recursive function for updating the measurement weights on-line. Simulated measurement data and state estimation results are used to test and verify the accuracy of the proposed method. The proposed method can be integrated into an existing WLS state estimator as an added feature.