Highly Efficient Implementation for Parameter Error Identification Method Exploiting Sparsity

Accuracy of the network parameters has a strong influence on the results of power system state estimation. It has been shown earlier that normalized Lagrange multipliers can be used as a systematic way for identifying errors in network parameters. However, this approach carries a rather heavy computational burden limiting its practical utilization to small-size systems. In this paper, a computationally efficient algorithm is proposed to address this limitation. The idea is to derive and compute only the necessary subset of the gain matrix and covariance matrix, thus avoiding the computation and storage of large dense matrices. The proposed efficient procedure can be applied either to the single-scan or multiple-scan schemes with equal ease. Test results confirm that the improvements in computational speed and memory requirements brought by the proposed algorithm are quite remarkable. The proposed implementation of the normalized Lagrange multipliers method is tested using a large utility power system. The effectiveness and limitations of the single-scan scheme, and the improvements brought by incorporating multiple measurement scans, are discussed in detail.