A comparative study of decentralized data reconciliation algorithms

A comparative study of data reconciliation methods based on decentralized calculus is presented. An efficient strategy for reconciling data requires a dimensionality reduction of the initial problem by observability and redundancy concepts. The decentralized calculus is applied to the redundant subsystem obtained by this decomposition. The result obtained from applying different algorithms to a numerical example shows the efficiency of using an analytical algorithm with coordination by coupling variables. If there is a incidence matrix of a subsystem which is not a full row rank matrix, the Gauss-Seidel algorithm can be used. A relaxation scheme guarantees the convergence of this algorithm.<<ETX>>