Mutiple Bad Data Identwication for State Estimation by Combinatorial Oftimization

The problem of multiple bad data in state estimation is thoroughly analyzed and a new approach to bad data identification is proposed. The method supersedes the largest normalized residual method as a special case for single or multiple noninteracting bad data. The approach borrows the framework from Decision Theory. The bad data identification is formulated as a combinatorial optimization problem. The optimization takes into account the reliability of the measurements. An efficient branch-and-bound algorithm fully exploiting the knowledge of the problem is developed. The method is reliable, efficient, and does not require separate testing of network observability.