A Novel Topology Determination Approach For Distributed State Estimation In Large Scale Plants

In this paper, a novel graph-based methodology is proposed to determine proper topology for distributed state estimation in large-scale industrial plants. For this purpose, the concept of energy flow has been utilized as a measure of observability to determine the closeness of plant variables, leading to avoidance of the worst observability possible situation. A weighted graphical representation of the plant is first prepared based upon the energy flow concept and a new proposed systematic algorithm is employed thereafter to partition the resulting main graph. The algorithm starts with assigning the plant output nodes as the initial target subgraphs centers. The remaining nodes are treated as free states which should properly be allotted to relevant subgraphs already been formed having plant outputs located at their centers. A merging routine has been proposed to efficiently reduce both the number of subgraphs and the amount of energy flows as much as possible. A computational constraint is also considered according to computational limitations on processor which restrict the size and number of subgraphs. The effectiveness of the proposed algorithm is demonstrated via an industrial chemical process.

Keywords:

topology , decentralized and distributed state estimation , graph theory , observability grammian