An Outer-Approximation based algorithm for solving integer non-linear programming problems for Optimal Sensor Placement

There is a growing need across GE businesses to push system operation performance closer to design entitlement through application of advanced controls and optimization. The performance requirements coupled with stringent operability boundaries necessitate systematic design of sensing networks for condition monitoring across applications such as power plants, aviation, etc. This paper addresses the important system level question of an optimal sensor network design through model-based estimation and optimization. The main contribution of this work is the development of an outer approximation based algorithm for solving a pure integer non-linear programming problem that arises when designing an optimal sensor network for on line condition monitoring of systems like a gasifier in an Integrated Gasification Combined Cycle (IGCC) plant. The effectiveness of the proposed methodology has been demonstrated in designing a robust and cost-effective sensor network using a simplified gasifier model for an IGCC plant.