Optimal inventory control for bottleneck isolation in general processes

Optimal inventory control seeks to isolate the economic effect of bottlenecks and maximise the throughput of processes. This is challenging in complex topologies with disturbances causing shifting bottlenecks. Decentralised and model predictive control schemes have been proposed for bottleneck isolation of sequential processes. Although decentralised control schemes work well for sequential processes, they are difficult to apply to more complex topologies such as parallel arrangement of units, flow splits, mergers, and recycles that are common in the industry. In contrast, such multi-input multi-output systems can be naturally handled with model predictive control schemes. This work extends a preliminary model predictive control scheme in the literature to achieve bottleneck isolation in general process topologies. In particular, a seriatim amongst inventories and system outflows is created using weights in the objective function. Our approach is simple to implement and is shown to optimally isolate bottlenecks on a wide range of case studies and topologies.