Two-stage optimal control of large-scale industrial pressure network

This paper presents a multi-stage framework for energy-optimal industrial compressor control in highly volatile demand conditions and analyzes the impact of forecast quality on closed loop performance. For compressor activation scheduling, a receding finite horizon dynamic programming formulation is used. The approach is coupled with an exhaustive load-sharing optimization and noise mitigation control. Within the scope of a case study for Europe’s largest manufacturing plant in northern Germany, the beneficial application of predictive activation scheduling compared to the commonly implemented reactive activation is showcased. Depending on the demand forecast quality, the approach achieved energy savings of 3.7 to 10 % compared to a purely reactive control strategy.