Predictive management of cogeneration-based energy supply networks using two-stage multi-objective optimization

A predictive management system for cogeneration unit-based energy supply networks using two-stage multi-objective optimization was developed to tackle a trade-off between energy savings and operating cost reduction. The developed system integrated support vector regression-based energy demand prediction, MILP (mixed-integer linear programming)-based schedule planning, and rule-based operation control. The contribution is to develop two-stage MILP-based multi-objective schedule planning, which is extension of an ε-constraint method, and operation control rule of multiple cogeneration units. In the first-stage schedule planning, primary energy consumption in the prediction horizon is minimized, and a reduction rate of primary energy consumption is calculated. In the second-stage schedule planning, an operating cost is minimized additionally subject to satisfaction of partial achievement of the reduction rate of primary energy consumption calculated in the first stage. An energy-saving achievement rate is regarded as a decision-making parameter to control a trade-off between energy savings and cost reduction, of which definition is quantitatively apprehensible for decision makers. Annual operating simulation of an energy supply network using four fuel-cell-based cogeneration units revealed that the developed predictive management system has high controllability to the trade-off between the energy-saving rates (18.9%–21.6%) and the operating cost reduction rate (19.0%–15.6%), caused by a time-of-use power tariff structure.