Decision tree-based optimization for flexibility management for sustainable energy microgrids

Abstract In this paper, we apply a flexibility based operational planning paradigm to microgrid energy dispatch. The classic energy dispatch problem with energy storage and dispatchable thermal generation assets requires the solution of mixed-integer optimization problems. Such approaches are not amenable to most remote microgrids and practical field microgrid implementations, where controls are rule-based and typically implemented by programmable logic controllers. Albeit such rule-based dispatch controls are always feasible, they cannot optimize fully over the availability of renewable generation and asset capacities of microgrids, especially energy storage. In this paper we propose a systematic method to generate the microgrid dispatch rule base with the objective of matching as much as possible the control performance obtained by full mixed-integer optimization. To achieve this we develop a rigorous control mapping method based on decision trees. The numerical results demonstrate that the decision tree-based dispatch strategy can provide feasible and near optimal dispatch decisions for microgrids. Its computational efficiency is very high, a feature promising for real-time in-field implementation.