Solving Two-stage Stochastic Milp Chemical Batch Scheduling Problems by Evolutionary Algorithms and Ordinal Optimization

Chemical batch scheduling is mostly solved for problems where all data is assumed to be known. While this assumption makes scheduling problems much easier to handle, it cannot be upheld in reality. A possible way to introduce uncertainties into scheduling problems is to use two-stage stochastic mixedinteger linear programming where the uncertainties are represented by a discrete set of scenarios. With an increasing number of uncertainties, the complexity of these models increases rapidly and makes it impossible to solve them in a monolithic fashion in a reasonable amount of time. In this contribution we present a new approach to solve chemical batch scheduling problems by combining a hybrid evolutionary algorithm with a scenario decomposition technique from our previous work with the ideas of Ordinal Optimization. The proposed heuristic replaces the exact MILP solution of the scenario problems by fast non-exact solutions to perform a ranking (with small errors) of different promising first stage solutions.