An application of dynamic programming to assign pressing tanks at wineries

This paper describes an application of dynamic programming to determine the optimal strategy for assigning grapes to pressing tanks in one of the largest Portuguese wineries. To date, linear programming has been employed to generate proposed solutions to analogous problems, but this approach lacks robustness and may, in fact, result in severe losses in cases of sudden changes, which frequently occur in weather-dependent wine factories. Hence, we endowed our model with stochasticity, thereby rendering it less vulnerable to such changes. Our analysis, which is based on real-world data, demonstrates that the proposed algorithm is highly efficient and, after calibration, can be used to support winery’s decision-making. The solution proposed herein could also be applied in numerous other contexts where production processes rely on outside supplies.