Economically optimal flooding probabilities in the IJssel Lake area : A review of the Diqe-Opt model and optimization solvers

This research aims to find a suitable optimization solver to replace CPLEX for optimizing the 1-barrier Diqe-Opt model in the IJssel Lake area, which involved integer programming. This model is used in policy making to determine the water safety norms. Suitability of alternative solvers is defined based on four criteria: being free, open source, compatible with Python, and capable of handling integer programming problems with exact algorithms. Additionally, alternative solvers are required to match CPLEX’s performance in terms of producing the optimal solution and computing time.

The methodology involves a thorough description of the existing model to understand its workings, followed by an exploration of potential alternative solvers. In the second part, the mathematical background of the optimization algorithms is investigated. Two potential alternative solvers (GLPK and CBC) are implemented in combination with two potential modeling packages (Pyomo and PuLP) to assess their viability as replacements.

The performance of the alternatives is tested against CPLEX, and the results indicate that the CBC solver with PuLP as a modeling package could match CPLEX’s performance in most cases, with only minor difficulties in handling extremely small numbers. CBC reports in some cases small deviations from the optimal solutions. However, these do not have a potential significant effect on policy making if we limit the discount rate or the time horizon. While CBC has slower computing time than CPLEX, it remains fast enough to be considered a suitable replacement. Consequently, CBC in combination with PuLP is recommended as an alternative for CPLEX, when keeping the limitations on the discount rate and time horizon in mind.