This paper considers algorithm selection for the berth allocation problem (BAP) under algorithmruntime limits. BAP consists in scheduling ships on berths subject to ship ready times and sizeconstraints, for a certain objective function. For the purposes of strategic port capacity planning,BAP must be solved many times in extensive simulations, needed to account for ship traffic andhandling times uncertainties, and alternative terminal designs. The algorithm selection problem(ASP) consists in selecting algorithms with the best performance for a considered application. Wepropose a new method of selecting a portfolio of algorithms that will solve the considered BAPinstances and return good solutions. The portfolio selection is based on the performance on thetraining instances. The performance is measured by the runtime and solution quality. In orderto select the portfolio, a linear program minimizing the solution quality loss, subject to overallruntime limit is used. Thus, the portfolio evolves with the runtime limit, which is a key parameter indesigning the port capacity simulations. For the training and validating datasets, random instancesand real ship traffic logs are used. A portfolio of heuristics is developed which can be used forsolving large instances of BAP, emerging when time horizons of months or years are considered.The evolution of the algorithm portfolios under changing runtime limits is studied. The portfolioabilities to solve new instances are assessed.