Development of a Vessel-Performance Forecasting System: Methodological Framework and Case Study

摘要

Terminals rely on optimization tools to determine merchandise location, quay occupation, or vehicle trajectories to minimize the movements and time dedicated to every task. However, operations are developed in an environment that induces variability to the theoretical model used to schedule and control the operations. Given the complexity of the port operations, artificial intelligence systems can act as a valuable tool to analyze such processes. Neural networks in particular are characterized by their capacity to establish nonlinear relationships (and consequently, nonintuitive ones) among the variables; this interaction generates a specific operational response. In the near future, the monitoring of operational variables has great potential to make a qualitative improvement in the operations management and planning models of terminals that use increasing levels of automation. This paper proposes a method to obtain operational parameter forecasts in container terminals. To this end, a case study is presented, in which forecasts of vessel performance are obtained. By doing so, the management strategies are supported by an expert system, grounded in the historical data series of quay operation and the climatic conditions observed, as well as the ordinary and extraordinary events that have happened in the past, from which the system is able to learn. This research was based entirely on data gathered from a semiauto-mated container terminal from Spain.