Deep Neural Network Enhanced Early Warning System for Ports Operations

Abstract

Downtime of port terminals results in large economic losses and has a major impact on the overall competitiveness of ports. Early Warning Systems (EWS) are an effective tool to reduce ports’ vulnerability by increasing their preparedness and planning capacity to either avoid or efficiently respond to emergency situations. The SAFEPORT EWS predicts met-ocean variables and their impact on ships (navigating, docking or moored) and port operations daily, Pinheiro et al. (2020), and provides alerts for emergencies and operational constraints. It is currently implemented in six ports, namely: Praia da Vitória, São Roque do Pico and Madalena do Pico, in the Azores archipelago, and Sines, Aveiro and Figueira da Foz, on the west coast of Portugal. All information provided by this EWS is available in a dedicated website and mobile application. In addition, an alert bulletin is sent by e-mail to interested parties. This provides port stakeholders with a decision support tool for timely implementation of mitigation measures to prevent accidents and economic losses. This system is now being enhanced with artificial neural network tools to obtain more accurate results from numerical wave propagation models and to allow the use of complex and time-consuming numerical models in an operational framework. As with any EWS, its usefulness depends heavily on its reliability, accuracy, and consistency. Two types of ANN have therefore been developed.