Eventos Anais de eventos
COBEM 2023
27th International Congress of Mechanical Engineering
OCEAN CURRENT ESTIMATION FOR A TURRET-MOORED FPSO USING NEURAL NETWORKS
Submission Author:
Pedro Felipe Lavra Dias , MA
Co-Authors:
Pedro Felipe Lavra Dias, Gustavo Bisinotto, Anna Helena Reali Costa, Eduardo Aoun Tannuri
Presenter: Pedro Felipe Lavra Dias
doi://10.26678/ABCM.COBEM2023.COB2023-0723
Abstract
This research deals with a data-driven model to estimate local ocean currents affecting a Turret-Moored FPSO. Strong environmental conditions impact the effectiveness of oil exploration at deep sea, reducing the uptime and increasing risks. With Turret–moored FPSOs, due to their hull-shaped format and the weathervaning property, crucial importance is given to onboard sensors capable of estimating these conditions and platform response. The accurate measurement of ocean currents is difficult with ordinary onboard devices. Hence, a motion-based Multi-Layer Perceptron Neural Network (MLP) is proposed to determine local current speed and direction. A dataset of environmental conditions observed in a Brazilian offshore oil basin was input to numerical simulations with two loading conditions: full-loaded and ballasted. Those simulations provided time-domain platform responses from which motion statistics were computed to generate a dataset that was associated with wind speed and direction to train four Neural Networks: two for each loading condition – one for current speed and the other for current direction. Results demonstrated a strong relationship between estimations and references (correlations around 70% and 90% for current speed and direction, respectively) and mean absolute errors of 0.09 m/s and 25° for full-load, and 0.09 m/s and 30° in ballast. Keywords: Turret-Moored FPSO, Neural Networks, Estimation of Ocean Current.
Keywords
FPSO, data-driven models, Ocean Current Estimation, neural networks
