Intelligent Optimization of Overall Layout for Ultra-Deepwater Subsea Dynamic Umbilicals

Abstract

Abstract This paper addresses the growing global demand for energy, specifically focusing on the ultra-deepwater regions of the South China Sea, a rich source of oil and gas. In this context, the study of ultra-deepwater umbilical cable systems, a critical component in offshore oil and gas field development, becomes increasingly significant. The paper analyzes the composition, design challenges, and performance under extreme conditions of ultra-deepwater umbilical cable systems. Utilizing recent data collected from umbilical cable systems that are tied-back to six floating platforms in the South China Sea and dynamic analysis data from the OrcaFlex software, the study employs an enhanced Backpropagation (BP) neural network and Bayesian Optimization method for multi-objective iterative optimization under specific target functions, achieving an optimal overall design for the ultra-deepwater dynamic umbilical cables in the region. The results indicate that the predictive accuracy and generalization ability of the BP neural network model are significantly enhanced through the method combining the Bayesian Optimization algorithm with the improved BP neural network. The refined BP neural network achieves a mean squared error (MSE) of 0.011 and an R2 score of 0.98, demonstrating high predictive accuracy and generalization capability. This optimization reduces the BP neural network's training time by over 60%, meeting the precision requirements for umbilical cable design optimization. After more than 5000 iterations using the improved BP neural network combined with the Bayesian Optimization algorithm, the study identifies an optimal set of design parameters for umbilical cables that are tied-back to floating platforms in the South China Sea. This optimization reduces the maximum effective tension and overall axial tension in the umbilical cable system by over 15%, decreases cable curvature by more than 25%, and lessens the impact of vortex-induced vibrations by over 30%. Furthermore, the BP neural network-based optimization method employed in this study is successfully applied to the design of umbilical cables for floating platforms in the South China Sea, providing scientific and technical support for the design and optimization of ultra-deepwater umbilical cables. This research has significant implications for the exploitation and utilization of China's deepwater marine resources.