| Offshore marine disasters have become an essential factor threatening the rapid development of China’s coastal areas.With characteristics of sudden,frequent,and destructive,disasters have caused significant direct economic losses and the number of deaths and disappearances in the past decade.Offshore marine disasters are induced by abnormal marine meteorology,biology,and geology activities.Timely and accurate disaster judgment and emergency plans are the keys to disaster management.With the rapid development of observation measurement,it is effective to reproduce the hazard through modeling,assessing the potential threat of disasters,and customizing the management strategy to achieve effective disaster prevention.The research objective of this thesis is to integrate the disaster evolution and disaster-causing process into a multiscale and multi-temporal "glass Earth".It can support valid disaster response based on the simulation,evaluation,and decision-making of offshore marine disaster management.The digital twin is a framework composed of the physical entity,virtual twin,and digital mapping relationship.The digital twin is suitable for disaster research due to multi-physical field coupling,multi-scale data fusion,and high-fidelity model characteristics.Digital twinning for offshore marine disasters has the following critical problems:(1)Marine disaster evolution is complex.Hazard monitoring data is challenging to organize,and the multiple numerical models are mutually coupled.How to reproduce the offshore marine hazard based on the numerical simulation method and intelligence method is a problem;(2)The topography of the coastal waters is complex.The features of coastal cities are diverse.How to design intelligent assessment methods according to the assessment needs of different scales to meet the accurate and efficient disaster prevention and early warning is a critical problem;(3)Sudden and severe marine disasters with long duration and vast influence.A critical problem is designing a robust hazard monitor system and effectively scheduling the emergency materials,providing efficient and reasonable decision support for pre-disaster prevention facilities deployment,disaster relief items relocation in advance,and real-time rescue in disaster.This thesis focuses on two typical disasters,the offshore oil spill disaster,and the storm surge inundation disaster.The main research contents and innovations are as follows:(1)Design of digital twin framework for offshore marine disasters.The characteristics of offshore marine disasters include many disaster-causing factors,complex evolution processes,intense regional outbursts,and extensive disasters.For disaster prevention and emergency preparedness,monitoring and early warning,emergency response,and rescue management in disaster prevention and control,the digital twin framework integrates the whole procedure of disaster prevention from monitoring simulation,analysis,and evaluation to decision support.According to the physical process of offshore marine disaster,disaster-related data are sorted out,the functions and results of assessment and decision support are sorted out and analyzed according to the management needs of specific disasters,and the conceptual framework and technical framework of digital twin are proposed.After the design and implementation of each sub-function module in the digital twin framework for offshore marine disasters is completed,the intensely interactive and high-fidelity offshore marine disaster twin supports disaster management and control.(2)Offshore oil spill simulation,assessment,and prevention.For the oil spill accident occurs around a drilling platform in a small region,solve the imprecise,inconsistent historical data,the distribution of the discrete,isolated problem through the Monte-Carlo sampling method and the numerical model of oil spill disaster evolution,fitting and twin simulation with high time complexity.The oil spill lookup table and the deep Q-transfer-learning network method were designed to locate the offshore oil spill source.For the oil spill oriented by the submarine pipeline rupture caused in the large region,a finite volume coastal ocean model is utilized to reproduce the physical evolution of the oil spill.The cross-entropy is proposed to tackle oil spill detection.The massively parallel computing model based on high-performance computing technology is further studied to improve efficiency.In order to enhance the disaster prevention and mitigation response of marine disasters,an oil spill monitoring system based on offshore sensors combined with a cloud server is studied.The target sea area is continuously monitored by laying buoy sensors to provide data support for disaster analysis.In order to improve the robustness of the whole monitoring system,the method of partial observation Markov decision process is utilized to evaluate the stability of the sensor network.(3)Storm surge inundation simulation,assessment,and decision-making support.Storm surge disasters arouse the sea level rise,heavy rain weather,and urban surface runoff.By coupling the marine current and urban flood model,the Markov decision process method simulates the underground drainage ability and constructs a model of twin storm surge inundation.Furthermore,a storm surge simulation coupled with an ocean wave and wind field model is designed to assess storm surge disasters.Based on the existing assessment system,a differential vulnerability assessment system is constructed by utilizing the point of interest data.Then,based on the urban inundation,the disaster risk and risk are designed and assessed to provide decision support for disaster prevention and mitigation.Based on the analysis and assessment of the impact of simulated storm surge on coastal cities,a disaster relief model for disaster prevention materials is proposed.The material scheduling scheme is planned to provide decision support for storm surge disasters through optimization based on a mixed integer linear programming method and a deep deterministic policy gradient method.This thesis proposes a digital twin framework for offshore marine disasters to meet the requirement of the whole-chain disaster management and control of analog,assessment,and decision support.Based on marine geological,hydrological,and meteorological data,urban geographic information,and disaster data,it integrates wave,wind field,diffusion,and surface runoff models.It combines intelligent methods,including lookup tables,cross-entropy,linear programming,and reinforcement learning to drive dynamic data.This ensures the high performance and high-fidelity disaster reconstruction,analysis,and decision of the twin offshore oil spill and storm surge disasters. |
PDF Full Text Request