| Natural gas hydrate is a non-stoichiometric clathrate crystal compound generated under high pressure and low temperature conditions.Natural gas hydrate is an unconventional energy source with huge reserves,and its mainly occurs in terrestrial permafrost and marine sediments,its appearance is soft and resembles ice and snow,so it is also called "flammable ice".After the methane released from the decomposition of natural gas hydrate is burned,only carbon dioxide and water are produced,and there is almost no environmental pollution.Therefore,natural gas hydrate is regarded as an ideal clean energy in the future.Exploring the theory of safe,efficient and low-carbon development of natural gas hydrate reservoirs is an significance strategy for realizing2035 deep-sea natural gas hydrate exploitation plan of China,achieving the strategic goals of "carbon peaking" and "carbon neutrality",and ensuring national energy security.The commercial exploitation of natural gas hydrate requires scientific exploitation strategies to achieve the purpose of reducing exploitation cost,improving production efficiency and ensuring operation safety.Based on deep learning and reinforcement learning theory,this research establishes a self-learning multi-neural network dynamic intelligent decision-making system suitable for gas hydrate reservoir exploitation,aiming to realize safe,efficient and low-carbon gas hydrate reservoir development and promote exploitation efficiency.The synergistic improvement of performance and safety will provide theoretical and technical support for the commercial exploitation of hydrate reservoirs.The main research contents and conclusions are as follows:(1)The numerical simulation model of natural gas hydrate reservoir exploitation is established on the basis of considering the multi-physics coupling of phase state changes,heat transfer characteristics and gas-liquid seepage involved in the process of hydrate reservoir exploitation.Based on the depressurization-thermal stimulation decomposition law of methane hydrate at the laboratory scale,the numerical simulation and experimental results under the same conditions were compared and analyzed,and the reliability of the numerical simulation model was verified.(2)Based on the natural gas hydrate exploitation numerical simulation model,an intelligent exploitation model of gas hydrate reservoir was established,which solves the problems of high-dimensional continuity of production state,continuous correlation of policy parameters,lagging effect of recovery policy,and long-term value of policies that is difficult to determine and are easily masked by follow-up policies.The multi-neural network intelligent exploitation model realized autonomous interactive learning,intelligently set numerical simulation parameters,automatically explore exploitation policies.At the same time,it realized the dynamic evaluation and visualization of the long-term "carbon-energy" and safety coordination comprehensive value of exploitation policies based on the perspective of big data.(3)Based on three exploitation scenarios,the exploitation policy optimization of gas hydrate depression-thermal stimulation is studied in laboratory scale.The results show that the proposed intelligent exploitation model achieves the coordination between the hydrate recovery objective(carbon)and the economic objective(energy)well,and achieves the expected effect of the model.Among them,in the dynamic optimization scenario of energy production-injection ratio of natural gas hydrate reservoir exploitation,the energy production-injection ratio is increased by 4.5 times;in the dynamic optimization scenario of CH4 recovery efficiency,the CH4 recovery efficiency is increased by 1.6 times;In the dynamic optimization scenario of the exploitation policy under the preset recovery plan,the model achieves the preset plan well under the condition of maximizing the reduction of thermal energy injection,and the control error is only about 1%.(4)The intelligent exploitation model learns and quantifies the comprehensive "carbon-energy" coordinated value of each production policy for hydrate recovery at the laboratory scale.Through the model learning,the quantitative relationship between each thermal stimulation policy and its developmentally changing energy utilization efficiency was established.The results showed that in the initial hydrate recovery period,thermal stimulation policies with a high injection temperature and low injection rate achieved greater energy utilization efficiency.As recovery progressed,gradually lowering the injection temperature and increasing the injection rate gave the policy a higher energy utilization efficiency.Moreover,the learning results of the AI model also showed that the reactor scale hydrate recovery had a size effect,such that the later the thermal energy injection was,the higher the energy efficiency that could be obtained.(5)Taking the exploitation of natural gas hydrate reservoirs in the Shenhu area of the South China Sea as the research background.On the basis of the intelligent exploitation model,it is coupled into the long short-term memory neural network model.It realizes the storage and evaluation of the influence relationship between each successive production state and the long-term effect of various exploitation policies in the process of dynamic reservoir production,and solves the continuous high-dimensional problem of field-scale hydrate reservoir production optimization.Under the preconditions of meeting the gas production efficiency demand,ensuring the wellbore pressure safety,minimizing the water production amount,and maximizing thermal energy utilization efficiency,this research obtains the optimal dynamic production policy and quantitatively presents the advantages and disadvantages of different exploitation policies in each reservoir exploitation state.Meanwhile,the model-based learning and evaluation completed the quantification and visualization of the comprehensive value for different reservoir development states and different exploitation policies.(6)Based on the evaluation results of the intelligent exploitation model,from the perspective of heat and mass transfer,the reasons for the dynamic changes of the policy value at the filed scale were systematically explored,and the methods for improving the exploitation efficiency were explored.The results show that: in the early stage of production,on the premise of ensuring the safety of wellbore pressure,the high temperature policy with increasing heat injection rate can provide faster heat and mass transfer efficiency,making the recovery of hydrate more efficient,economical and safe;The huff and puff heat injection method can effectively reduce the permeability decline caused by the formation of secondary hydrate,and enhance the safety of the wellbore during the heat injection process;the higher solubility of methane is easy to cause dissolved methane to precipitate and accumulate around the production well,increasing the blockage risk of production wells.In the middle stage of production,the production efficiency of hydrate is controlled by the density of energy injection.More intensive heat injection per unit time will bring higher production efficiency and lower water production.In the late stage of production,the rate of methane recovery is not depending on the injection of thermal energy,more thermal energy supply can only lead to lower exploitation returns;meanwhile,the injection of heat water will cause the increase of reservoir pressure,but it is difficult to cause the temperature change at the decomposition front,which will lead to a decrease in production efficiency. |
PDF Full Text Request