Research On Trajectory Optimization Of Horizontal Well Based On Results Of Drilling,Well Logging,seismic And Mudlogging

The development of horizontal well drilling technology has greatly promoted the development process of special oil and gas reservoirs such as thin layer oil and gas reservoirs in my country.Horizontal wells have the characteristic of expanding the contact area between the underground oil layer and the wellbore,and have great advantages in increasing the production of a single well in an oil and gas field.Due to the complex and changeable underground factors,the accuracy of horizontal wellbore trajectory design is very critical.The optimal design of wellbore trajectory can not only reduce the difficulty of engineering construction but also further improve the benefit of reservoir development.The conventional optimization design method is to artificially design the wellbore trajectory given the coordinates of the geological target point,and continuously optimize and adjust it.However,due to the constraints of target location,geological factors and engineering factors,it is often difficult to quickly find an optimal wellbore trajectory that satisfies engineering geological factors and can accurately enter the target.To solve this problem,aiming at improving the drilling rate of horizontal wells,this paper studies a horizontal well trajectory optimization method based on deep reinforcement learning algorithm and a horizontal well trajectory parameter optimization method based on improved artificial bee colony algorithm.The specific research content is as follows.First,a multi-source information fusion model was constructed based on drilling,logging and seismic data.A multi-source information fusion model is constructed by combining drilling,well logging,mud logging,seismic multi-information source data and three-dimensional geological attribute volume data through data association and information fusion.The constructed multi-source information fusion model provides data support for building a deep reinforcement learning environment.Second,a deep reinforcement learning algorithm based on an improved reward mechanism is proposed.Based on the 3D geological attribute volume data in the multi-source information fusion model,the simulation experiment environment space,action space and reward and punishment function are designed,and the horizontal well trajectory optimization model is constructed by training the model and continuously adjusting the model parameters.By comparing various deep reinforcement learning algorithms,the deep reinforcement learning model with improved reward mechanism proposed in this paper has better convergence performance.Third,an event-triggered artificial bee colony algorithm is proposed for well trajectory parameter optimization.In the evolution process of honey bees,the influence of other bee colony individuals and the current optimal bee colony individual on its evolution is comprehensively considered,and an event trigger mechanism is added to make it adaptive to mode switching.Simulation experiments show that the artificial bee colony algorithm based on event trigger mechanism has better optimization performance.The improved artificial bee colony algorithm is used in the optimization of horizontal well trajectory with the measured length of well trajectory as the optimization goal.Fourth,based on the above research content,this paper builds a geological engineering integrated system platform to comprehensively manage drilling,logging,mud logging,and seismic multi-dimensional data,and realizes the analysis,storage,and data visualization modules of multi-information source data.And completed the well trajectory optimization module design.This paper makes a preliminary exploration on the direction of well trajectory optimization based on deep reinforcement learning.The research results show that it is feasible to use deep reinforcement learning to optimize well trajectory.This method has important guiding significance in guiding the trajectory control of horizontal wells.