Research On Seismic Attribute Prediction Method Of Coal Seam Gas Content Based On ABC-BP Model

The prediction of coalbed methane content is one of the important research contents in the initial stage of exploration and development of coalbed methane resources.Using logging data to constrain seismic attribute inversion combined with linear mapping model is one of the common methods to predict coal seam gas content.However,the prediction accuracy of this method is difficult to control and its universality is greatly limited.BP neural network model is often used in the field of coal seam gas content prediction.However,in the face of complex scenarios,the traditional BP model is prone to slow convergence speed,the prediction results are greatly affected by the initial value of the network and fall into local optimization.Based on this,this paper puts forward an improved BP neural network prediction method which is characterized by artificial bee colony algorithm,and combined with the optimization of seismic attributes,applies it to the field of coal seam gas content prediction.Taking No.3 coal seam in a working area of Qinshui Basin as the research object,firstly,according to the specific geological conditions of the study area,we extracted pre-stack and post-stack seismic attributes slices of the target reservoir based on 3D seismic exploration data,and analyzed the relationship between seismic attributes and coal seam gas content.Secondly,R-type clustering analysis was used to classify various types of seismic attributes extracted from the target coal reservoir,and five types of seismic attributes which are most sensitive to the change of coal gas content and independent of each other has been selected.Then,the artificial bee colony optimization algorithm(ABC)is used to determine the optimal connection weight between the input layer and the hidden layer of BP neural network and the optimal threshold of the hidden layer,and a robust prediction model of ABC-BP neural network was constructed.The model was trained with the optimal seismic attributes of well location and gas content data as samples.Finally,the optimal seismic attributes of the target reservoir in the whole working area are used as input to predict the gas content of coal seam in the working area.To further verify the prediction accuracy and improvement effect of the ABC-BP model,the same data were brought into the traditional BP neural network prediction model for training in this study,and the prediction of coal seam gas content in the target reservoir in the work area was completed for comparative analysis.The analysis of related prediction results shows that compared with the traditional BP neural network prediction model,the improved ABC-BP prediction model has higher prediction accuracy,more stable error range and more ideal prediction effect.Among them,the prediction results of ABC-BP model are basically consistent with the variation trend of gas content of each well.The average error rate at the training well is 0.11%,which is 0.72%lower than that of the BP model,and the average error rate at the validation well is 2.22%,which is 1.57%lower than that of the BP model.In addition,the prediction results of ABC-BP model,whether high value,low value or medium value,have high prediction accuracy and stable error range,while the prediction results of BP neural network prediction model have poor prediction effect for high value and low value,and the error range is unstable.Therefore,the improved ABC-BP prediction model has high reliability and strong applicability,and can be effectively used in the prediction of coal seam gas content.Figure[33]table[6]reference[119].