Research On The Seismic Attribute Optimization And Integration Of Yan Chuanan Coalbed Methane Exploration

Abstract:As our country is poor in oil but rich in coal, our coalbed methane reserves is quite rich and can alleviate the energy shortage crisis. Therefore, the study of coalbed methane seismic exploration technology becomes an urgent task. The East China Branch of China Petroleum&Chemical Group decided to process large-scale prospecting and exploitation of coalbed methane in Ordos Basin.In seismic exploration of coalbed methane, the stronger amplitude seismic reflection of the coal seam is easier to identify because of its low-density, low-speed, low wave impedance characteristics, but the coalbed methane is difficult to identify from the background of strong reflection. Therefore, the paper involves the series studies of seismic attributes, and provides some system approaches and technologies of seismic attribute extracting, optimizing, and using the neural network method and its improved method to predict the reservoir parameters.The beginning of the paper introduces some basic geological theories, including the genesis of formation, the existing condition, the movement mechanism and the enrichment condition of coalbed methane. Studies show that the main factors affecting coalbed methane production includ the metamorphic grade of coal and the change of coal seam thickness. The higher metamorphic degree of coal is and the greater the thickness of coal seam is, the more coalbed methane form. If the coal seam thickness can be effectively predicted, coalbed methane prediction become very meaningful. Then the paper research on using the seismic attributes to predict the reservoir parameters of the coalbed methane. Firstly, seismic attributes are more profoundly understood through studying signification, classification, extraction methods and influence factors of seismic attributes. In addition, the paper Summarizes the selection principle of seismic attribute extraction window. Secondly, use the correlation analysis and the principal component analysis to select the seismic attributes which has good correlation with the coal seam thickness but is independent from each other. Then use the K-L transform to reduce the dimensionality of the training samples. After these three steps, the training samples remove the redundant data and realize the optimization of composite sample. Lastly, use the optimized samples to train the BP neural network and its improved method. Six well data is validation samples to detection the prediction effect and use the three neural network model to predict the coal seam thickness in the work area. After comparison, the BP neural network optimized by genetic algorithm has the highest prediction precision and the most ideal prediction effect.