Font Size: a A A

Research On Determination Of Secondary Porosity Of Volcanic Reservoirs

Posted on:2011-12-06Degree:MasterType:Thesis
Country:ChinaCandidate:X H WangFull Text:PDF
GTID:2120360305955291Subject:Earth Exploration and Information Technology
Abstract/Summary:PDF Full Text Request
By the ongoing found of volcanic reservoirs, the volcanic rocks can not be ignored in the oil and gas exploration. The porosity prediction of volcanic reservoirs becomes an important concern and issue attracting lots of domestic and foreign oil workers. Since the secondary porosity of volcanic reservoirs is the major oil and gas reservoir spaces all the time, for the petroleum exploration and development, it is very important to predict secondary porosity of volcanic reservoirs accurately.The evaluation of reservoirs is very important since the complex porosity spaces of volcanic reservoirs. The conventional log interpretation of secondary porosity of volcanic reservoirs has not yet formed a mature and effective method. Imaging log methods are restricted by areas and economic conditions, which make part of the region can not explain the volcanic secondary porosity using imaging log data. For the past few years, it is more and more common in the evaluation of volcanic secondary porosity using neural networks, and the methods achieve good results. The study area according to this paper locates in the southern of Song-Liao basin, where regional fractures grow and has imaging log data. The main aim of this study is: predict the secondary porosity of volcanic reservoirs based on the conventional log data of study area, according to imaging log data and using BP neural networks technology.Firstly, in this paper, compared with the previous research methods, the paper describe the region general situation of volcanic reservoirs on the study area, including the description of fractures types in volcanic reservoirs and the logging response characteristics of volcanic fractures. The paper divided the volcanic lithology of the study area using the ECS method with actual log data, and counted the skeleton parameter values of different lithologic characters, laying a foundation to the using of conventional log method to determine secondary porosity. In this paper, the methods of using conventional log data and imaging log data were summarized respectively to calculate the secondary porosity of volcanic reservoirs. However, there are large errors when these methods used alone, thus, we use BP neural networks technology to determine the secondary porosity of volcanic reservoirs in this paper. The paper briefly describes the basic structure of BP neural networks and learning algorithms. For the defects of BP learning algorithm, a genetic algorithm (GA) is informed. Genetic algorithm is proposed to optimize the initial weights of BP networks. Finally, programming procedures for the GA-BP neural networks model and process well log data of XX on the study area which is so much better. The above proves the feasibility of using GA-BP neural networks technology to predict secondary porosity of volcanic reservoirs.The technical superiority of this paper is reflected in the part using genetic algorithm to optimize BP neural network initial weights and thresholds. We mainly use the global search ability of genetic algorithms and neural networks convergence speed and learning ability characteristics. In this paper, we program in the Visual C++ environment and achieve the optimization of the BP networks initial weights and thresholds. After training we get the mean square error of change maps of optimized training model. We compare the change of the mean square error and predicted result of the optimized BP network model and the traditional BP network model and find that the BP network model optimized by genetic algorithm is superior to traditional BP neural network model. It is much faster in learning and has little forecast errors. We compare the predicted results from GA-BP networks model and the secondary porosity calculated with the conventional log method. It is showed that the results from GA-BP networks prediction model are much closer to the secondary porosity obtained from imaging log data. The conventional logging method has many parameters and it is affected much more by human factors and it has more error.The advantage in predicting secondary porosity using GA-BP neural networks model is manifested below:1. It has much better objectivity predicted results and less effect on human.2. The calculation method of the model is simple and fast. For the trained model, we only need to enter the log data and then quickly obtain the predicted results and it greatly simplifies the calculation.
Keywords/Search Tags:volcanic rocks, secondary porosity, imaging well log, genetic algorithm, BP neural networks
PDF Full Text Request
Related items