| Sedimentary microfacies pattern recognition is the basis of studying stratal configuration,consturction and analysis container rock. It's also the important reference evidence for discovering new oil reservoir. Because most of the sedimentary microfacies pattern recognition was done by hand, it has some limitations such as ineffectiveness,anthropic factor,so it can't satisfy the Dagang oilfield's request. To solve this problem, we should find and implement a practicable method to automatic recognize the Sedimentary microfacies pattern with computer.First, we adopt an automatic recognizing sedimentary microfacies pattern method which combine the BP neural network with image process technical. When we choose the BP neural network model, the network studying may cause local minimization problem with incorrect initial value. So we use the method combining super liner BP algorithm with genetic algorithm. We use the genetic algorithm to choose initial value rough grade,then use super liner BP algorithm to ensure the expired grade. And instead of the method which use BP neural network to recognize sedimentary microfacies and extract curve character, we convert digital well logging curve to binary lattice image mode, encode and compress lattice data, then input them to the BP neural network, the BP neural network could extract and remember the stratum mode character with sample training.Second, we improve the ratio of recognition by optimizing the parameters of the neural network. Through the large amount of test with oilfield's hundreds of digital well logging curve, the result is improved to be correct and feasibility. With the trained BP neural network, We could simplify the traditional pretreatment process, increase operation rate of original data and recognizing accuracy.Finally, we describe the function and system structure, introduce the tools and database during the procedure of software development, and the database model and software design.
|
PDF Full Text Request