The Method Of Oilfield Development Index Prediction And Condition Assessment Based On The Reservoir Modeling Results

With the depth of oilfield exploratory development, the major blocks of many oilfields in our country have gradually entered high water-cut stage, the subsurface reservoir distribution has become more complicated and the moisture content has gradually moved upward, even the recoverable reserves and oil production are declining. Therefore, to improve the accurate understanding about oil development production regularities and exploitation conditions of the present stage, to timely research the change regularities of oilfield performance and to adjust corresponding oilfield development program is very significant. The oilfield development index prediction and condition assessment is important in the oilfield performance research, whose accuracy degree can directly affect the program arrangement formulation and implementation adjustment in the oilfield development. Yet choosing what kind of models and algorithms has a direct impact on the results of prediction and assessment.Based on the results of reservoir modeling and numerical simulation, the intelligent models and algorithms are established for oilfield development index prediction and condition assessment of meeting actual oilfield performance regularities, which can provide a research method for oilfield development index prediction and condition assessment and have a certain practical meaning.This thesis selects the intelligent analysis model and algorithm to research the oilfield development index prediction and condition assessment method by analyzing the regularity of the oilfield development index data and the limitations of traditional prediction methods. The traditional prediction methods are always based on blocks or well, while the thesis’ s research is mainly refining to layer on the basis of reservoir geologic modeling and numerical simulation. The practical problems can be more fine and targeted by researching layer index data. For the layer development index prediction, the development index prediction model is established by using the discrete process neural networks model on the basic of the Walsh transform. In order to improve the adaptability and the predicted ability of the model, the predictive model parameters are optimized according to the improved particle swarm optimization algorithm.As to the abnormality diagnoses of oil well in oilfield development,the process support vector machines classification model is proposed which is based on the improved Gaussian kernel function. The model takes the oilfield actual development data and prediction data to build the training sample set. And the genetic algorithm simulated annealing is used to optimize the model kernel parameter, in order to improve its global optimizationability and learning efficiency. The diagnostic model has achieved good results in the actual application material handling.On the basis of oilfield development index prediction and condition assessment model and algorithm, in the thesis the software system for oilfield actual material handling is designed and developed and achieves good experimental analysis results. The research of the thesis enriches the method of index prediction and condition assessment to some extent and has certain theoretical significance and reference value in guiding the oilfield development plans designing and adjustment.