Applied Research Of Support Vector Machine Based On Ant Colony Algorithm In Reservoir Parameters

An oil field filled the oilfield with water for a long time, oil reserves were atdepressed levels. By the late mining, The long-term water makes water ratio of oilincreases, the distribution of remaining oil underground complex, increasing thedifficulty of oil exploitation.So how to calculate accurately reservoir parameters(porosity, permeability, oil saturation)at this stage, to re-learn some rules about,lithology, electrical, physical properties, and accurately calculate the distribution ofthe remaining oil is in urgent need to solve problems.At present, the thought of reservoir parameters calculating is that. First, we usedthe curve data of inspection wells to build traditional modeling for porosity,permeability and oil saturation. Then, we predicted some reservoir parameters of theunknown well though the traditional model. The traditional modeling is linearmodeling. That is, through a curve of the deep lateral resistivity (RLLD), spontaneouspotential (SP), gamma ray (GR) and slowness (AC) to fit the porosity, permeabilityand oil saturation. And then using the established model to predict the porosity,permeability and oil saturation of the unknown Wells. But there is a big differencebetween the value of the reservoir parameters calculated by the traditional linearmodel and the actual value. In order to find more high-precision models to predictreservoir parameters. We use the support vector machine to build the nonlinearmodeling about parameters. Support vector machine is a machine learning approach.Is a kind of machine learning also. Its main idea is statistical learning theory. Thenumber of training samples in traditional statistics is very big, closely to the infinite.However, compared to some traditional statistics, the size of training sample in thestatistical learning theory is often limited, such as the log data in this paper. Therefore, in this article using log data which number is often limited to model, we will get abetter effect by using the support vector machine (SVM).Support vector machine is mostly applied in classification. In this paper weapplied the support vector machine regression. In the process of support vectormachine regression modeling, we need to set two parameters, the penalty factor C andinsensitive coefficient parameters.In this paper, we optimize these two parametersthrough ant colony algorithm. Ant colony algorithm is a behavior of simulating antsforaging. Ants always find the shortest path between the nest and the food sourcethrough information. We use ant algorithms to optimize parameters of support vectormachine by grid method. Then support vector machine parameters that ant colonyalgorithm optimize build porosity, permeability and oil saturation modeling, andtested the accuracy of the model through method of cross plot and histogram. Finally,we used the model building to predict reservoir parameters of the actual wells, andcompared with oil reservoir parameters through traditional modeling predicted. Toillustrate the higher accuracy of the model that support vector machine built throughant colony algorithm optimize parameters. The prediction of reservoir parameters isthe more accurate. As well as it better reflect the distribution of remaining oilunderground.