Support Vector Machine And Applicaiton In Reservoir Capacity Prediciton

Traditional learning method of predicting Reservoir parameter are mostly based on the rule of Empirical Risk Minimization,the result is not satisfactory. Support vector machine is one kind of machine learning method that is based on statistic learning theory and structure risk minimum,it can reach the max generalization ability and global optimum by maximizing the upper bound of the promotion of error.Especially for Small sample of cases,it shows a good performance with a reliable predictive capability.At the same time, the problem of choosing params in Support Vector Machine is very important.because the params determine the model.the research about choosing the params of SVM is developed wildly.The chaos Particle swarm optimization have both the fast convergence and simple advantages, Meanwhile it can solve the precocious of the PSO algorithm.First of all the papers does some discussion on the principle of the support vector machine (SVM) Respectively, do some fitting work for the known function sinc using support vector machine method (SVM), analysis their predictive power,the Results show that support vector machine (SVM) has reliable prediction,Then, using this method to predict reservoir capacity.Taking the known five exploration wells in JiYuan as the training sample data,after normalizing the sample data,choosing the Effective porosity, reservoir permeability, water saturation as the input parameters,And then take chaos-particle swarm optimization algorithm to determine the parameters of support vector machine regression (SVR),Then we get the regression function, using the regression function to predict the capacity of one known well, The results have good agreement with known results,Verified the support vector machine (SVM) method is practicable in the predictions of reservoir prediction.