The use of support vector machine for oil and gas identification in low-porosity and low-permeability reservoirs (Q2655888)

Summary: Well-log interpretation becomes very complicated in low-porosity and low-permeability reservoirs, due to the strong non-linear relationship between oil/gas identification and well-log interpretation results. To find a method for predicting oil/gas identification, multiple regression analysis (MRA), backpropagation neural network (BPNN) and support vector machine (SVM) have been applied to two case studies based on well-log interpretation results and oil/gas test data. The specific MRA adopted is the technique of successive regression analysis, and the particular SVM employed is the technique of C-SVM binary classifier. The two case studies show that: 1. For the learning samples, the results of SVM and BPNN show a far more precise fit than MRA 2. For the prediction samples, the SVM predictions coincide with oil/gas test results and in fact correct some erroneous well-log interpretations, but the predictions of both MRA and BPNN do not.