A prediction model of specific productivity index using least square support vector machine method

In the design of oilfield development plans, specific productivity index plays a vital role. Especially for offshore oilfields, affected by development costs and time limits, there are shortcomings of shorter test time and fewer test sampling points. Therefore, it is very necessary to predict specific productivity index. In this study, a prediction model of the specific productivity index is established by combining the principle of least squares support vector machine (LS-SVM) with the calculation method of the specific productivity index. The model uses logging parameters, crude oil experimental parameters and the specific productivity index of a large number of test well samples as input and output items respectively, and finally predicts the specific productivity index of non-test wells. It reduces the errors caused by short training time, randomness of training results and insufficient learning. A large number of sample data from the Huanghekou Sag in Bohai Oilfield were used to verify the prediction model. Comparing the specific productivity index prediction results of LS-SVM and artificial neural networks (ANNs) with actual well data respectively, the LS-SVM model has a better fitting effect, with an error of only 3.2%, which is 12.1% lower than ANNs. This study can better reflect the impact of different factors on specific productivity index, and it has important guiding significance for the evaluation of offshore oilfield productivity.