Prediction of Production in Multiple Clusters Stages Fracturing Horizontal Well by Support Vector Machine

Conventional production prediction of multi-cluster stages fractured horizontal well is based on numerical simulation technology. While using this method, a large number of parameters needed, such as the reservoir parameters, fracturing treatment parameters, geological parameters etc. The huge computational time consuming of numerical method makes it too difficult for quick filed application. Against these deficiencies, the paper gives full consideration to the effects of reservoir, geology, and multi-cluster stages fractured parameters on productivity. A production prediction model of multi-cluster stages fractured horizontal wells is built by using SVM based on statistical theory and kernel function. First, its training algorithm is used to train the model. Then, samples are used to predict the production. Finally, production data is used to verify the model. Analysis the results show that the SVM model does not only have the advantage of quick prediction application, but also the prediction results obtained by the model have high consistency with the actual production data. It indicates that this method has good engineering practicability in production prediction of multi-cluster stages fractured horizontal wells.