Gated recurrent unit neural network (GRU) based on quantile regression (QR) predicts reservoir parameters through well logging data

The prediction of reservoir parameters is the most important part evaluation, and porosity very among many parameters. In order to accurately measure core, it necessary take cores for indoor experiments, which tedious difficult. To solve this problem, paper introduces machine learning models estimate through logging paper, gated recurrent unit neural network based on quantile regression method introduced predict porosity. Porosity measurement implemented by taking experiments. data divided into training set test set. are used as input model, measured in laboratory output Experimental results show that improves accuracy network, RMSE (Root Mean Square Error) unoptimized GRU 0.1774, after optimization, 0.1061. By comparing with widely BP proposed much higher than network. This shows excellent predicting