Porosity prediction method based on multi-layer long short-term memory neural network model

Abstract

The invention belongs to the technical field of reservoir parameter prediction, and particularly relates to a porosity prediction method based on a multilayer long-term and short-term memory neural network model. A neural network model based on multi-layer long short-term memory comprises an input layer configured to represent original logging parameters of porosity; the hidden layer is formed bysuperposing a plurality of long short-term memory (LSTM) models; and the output layer is used for outputting a predicted value of the output porosity through the full connection layer from the last hidden layer. According to the invention, the LSTM is superposed; a plurality of LSTM models are used for framework prediction; the output of the front layer is used as the input of the LSTM model of the rear layer; by using the deep LSTM model, long-term information can be memorized continuously, time information can be screened more strictly, prediction can be performed well at the inflection point of porosity change, and particularly, a high-precision target parameter prediction value can be obtained under the conditions of few well positions and few well logging parameter dimensions.

Description

technical field [0001] The invention belongs to the technical field of reservoir parameter prediction, and in particular relates to a porosity prediction method based on a multi-layer long-short-term memory neural network model. Background technique [0002] Deep learning is derived from the method of data representation learning in machine learning, and it is also one of the most modern and practical machine learning methods. High-level abstract attribute types or features are formed by recombining from lower-level features. Breakthrough achievements have been made in many scientific fields. The prediction accuracy and recognition ability based on deep learning continue to improve, and more and more applications are applied to practical development problems in various fields. In the past, deep learning has been drawing on the working principle of the brain, and applying the theoretical knowledge of statistics and applied mathematics. A large number of experimental studie...

AI Technical Summary

Problems solved by technology

The prediction of reservoir parameters is only related to the logging data at the corresponding depth, thus ignoring the impact of the logging data before and after the depth, and it is difficult to guarantee the authenticity of the predicted results

Artificial neural networks and BP neural networks tend to fall into local minimums when dealing with very large data and the accuracy is not high, making it difficult to effectively predict sequence data

There are also many researchers who make predictions by improving the artificial neural network and BP neural network, but the implementation process is very complicated

Images