An oil field development scheme optimization method based on a self-learning mechanism

Abstract

The invention provides an oil field development scheme optimization method based on a self-learning mechanism. The method is characterized in that a neural network oil reservoir simulation module, a Monte Carlo production parameter simulation module, an RNN result inference training module and a prediction module are included, and the method comprises the following steps that in the neural networkoil reservoir simulation module, nerves are trained through historical data, and a neural network oil reservoir simulator is obtained; In the Monte Carlo production parameter simulation module, various attributes in the production process are simulated through a Monte Carlo search tree method, different values are simulated for all attribute variables, and tens of thousands of mining schemes areobtained; an RNN result inference network is trained through the historical production data to obtain a mining scheme inference device; and the mining scheme generated by simulation is inputted into amining scheme inference device, and the optimal mining scheme is returned. According to the method, the neural network model is utilized to approach the unknown physical process, the constraint of oil reservoir numerical simulation simplified modeling formula is eliminated, and the accuracy is improved. Meanwhile, the training speed is greatly improved, and an optimal scheme can be selected froma large number of development schemes.

Description

technical field [0001] The present invention relates to deep learning and reinforcement learning, in particular to a method for optimizing an oilfield development plan based on a self-learning mechanism Background technique [0002] The selection of the traditional oilfield exploitation plan is based on the detailed exploration results and the necessary production test data, on the basis of comprehensive research, for the oilfields with industrial value, starting from the actual situation and production rules of the oilfield, with the aim of improving the ultimate recovery rate , to formulate a reasonable development plan. With the rapid development of deep learning, it has been widely applied and achieved good accuracy in the fields of recognition, classification and natural language processing; [0003] The technologies closest to the present invention in recent years are: [0004] (1) Single LSTM model: The long-short term memory model (long-short term memory) is a spec...

AI Technical Summary

Problems solved by technology

The basic data of the model is relatively simple, and the accuracy needs to be considered

Images