A steam injection distribution optimization method for SAGD oil field using LSTM and dynamic programming

Steam injection distribution optimization refers to the process of distributing steam injection in steam assisted gravity drainage (SAGD) oil field to maximize the total oil production. A novel optimization method that integrates long short-term memory (LSTM) neural network and dynamic programming is presented in this paper to solve the steam injection distribution optimization problem for the first time. In the proposed method, LSTM is used to construct the prediction model to predict oil production of the wells. With the prediction result, dynamic programming optimizes steam injection distribution in the oil field to maximize total oil production. Convergence stability and computational complexity of the dynamic programming method have been analyzed and presented in this research. A web-based geographical information system called Petroleum Explorer is also developed based on the proposed method. Experiments on two pads of a real-world SAGD project demonstrate that LSTM model gives better prediction result than other five existing models and production improvement of the proposed method is highly related to parameter setting of the optimization process.