STRUCTURAL OPTIMIZATION OF JET PUMP BASED ON BP NEURAL NETWORK

Abstract

With the increasing drilling depth, problems follow, such as the lifting height of the liquid in the well and increased pump operation. Therefore, improving the production efficiency of deep wells is a hot spot in the current oil drilling and production industry. This paper designs a new jet pump that can be mined for high and low-pressure oil layers according to the oil wells’ inter-layer contradiction. The FLUENT software is used to simulate the new jet pump, analyze the pump efficiency factors of the jet pump and the physical properties of the fluid, and the BP neural network is used to optimize the structure of the jet pump. The results show that the maximum pump efficiency of spout distance, duct diameter, duct distance, and spread angle is 2.61-5.22 mm, 6.393-8 mm, 42.5-53.2 mm, and 6-9°, respectively. The best spout distance, duct diameter, duct distance, and spread angle are 2.74 mm, 6.80 mm, 46.5 mm, and 7.4°after being optimized by the BP neural network model presented in this paper and the optimized pump efficiency is improved by 9.45%.