Research on wellbore trajectory control of Rotary Steerable System using back-propagation neural network-fuzzy method

Abstract

The core tasks of wellbore trajectory control are to control inclination and azimuth of steerable drilling tool. During the drilling process, RSS (Rotary Steerable System) constantly changes the target inclination and azimuth. The non-intelligent downhole closed-loop control method often leads to larger hysteresis of wellbore trajectory control and an increase in non-productive time. This study proposed a control methodology for RSS downhole closed-loop control, which combined a back-propagation neural network with a fuzzy control system (BP-Fuzzy). This paper also investigated the control method of PID, fuzzy, and BP. In the simulation experiments, both inclination and azimuth assigned new targets, and the performance of four control methods were evaluated with a RSS dynamic model. Furthermore, in the simulation, the fuzzy method initializes control parameters using PID values, while the BP-Fuzzy method adopts the same fuzzy rules as the fuzzy method and the same neural network structure as the BP method. Therefore, the simulation experiments are methodically sequential and control other variables. In multiple simulations, BP-Fuzzy method shows better control effect in response speed, overshoot, steady-state error and disturbance resistance. Finally, a three-dimensional drilling trajectory, encompassing vertical drilling, build-up, and horizontal drilling, was planned and implemented, with random disturbance introduced throughout the process. The BP-Fuzzy method exhibited superior performance in tracking the target attitude and demonstrated enhanced disturbance suppression capabilities during the entire drilling operation. This method can be applied to downhole closed-loop control to enhance the automatic performance of RSS and establish the foundation for future autonomous drilling.