(534h) Intelligent Pressure Control Method for Managed Pressure Drilling

Tradition drilling technique using manual choke poses problems with the regularity of pressure management and higher unproductive drilling time. Managed Pressure Drilling (MPD) with automated choke controller has become increasingly popular as it provides precise control over Bottom Hole Pressure (BHP) during various drilling activities. The main challenges with automated choke control are the ability of the system to work effectively in highly dynamic operating conditions and measurement of BHP, which can be viewed as an unmeasured state. While measuring BHP using downhole sensors, the measurements are communicated with slow mud pulse telemetry. In addition, several uncertain factors such as drill pipe movement and reservoir influx lead to uncertainty. Invariably, measurement is not available when circulation is low or during pipe connection. Even models used for predicting BHP are unreliable due to uncertainties in friction loss and fluid density change.

This study presents a robust and accurate intelligent choke controller. The system uses a simple hydraulic model for the estimation of the BHP. Real-time tuning of the hydraulic model is performed using an Unscented Kalman Filter (UKF) optimization technique. Once the sensor measurements are available, UKF makes an estimate of real-time density and friction factor and makes an update to the hydraulic model. The control of the BHP is performed by using L1 adaptive control algorithm. It is a robust control method suitable for a highly non-linear system like drilling. The main components of the L1 adaptive algorithm are a state predictor, an adaptive law, and a low pass filter. The low pass filter prevents high-frequency noise from entering the controller. The adaptation law allows for the faster adaptation of the system, which in turn allows for the faster convergence of the bottom hole pressure. The L1 adaptive architecture permits the adaptive gain to be relatively high. In fact, the only limitation to the adaptive gain is determined by the hardware. This would allow for faster adaptation without losing the robustness.

Method, Procedure, Process

Governing equations were derived for the drilling system using the mass and momentum balance of the fluid flow. These equations are used for the design of estimation and control methods. L1 adaptive algorithm consists of a state predictor, an adaptive law, and an estimator. For the state predictor design, the equations were rearranged such that it contains bottom hole pressure as a state variable and three uncertain system parameters. The adaptive law makes prediction of these uncertain parameters. The estimator of L1 adaptive control law consists of the hydraulic model which makes prediction of the bottom hole pressure. Furthermore, the estimation performance is improved by real-time tuning of the method by using Unscented Kalman Filter (UKF). A process noise covariance matrix and measurement noise covariance matrix need to be designed for UKF. Measurement noise depends upon the characteristics of the sensor. Determining process noise matrix is more cumbersome. Inverse quadratic method was used for determining a relationship between elements of process noise matrix. In addition, a tuning parameter is introduced to scale the process noise matrix in relation to measurement noise matrix. The performance of UKF depends upon this scaling parameter.

Complete design of the system is implemented as blocks in Simulink. The plant was designed to simulate drilling condition. The simulation of the system was done to study its performance in various drilling scenarios like normal circulation condition, pipe connection, time delay, and in the presence of noise in the system.

Results

An innovative control method was designed for real-time estimation and control utilizing a combination of L1 adaptive algorithm and UKF. The system performance was studied with operating parameters of a drilling operation in Marcellus Shale play. UKF was observed to be effective for estimation of uncertain system variables of the drilling system. The estimation error for UKF was obtained to be less than 5 % with convergence time of 10 seconds. The simulation of combined L1 adaptive algorithm and UKF showed a stable performance for highly transient input parameters. The performance of the system was studied with changing set point of 420 bar, 460 bar, and 480 bar. The system was able to maintain a new set point within 5 seconds. In addition, the steady state error was observed to be less than 1 %. For the simulation of pipe connection, the pump flow rate was ramped down from 0.01 m3/sec to 0 m3/sec. A small amount of overshoot of about 10 bar was observed during the transient phase. For the simulation with time delay added to the sensors, stable output response was obtained for a delay of up to 3 seconds.