Statoil: Automatic MPD a coming reality

R&D focuses on developing hydraulic model, choke controller using intelligent automatic control solutions

By Alexey Pavlov, Glenn-Ole Kaasa, Statoil Research Centre

Managed pressure drilling (MPD) is a technology that enables fast and accurate control of the pressure in the well during drilling and completion operations. This is achieved by sealing the annulus and diverting the mud outflow to a choke manifold. By opening or closing the choke, one can actively control the pressure at a desired location in the well.

The main benefit of MPD is that it enables drilling wells with narrow pressure margins that are undrillable with conventional technology. Additionally, MPD technology’s hydraulic models and measurements allow for flow and pressure conditions in the well to be monitored in an accurate manner, thus allowing for kick and loss situations to be detected at an early stage. Once detected, they can be handled quickly through active pressure regulation. This leads to reduction of nonproductive time.

All these benefits of MPD have been demonstrated by numerous successful MPD applications worldwide, both onshore and offshore. This may be perceived as a proof of maturity of MPD technology, which is indeed the case for manual MPD. Even though in the future we will definitely witness improvements of MPD hardware (e.g., annular seal and choke), they will neither change the essence of manual MPD operations nor significantly improve its efficiency. It is the human factor that to the largest extent limits the efficiency of manually controlled MPD.

For automated MPD, the situation is different. The automatic control system is the component that allows one to get the most out of the MPD hardware. The control system, which provides accurate pressure control in an automatic way, resembles to some extent an autopilot in an airplane. Still, the level of maturity of MPD control systems, compared with automatic control systems like the autopilot, is rather low. An autopilot can control the process of flying with very little attention from the pilot.

In many cases, the autopilot controls the airplane in a safer and more efficient way than pilots are able to do manually, and even landings in bad weather conditions are nowadays safer with autopilots. Despite all the complexity of modern airplanes, the operation of an autopilot is as simple as setting the desired course and flight parameters, and pressing an on/off button.

This simplicity of operation in combination with high safety and efficiency of autopilots is made possible due to advanced automatic control systems and is one of many examples that demonstrate the power of modern control solutions.

In comparison, an automated MPD system, even though equipped with an automatic control system, still requires a whole crew to operate it. It needs constant attention and adjustments (e.g., to update well and mud parameters that vary during drilling), and depends to a large degree on qualified manual interaction. This comparison leads to the conclusion that simplicity (or currently complexity) of MPD operations, as well as their efficiency, can be significantly improved by applying advanced automatic control solutions.

This conclusion motivated the development of an MPD control system in Statoil R&D, which started in 2007 with the goal to improve efficiency, safety and simplicity of MPD operations by intelligent automatic control solutions.

Following theoretical development and numerous simulations studies, the first version of this control system was successfully tested in full-scale experiments at a drilling test facility in Stavanger, Norway. Some components of the control system were also tested against field data from MPD operations in the North Sea. To present these results and the benefits of this control system, let’s first have a closer look at an MPD control system in general.

The main objective of an MPD control system is to maintain the pressure in the well, typically at the casing shoe or the bottom of the well, within the pore and fracture pressures. This is achieved by controlling the pressure to a specified value within these margins, which is determined in the operation planning phase. As shown in Figure 1, the control system consists of two parts:

• Hydraulic model,

• Choke controller.

The hydraulic model computes a reference value pRef_c for the backpressure (upstream the choke) needed to achieve the desired pressure pSP_dh in the well. This is done based on well parameters, top-side measurements and, when available, downhole measurements (received either through mud-pulse telemetry or a wired pipe). The choke controller automatically opens or closes the choke to bring the backpressure to the reference value generated by the hydraulic model.

In this way one can control the pressure in the desired location of the well. Efficiency, robustness and usability (i.e., how easy it is to operate the control system and how much attention it requires) of the overall control system depend primarily on the hydraulic model and the choke controller. Therefore, in the work of developing an MPD control system, our attention has been focused in particular on developing and improving these two parts of the control system.

Existing automated MPD systems are mostly based on conventional automatic control solutions developed for process industries, where processes are usually slow. Since pressure transients during drilling operations are typically much faster than in the process industry, these conventional control solutions are not particularly suited for MPD. Conventional proportional-integral (PI) controllers are reactive by nature. This means that a PI controller starts acting only after deviations of the backpressure from its desired value are observed. It does not take into account any other information, even though it may be available.

This limits the ability of the PI controller to compensate for fast changes in the operating conditions. This in turn impos es constraints on operations, such as ramping of rig pumps and tripping (surge and swab), which reduces the attainable efficiency of the operation. In addition to that, a PI controller has to be tuned for particular well conditions, such as mud compressibility, well length (volume), and pressure conditions. Significant changes, for example, in pressure conditions, which may occur in a critical situation, may result in poor performance or even instability of the control system.

The advanced choke control system developed by Statoil effectively utilizes available top-side measurements to compensate for changes in the operating conditions in a proactive way. It starts acting the moment operating conditions start changing and before the effect of this change becomes noticeable in the backpressure. The control system thus compensates for these changes more efficiently.

This enables, in particular, faster operations, such as faster ramping of the rig pump (e.g., reducing the connection time) and faster tripping (surge and swab) operations, while maintaining accurate pressure regulation. Moreover, the control system has uniform performance over the full range of pressures, which eliminates the need for re-tuning the controller for different pressure conditions. More importantly, this means that the control system provides reliable performance in case of large pressure variations, thus improving safety and handling of well control situations.

This control system may seem more vulnerable to the loss of measurements compared with a PI controller, which uses only the backpressure. Even if all the additional measurements used by the advanced controller are lost, it will lose only its proactive capabilities but retain its uniform performance, i.e., it will still perform better than the conventional PI controller.

Performance of the controller was tested in experiments at a full-scale drilling rig in Stavanger. Figure 2 shows test results with stepping reference value for the choke pressure (also known as backpressure) and demonstrates uniform performance in the pressure from 15 to 90 bar, which was the maximum allowed on the rig, without any re-tuning for different pressures.

Similar tests performed with a PI controller tuned for 50 bar, demonstrated good performance in the pressure range from 40 to 60 bar, slow performance around 10 to 30 bar, and instability, i.e., violent oscillations resulting in control system shut down, around 80 bar.

The next test (Figure 3) emulates a connection scenario where the rig pump is ramped down from 1,000 l/min to zero flow, and then ramped up again to 1,000 l/min. The test was performed without a backpressure pump, which makes the task for the controller even more challenging. The figure illustrates excellent performance in the case with 60 sec ramping time (period from 0 sec to 200 sec in Figure 3) with 0.6 bar regulation error in steady-state (no flow conditions) and 1.9 bar peak regulation error.

A similar test was performed with a PI controller, resulting in 5.6 bar steady-state error. Clearly the advanced choke controller demonstrated much better performance than the PI controller. It demonstrated good performance even in an extreme connection test with 30-sec ramping time (see Figure 3, period from 200 sec to 350 sec), which resulted in the steady-state error of 1.9 bar and a peak error of 3.4 bar. The PI controller was not tested with 30 sec ramping time due to the poor performance observed in the 60 sec test.

Other tests with varying drillstring rotation (up to 150 rpm), surge and swab with 16 m/min block velocity, clearly demonstrated a superior performance of the advanced choke controller compared with the PI controller.

The hydraulic model is an essential part of an automated MPD control system. In many cases it is the limiting factor for achievable accuracy of the system. A lot of effort has therefore been put into developing advanced hydraulic models that capture all aspects of the drilling hydraulics. Such models are well suited for the planning phase of MPD operations. In real-time MPD, however, the main drawback of these models is the resulting complexity, which requires expert knowledge to set up and calibrate, making it a high-end solution with reduced usability.

In practice, much of the complexity does not contribute to improve the overall accuracy of the pressure estimate, simply because conditions in the well change during MPD operations, and because there are not enough measurements to keep all parameters of the advanced model calibrated. Moreover, the output of the hydraulic model is used by the choke controller and eventually by the choke actuator, which has its limitations in the opening/closing rate. Therefore, fast variations of this output, which result from some fine details of an advanced hydraulic model, eventually do not contribute to the overall accuracy of the pressure regulation. The choke does not simply compensate such fast variations.

To reduce the complexity of a hydraulic model while preserving accuracy, Statoil developed a simplified hydraulic model. This model, which is based on basic fluid dynamics, is able to capture the dominating hydraulics of an MPD system. Further, its simplicity (and therefore its very few parameters) enables its automatic online calibration. This is done by applying algorithms for online parameter estimation similar to those used in advanced control systems in the automotive and aerospace industry.

The resulting level of accuracy is comparable to that of a calibrated advanced hydraulic model (details on the hydraulic model can be found in IADC/SPE 143097, “Intelligent Estimation of Downhole Pressure Using a Simple Hydraulic Model,” by G. Kaasa, O.N. Stamnes, L. Imsland, O.M. Aamo).

Moreover, simplicity of the model is essential in analyzing its robustness. In a failure-critical application like MPD, all parts of the control system must be analyzed for performance in case of conditions diverting from normal operation. While complexity of an advanced model is prohibitive for such an analysis, it is possible for the simplified model.

Performance of the simplified model has been successfully tested against field data from MPD operations in the North Sea and during dedicated experiments at the full-scale drilling rig in Stavanger. Figure 4 illustrates performance of the hydraulic model for field data logged during MPD commissioning tests in the North Sea. In this case, uncertainty of the model is lumped into three parameters: θ_F, representing uncertainty in the friction model (both for the drillstring and annulus); θ_FA, representing uncertainty in the friction model in the annulus only; and the uncertainty in the mud density θ_ρ.

In the figure, these uncertain parameters have been scaled such that they correspond to the correct value when they are equal to 1. In the hydraulic model with automatic calibration, θ_F and θ_ρ are updated based on the top-side measurements only. The calibration of the parameter θ_FA is based on PWD measurements when they are available.

In the case presented in Figure 4, the initial value of the mud density has been chosen 10% higher than its nominal value, and friction parameters have been chosen 50% higher than their nominal values. This results in the initial pressure overestimation by almost 30 bar.

Once the autocalibration functionality is activated, the uncertain parameters quickly converge to their nominal values and, as a result of that, the estimate of the downhole pressure converges to and stays on the actual downhole pressure (its measured values are shown in the upper plot of Figure 4). This example demonstrates excellent performance of the simplified hydraulic model with autocalibration even with such a significant initial offset in the uncertain parameters.

The advanced choke controller and the fit-for-purpose hydraulic model developed by Statoil clearly demonstrate the potential of advanced automatic control solutions in MPD. In fact, this establishes a trend for further development of MPD systems: higher efficiency, reliability and simplicity of operation achieved through advanced automatic control solutions. Following this trend, automated MPD operations in the relatively near future must be as simple and routine as flying an airplane with an autopilot.

As a part of the normal drilling operation, a driller will specify desired pressure parameters and activate a “drilling autopilot” – an intelligent automatic control system running the drilling process with full control of the well hydraulics and requiring attention from the driller only in exceptional situations. Numerous examples of successful applications of advanced control systems in other industries as well as in-depth academic studies and our own expertise and experience within control systems indicate that this is not just a vision, but a coming reality.