The Impact of Riser-Induced Slugs on the Downstream Deoiling Efficiency

Abstract

In Oil and gas productions, the severe slug is an undesired flow regime due to the negative impact on the production rate and facility safety. This study examines the severe riser-induced slugs’ influence on a typical separation process, consisting of a 3-phase gravity separator physically linked to a deoiling hydrocyclone. Four inflow scenarios are compared: Uncontrolled, open-loop, feasible, and infeasible closed-loop anti-slug control, respectively. Three PID controllers’ coefficients are kept constant for all the tests: The separator pressure, water level, and hydrocyclone pressure-drop-ratio (PDR) controllers. The simulation results show that the separation efficiency is significantly larger in the closed-loop configuration, probably due to the larger production rates which provide a preferable operation condition for the hydrocyclone. It is concluded that both slug elimination approaches improve the separation efficiency consistency, but that the closed-loop control provides the best overall separation performance.

1. Introduction

In offshore Oil and gas installations, severe slug is an undesired flow regime in the well-pipeline-riser systems, as it has proved to have negative impact on the daily production [1,2,3]. The issues related to the severe slugs are numerous [4]: Overload on gas compressors, fatigue in the transporation pipelines, increased corrosion [5,6,7], production reduction [8], production slop and high pressure and liquid overflow in the downstream gravity separators [9]. Anti-slug feedback control is one effective solution for changing the slug flow to a stable flow regime [10]. A common approach is to stabilize the fluctuating pressure and/or flow by manipulating the topside choke valve at the riser top [11,12]. However, as the controllers lack robustness to process or condition changes, the operators sometimes manually choke the valves to conservatively low opening degrees to eliminate the severe slugs with the consequence of simultaneously reducing the productions [13].

The traditional separation technology in the North Sea consists of 3-phase gravity separators and deoiling hydrocyclones. This configuration represents 90% of the existing deoiling technologies in the North Sea [14]. The outlet of a well-pipeline-riser transportation process is typically physically linked to a 3-phase gas/water/oil gravity separator. The first step in the separation consists of a single or multiple stage 3-phase separators to completely separate the gas and separate most of the oil in the water. The water outlet of the last 3-phase separator is connected to a deoiling facility consisting of multiple hydrocyclone liners [15,16].

The highly fluctuating production rate induced by severe slugs occurring in the riser can cause liquid overflow in the 3-phase separator if the separator’s size is designed for a non-slugging buffer time [9]. Furthermore, the study from [15] proved that a poor separation in the separator will affect the performance in the rest of the typical produced water separation process. Thus, handling the slugs upstream the separator would be preferable as the slugs can reduce the separation efficiency of the separator, ultimately resulting in a limited production rate, reduced production quality, as well as difficulties and challenges for the produced water treatment. The study in [17] proposed control methods for the gravity separator outlet valves to handle the large slug disturbances to the separation process. However, the work only investigate the gravity separator isolated with no downstream deoiling separation included. To the best of the author’s knowledge there are no practical studies on slug’s effects on deoiling performance of the hydrocyclone; however, there has been several empirical [18,19,20] and mathematical [21,22,23,24,25] on hydrocyclone operation and efficiency. Generally they agree that the splitting of the flow in the hydrocyclone is proportional to the pressure drop ratio [26,27].

This paper will examine the relationship between severe slugs (under different running conditions) and the associated produced water treatment performance. A testing facility is used to test both the transportation and the separator-hydrocyclone separation systems [28,29,30]. The work examined in [31] concluded that the separation performance decreased when severe riser-induced slugs are presents, however, the oil-in-water (OiW) concentration was not measured or estimated due to the challenges and limitations associated with online OiW monitoring [32]. Hence, in this paper, the recently developed hydrocyclone model from [33,34] will be utilized to quantify the riser-induced slug’s impact on the OiW concentrations after the downstream separation processes.

Two separation performances will be evaluated by the models during the simulations. The first separation performance metric is concentration reducing separation efficiency

$${\epsilon }_{CRE}=1-\frac{C_u}{C_i},$$

(1)

which describes the percentage of OiW concentration reduction from the hydrocyclone inlet to hydrocyclone underflow, where $C_i$ and $C_u$ are the volumetric OiW concentration at the inlet and underflow, respectively and $Q_o$ and $Q_i$ is the hydrocyclone inlet and overflow flow rate, respectively [35,36]. The second separation performance metric is the oil removal efficiency

$${\epsilon }_{ORE}=\frac{C_o{Q}_o}{C_i{Q}_i},$$

(2)

which describes the volumetric percentage of oil that leaves the overflow [37,38].

The rest of the paper is organized as follows: Section 2 briefly describes the testing facility, Section 3 describes the mathematical model of the separation units, Section 4 describes the emulated and modelled scenarios with associated operation conditions, Section 5 examines the simulation results and finally a conclusion is carried out in Section 6.

2. Testing Facility

The testing facility described in this section is an extension of the facility examined in [39,40,41]. Figure 1 from [31] shows P&ID drawings of two parts of the laboratory testing facility where all the examined experiments are executed. The facility consists of a pipeline-riser-separator-hydrocyclone system in a complete flow-loop. Each individual part of system can be tested respectively. Figure 1a shows the pipeline-riser with the anti-slug control configuration examined in this paper, and Figure 1b illustrates the gravity separator with water level control loop and the hydrocyclone with the pressure drop ratio (PDR) control loop.

All data acquisition and control are performed using a standard PC running Simulink Real-time (xPC) through a target PC which guarantees real-time simulations. The transmitters and actuators are connected to the target PC through National Instruments (NI) data acquisition and output PCI cards, which are installed in an electrical distribution box. All input and output signals of the facility are sampled at 100 Hz.

3. Model

This section describes the model framework for used in this paper. This model framework is based on previous works, which is well-described and validated in [33], expanded by a separator-tank model and two valve controllers. The model framework is illustrated in Figure 2.

The flow rate of water entering the separator tank ($Q_{in}$) is a non-controllable input. The flow rate leaving the separator tank is the same flow rate that enters the hydrocyclone ($Q_i$). The opening degree of the underflow valve ($V_u$) is actuated by the water level controller ($P{I}_{H_w}$), which aims to maintain a constant water level ($H_w$) at the water level reference ($H_{w_{ref}}$). The opening degree of the overflow valve ($V_o$) is actuated by the pressure difference ratio controller ($P{I}_{PDR}$), which aims to maintain a constant $PDR$ at the $PDR$ reference ($PD{R}_{ref}$). The two valve models $f_{V_u}$ and $f_{V_u}$ are modelled as first order transfer functions, to emphasize that the valves are physical systems that cannot move infinitely fast. A virtual flow resistance (VFR) model, proposed in [27], estimates the flow rates leaving the hydrocyclone: Underflow flow rate ($Q_u$) and overflow flow rate ($Q_o$). Finally, an oil droplet trajectory (ODT) model, proposed in [42], extended in [43], and validated in [33], estimates the deoiling performance of the hydrocyclone.

3.1. Separator Tank

The water level in the separator tank is modeled by the relationship between accumulated volume and water level of a horizontal cylinder tank:

$$\frac{d{H}_w}{dt}=(Q_{in}-Q_i)\frac{\sqrt{H_w(2r_T-H_w)}}{r_T^2{L}_T(1-\cos \left(2{\cos }^{-1}\left(\frac{r_T-H_w}{r_T}\right)\right))},$$

(3)

where $r_T$ and $L_T$ are the radius and length of the horizontal cylinder, respectively. The accumulating volume is $Q_{in}-Q_i$ as $Q_i$ is the flow rate entering the hydrocyclone.

3.2. Hydrocyclone Part 1: Virtual Flow Resistance

The hydrocyclone model for this work is divided into two parts, where the first part solves the relevant flow and pressures, and the second part computes oil droplet trajectories and estimates separation efficiency.

The relationship between the flow rates $Q_i$, $Q_u$, and $Q_o$ and the pressures $P_i$, $P_u$, $P_o$, $P_{ub}$, and $P_{ob}$ is approximated by three virtual orifice equations, two valve equations and the continuity equation $Q_i=Q_u+Q_o$, such that $Q_u$ and $Q_o$ and the internal pressures $P_u$ and $P_o$ can be solved at every time step. This system of pressure drops is illustrated in Figure 3.

The three virtual resistances $R_{i,u,o}$ are assumed governed by

$$\Delta P=\frac{Q^2}{K},$$

(4)

where $\Delta P$ is the pressure drop over a resistance, Q is the flow rate passing through it, and K its virtual flow permeability constant.

The pressure difference over the valves are assumed as

$$\Delta P_{V_u}=\frac{Q_u^2}{{\left(K_{V_u}{V}_u\right)}^2}$$

(5)

and

$$\Delta P_{V_u}={\left(\frac{Q_o}{K_{V_{o}1}{V}_o^{\frac{1}{2}}}\right)}^2+\frac{Q_o^2}{K_{V_{o}2}^2}.$$

(6)

which are are chosen to emulate the valve properties of the used pilot plant, as described in [33].

To solve this system of equations, the boundary pressures $P_i$, $P_{ub}$, $P_{ob}$ are required to be known. The internal pressures are used to generate $PDR$ by:

$$PDR=\frac{P_i-P_o}{P_i-P_u},$$

(7)

which will be sent to $P{I}_{PDR}$ to be used as feedback.

3.3. Hydrocyclone Part 2: Oil Droplet Trajectory

With $Q_u$ and $Q_o$ defined, the ODT model estimates the velocity fields inside the hydrocyclone geometry and computes the oil droplets’ trajectory to evaluate the deoiling performance as extensively described in [33].

The principle of the model is to estimates ${\epsilon }_{ORE}$

$${\epsilon }_{ORE}={\int }_0^{\infty }G\left(D\right){\varphi }_i\left(D\right)dD,$$

(8)

where D is droplet size and $G\left(D\right)$ is the grade efficiency, also known as mitigation probability. G is estimated based on droplet trajectories [33]. To compute these trajectories, $Q_u$ and $Q_o$ from the VFR model are needed to solve the estimations of the velocity fields inside the hydrocyclone.

4. Scenario Design

The scenarios in this study are based on defined operating conditions, such as fixed pressurization of the separator and common controllers for both hydrocyclone and separator. Detailed description of the separation controllers can be found in Section 4.1 and description of the separator inflow conditions can be found in Section 4.2.

4.1. Separation Dimensions and Controllers

Common for all scenarios are the controllers as listed in

Table 1. Common controller description.

Controller	${\mathbf{PI}}_{\mathbf{PDR}}$	${\mathbf{PI}}_{{\mathit{H}}_{\mathit{w}}}$
Control variable	$u_{V_o}$	$u_{V_u}$
Process variable	$PDR$	$H_w$
Controller type	PI	PI
Controller coefficients	P = 0.011; I = 0.067	P = 81.08; I = 1.466
Setpoint	2.5	0.25 m

. The separator tank has the following dimensions: $L_T=1.5\mathrm{m}$ and $r_T=0.3\mathrm{m}$ and the separator pressure controller (controlled by the gas outlet valve) is fast compared to the dominant dynamics, such that the tank pressure is assumed constant. For all the scenarios, $P_i$ = 7 $\mathrm{bar}$.

The droplet size distribution of the mixture entering the hydrocyclone (${\varphi }_i$) is assumed as a time invariant random variable distributed according to a log-normal distribution with mean 20 μ$\mathrm{m}$ and variance 0.0005 μ${\mathrm{m}}^2$. The inlet oil concentration is assumed to be 400 ppm and constant for both experiments.

4.2. Separator Inflow Generation

The scenarios included in this paper are based on the inlet operation conditions to the separator. Three different flow conditions are considered: 1. No control, where there are fluctuating slug flow, 2. Open-loop anti-slug control where the topside valve located upstream the separator is choked below the bifurcation point, and 3. closed-loop anti-slug control, where $P_b$ is used as a feedback signal for actuating the valve.

The slugs are generated from the flow conditions from [31] (scenario 1), where a typical severe slug scenario is emulated in the same testing facility as used in this study; see Section 2. Only the flow and pressure conditions are required for simulating the impact on the downstream 3-phase separator and hydrocyclone. Therefore, the time series are included as inputs to the separator and hydrocyclone models instead of the actual slug model (described, identified and validated in [44]). Two tests are considered: One open- and one closed-loop anti-slug control test, respectively.

•

An open-loop anti-slug experiment is shown in Figure 4, where $t<400$ s is slug flow and $t>400$ s is bubble flow caused by choking below the bifurcation point of the topside valve located upstream the separator.

•

A closed-loop anti-slug control experiment is shown in Figure 5, where the controller activated at $t=400$ s eliminates the slug flow until it is given a infeasible setpoint step at 975 s which caused the anti-slug controller to fail at stabilizing the flow and pressure. The setpoint given at $t>975$ s cannot be attained due to the control valve saturation constraints of $0\%\le V_{topside}\le 100\%$.

The two experiments (in the following referred to as experiment 1 and 2, respectively), have different lengths, and thus, in Section 5 experiment 1 ends at $t=975$ s, which is exactly the point where the closed-loop controller changes setpoint and becomes unstable in experiment 2.

5. Simulation Results

This section presents the results from the two simulations using $Q_{in}$ obtained from the inflow experiments examined in Section 4.2 with the operating conditions and controllers described in Section 4.1. All the results examined in this section are summarized in

Table 2. Scenario comparison.

	Unit	Uncontrolled	Open-Loop	Closed-Loop I	Closed-Loop II
Experiment	-	1	1	2	2
Slugging	-	Yes	No	No	Yes
$t_{start}$	$\mathrm{s}$	246	600	600	1250
$t_{end}$	$\mathrm{s}$	372	900	900	1380
$Q_{in}$	$\mathrm{L}/\mathrm{s}$	0.375	0.263	0.357	0.366
$Q_i$	$\mathrm{L}/\mathrm{s}$	0.376	0.262	0.359	0.367
$Q_u$	$\mathrm{L}/\mathrm{s}$	0.364	0.254	0.347	0.355
$Q_o$	$\mathrm{L}/\mathrm{s}$	0.012	0.009	0.012	0.012
$V_u$	%	38.7	32.5	38.3	38.2
$V_o$	%	1.2	0.4	0.8	1.14
$PDR$ *	-	2.5 ± 0.185	2.5 ± 0.03	2.5 ± 0.06	2.5 ± 0.185
$F_s$	%	3.22	3.24	3.25	3.22
$C_u$ *	$\mathrm{ppm}$	71.2 ± 23.4	84.3 ± 4.80	67.2 ± 4.84	72.2 ± 21.9
${\epsilon }_{CRE}$	%	82.2	78.9	83.2	82.0
${\epsilon }_{ORE}$	%	82.8	79.6	83.8	82.5

* The numeric value following ± is the standard deviation.

where the steady-state results of experiment 1 and 2 are compared, where experiment 1 is separated into uncontrolled (100% valve opening) and open-loop control (20% valve opening) and experiment 2 is separated into closed-loop I, where the anti-slug controller works, and closed-loop II, where the change in the anti-slug controller’s setpoint causes the slug reoccur.

The controller actuating $V_u$ was able to maintain $H_w$ within 2 mm from the 25 cm liquid height reference as seen in Figure 6. The largest deviations from the reference occur during the first 400 s of both experiments, and after 2000 s in experiment 2.

Similarly, the controller actuating $V_o$ was able to maintain its 2.5 $PDR$ reference in the non-slugging period of both experiments, but had large relative deviations in the slugging periods of both experiments as seen in Figure 7. It is clear that the PDR oscillations are relatively larger than $H_w$, however, this is a consequence of the impact from varying inflow to the hydrocyclone from the separation. This is clear from Figure 8 where fast reference tracking of the level controller results in $Q_i$ very similar to $Q_{in}$ during both experiments. Hence, if the $H_w$ PI controller coefficients are decreased the oscillations will increase for $H_w$ and decrease for PDR. A tradeoff between the two responses must therefore be made, and in this study the tradeoff is based on knowledge from offshore oil and gas operators in the North Sea which both the experiments and simulations are intended to emulate.

The opening degree of $V_u$ and $V_o$ during both experiments are shown in Figure 9. As low flow rate is required to recreate slugging conditions, $V_o$ operates at low openings around 1%, whereas $V_u$ operates around 30% to 50% open. This operational range of the valves are a consequence of the physical testing facility, where the flow rate range that result in slugging flow is governed by the 6 m riser [31,44]. Although this is clearly not in the optimal operational range of the hydrocyclone, it is still sufficient for quantifying the severe slug’s impact to the separation compared to non-slug, however, the absolute separation efficiency can be improved in both experiments by increasing inflow rates and separator pressurization.

During both experiments, the underflow OiW concentration from about 45 ppm to 125 ppm and with the largest fluctuations occurring during the slugging periods as seen in Figure 10. The closed-loop slug elimination is clearly superior to the open-loop elimination although both approaches stable inflow conditions to the separator and hydrocyclone. Hence, the lower inflow rate caused by open-loop choking clearly decrease the separation efficiency, due to the hydrocyclone’s demand for not only constant, but also larger, inflow rates from the separator. Moreover, it is clear that the OiW concentration follows the slug frequency, which again demonstrates that the hydrocyclone is inflow sensitive. This applies for both the uncontrolled part of experiment 1 and 2 and the closed-loop II part. From Table 2 it is observed that the average values of $C_u$ are primarily linked to the flow rates, where the standard deviations are linked to the oscillations in flow and, thus, whether slug occurs or not.

Equivalently, the concentration reducing efficiency and the volumetric oil removal efficiency during both experiments are shown in Figure 11 and Figure 12 respectively. As expected the two graphs follow Figure 10 while it is observed that both efficiencies fluctuate between approximately 70 and 90%. This is a low efficiency range compared to typical offshore processes, but it is important to highlight that the absolute concentration cannot be directly compared to real facilities due to the laboratory scaling and the bounded operating conditions. Table 2 shows the average efficiencies and it is surprising that Closed-loop II has similar efficiencies to Uncontrolled as the Uncontrolled has significantly larger inflow rates to the separator which intuitively should results in better operating conditions for the hydrocyclone. However, it can be seen that Closed-loop I provides the best and Open-loop provides the worst separation performance.

Figure 13 shows the average oil droplet distributions of the four scenarios as well as the droplet distribution entering the hydrocyclone from the separator. Open-loop causes the largest droplets on average, while Closed-loop I both provides the smallest droplet diameters ion average as well as the smallest standard deviation.

Figure 14 visualises the OiW concentration distributions. Here, it is obvious that the slug present in both Uncontrolled and Closed-loop II leads to flat broader curve, while the non-slug in both Open-loop and Closed-loop I has a narrower span. In reality, operators are interested in keeping the OiW concentrations on the underflow as stable as possible due to issues related to OiW monitoring and the severe negative impact to the environment of large OiW concentrations discharged into the ocean, even for a short time period. Hence, even though the Open-loop scenario clearly increases the average OiW concentration it might be preferred over the alternatives where slug is present. Still, Closed-loop I is the best of all the four scenarios on both parameters.

6. Conclusions and Future Work

Previous work on slugging suppression by actuating the topside valve using the bottom riser pressure as feedback successfully concluded that the amplitude of the slugging can be reduced experimentally. However, the limitation of real-time OiW monitors, left an open question as to how much this affects the separation performance of the deoiling produced water treatment system, especially droplet diameter size, OiW concentrations and separation efficiency. To quantify the performance improvement, a grey-box model of the combined separator tank plus hydrocyclone was utilized and simulated under the same conditions as the previous experimental works.

The results demonstrate that the slug flow regime causes large variations in the OiW concentrations in the hydrocyclone’s underflow. This applies both for the uncontrolled slug and the infeasible closed-loop slug control scenarios. The average flow rate entering the separator has a substantial impact on the separation efficiency, which is clear when comparing the open-loop and successful closed-loop anti-slug control methods. Here, the closed-loop approach is superior to the open-loop approach due to the lower flow caused by the choked valve. This can be explained from the hydrocyclone demand of both pressure and inflow from the separator for optimal operation.

It can be concluded that the riser-induced slugs cause large fluctuations following the slug frequency in OiW concentrations downstream the entire separation process, ultimately decreasing the separation efficiency. Moreover, the absolute flow rate into the separator significantly impacts the separation efficiency, which demand the anti-slug technique to not limit the production rate. This was demonstrated by the simple feedback controller which was superior to the open-loop slug elimination method.

Even though this work is based on a validated experimental test from [31] and a validated model from [34], the accuracy of model’s parts and the experimental measurements can always benefit from further validation. In future work the simulations will be tested experimentally with real-time OiW monitors to further validate the results. Moreover, an extended separator model can be included for extending the separation efficiency of the separator and not only the hydrocyclone. Lastly, more advanced anti-slug controllers can be implemented to compare the control solution’s impact on the downstream separation performance.

The main contributions of the paper can be summarized as:

•

Four scenarios were experimental tested on a pilot plant emulating different multi-phase flow regimes and rates into a typical deoiling separation process. This differs from most existing literature where the separation units only have been examined isolated and not in combination [18,19,20,21,22,23,24,25].

•

The experiments were used in combination with simulations to demonstrate how severe riser-induced slugs effect the OiW downstream the separation units. Measuring low OiW concentrations has many challenges and limitations [31,32], and therefore, a validated separator and hydrocyclone model were used to simulate the OiW concentration and droplet distribution.

•

Although previous studies qualitatively showed that both the separators’ and hydrocyclones’ individual performances are linked to flow rates and pressure ratios [15,17], this study quantitatively examined these effects as well as the coupling between the units.