AUTOMATED METHOD FOR OPTIMIZING ADJUSTMENT FACTORS OF FLOW MODELS

Abstract

The present invention consists of a method of automatic adjustment of multiphase flow models using the principle of least squares in the correction of the systematic error of simulated pressure drop and temperature drop values. This has been implemented and automated in the form of a computational algorithm and applied in the case study of an actual Production System using Marlim II simulator. For all four multiphase flow correlation sets considered, the adjustments followed each other stably, converging after a few iterations. At the end of the activity, the four sets were found to perform better than the best unadjusted set of correlations. In addition, the method provides consistent results, which is an advantage over the manual adjustment method.

Accordingly, the present invention has drastically reduced the time required for optimizing the adjustment factors of flow models and has improved quality of the adjusted model as compared to the final model obtained with manual adjustment. By better quality of the model is meant that the simulated results are closer to the measured results, that is, the model is more capable of representing the flow dynamics verified in the field. In cases with a high number of operating spots in the real system, reduction in the time required by the activity is even more significant.

Claims

1. An automated method for optimizing adjustment factors of flow models, by arranging in a scatter plot the variable values referring to the real system on the x-axis, and those obtained by the flow model on the y-axis, characterized by: a) defining an adjustment factor from a Sample Regression Function (SRF); b) defining a systematic error correction; c) adjusting the multiphase flow models applied simultaneously to the calculation of pressure drop and temperature drop in the pipeline and production column; d) inverting and transferring the values from step c) to the flow simulator and adjusting the calculated pressure and temperature gradients.

2. The method, according to claim 1, characterized in that step a) follows the steps defined by equations (3) to (10).

3. The method, according to claim 1, characterized in that step b) shifts the SRF so that it coincides with the ideal regression line, following the step of equation (11).

4. The method, according to claim 1, characterized in that step c) follows the steps defined by equations (12) to (15).

5. The method, according to claim 1, characterized in that step d) follows the steps defined by equations (16) to (19).

6. The method, according to claim 1, characterized in that the pressure drop on the temperature drop models and vice versa transform equations (12) to (15) into approximate expressions.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The present invention will be described in more detail below, with reference to the attached figures which, in a schematic and non-limiting manner of the scope of the invention, represent examples of embodiments. In the drawings:

[0022] FIG. 1A illustrates the calibration of the simulated values of pressure drop in the production pipeline;

[0023] FIG. 1B illustrates the calibration of the simulated values of temperature drop in the production pipeline;

[0024] FIG. 1C illustrates the calibration of the simulated values of pressure drop in the production column;

[0025] FIG. 1D illustrates the calibration of the simulated values of temperature drop in the production pipeline. Model with no adjustment;

[0026] FIG. 2A illustrates the adjustment of the simulated values of pressure drop in the production pipeline, where adjustment factors are shown between parentheses;

[0027] FIG. 2B illustrates the adjustment of the simulated values of temperature drop in the production pipeline, where adjustment factors are shown between parentheses;

[0028] FIG. 2C illustrates the adjustment of the simulated values of pressure drop in the production column, where adjustment factors are shown between parentheses;

[0029] FIG. 2D illustrates the adjustment of the simulated values of temperature drop in the production pipeline using the proposed methodology, where adjustment factors are shown between parentheses; and

[0030] FIGS. 3A and 3B each illustrate an example of adjustment of simulated values in the flow model: displacement of the theoretical regression line to the ideal regression line.

DETAILED DESCRIPTION OF THE INVENTION

[0031] Below is a detailed description of a preferred embodiment of the present invention, which is given by way of example and is in no way limiting. Nevertheless, possible additional embodiments of the present invention still comprised by the essential and optional features below will be clear to a person skilled in the art from reading this description.

[0032] The Production System chosen is in the Cachalote field, in the south of the State of Espirito Santo, and is connected to the FPSO Capixaba. As a lifting method, the Continuous Gas-Lift (CGL) is used throughout the field, being injected at two depths. Other basic aspects of the system are listed in Table 1. The representative fluid of the field is a heavy oil, whose main properties are shown in Table 2.

TABLE-US-00001 TABLE 1 Basic aspects of the production Unit Value Production pipe diameter [in] 8 (F) , 6 (R) Production pipe length [m] 4900 Water depth at WCT [m] 1450 Production column diameter [in] 6625 Depth measured at upper GLV [m] 2000 Depth measured at lower GLV [m] 2200 Depth measured at the PDG [m] 2520 system

TABLE-US-00002 TABLE 2 Summary of fluid properties Property Value Oil Density [API] 19 Gas density [−] 0.65 Gas-Oil Ratio [Sm.sup.3/Sm.sup.3] 70 Bubble point at 56° C. [bar] 255 Dead oil viscosity at 30° C. [cP] 320 Dead oil viscosity at 50° C. [cP] 95

[0033] Model calibration in Marlim II simulator was carried out based on 9 system operational conditions in a stabilized regime. These are recorded in Production Test Reports (PTRs), and represent a production period of approximately 28 months. Table 3 summarizes them.

TABLE-US-00003 TABLE 3 System operational conditions. Pressure on Liquid production Injected gas flow rate BSW choke flow rate Number [Sm.sup.3/d] [%] [barg] [Sm.sup.3/d] 1 4150 0 27.1 269000 2 4140 13 20.0 216000 3 4350 19 13.9 257000 4 4360 32 13.9 261000 5 4300 38 12.5 254000 6 4180 40 12.7 261000 7 4150 46 12.6 258000 8 4190 44 11.8 257600 9 4160 46 12.0 264000

[0034] In total, four vertical flow correlations were tested, namely: Beggs & Brill (1973), Duns & Ros (1963), Hagedorn & Brown (1965) and Orkiszewski (1967). For the horizontal section, the Dukler, Eaton and Flanigan (1969) correlation was chosen, with a transition angle equal to 15°. Comparison of the simulated values with the respective operational readings of the Production System is made in FIG. 1.

[0035] The final calibration result is shown in Table 4. As can be seen, the Duns and Ros correlation was the one that achieved the best results among those tested.

TABLE-US-00004 TABLE 4 Average percentage errors of the flow model in MArlim II simulator, with no adjustment. Average percentage error Vertical Horizontal ΔP in the ΔT in the ΔP in the ΔT in the flow flow production production production production correlation correlation pipeline pipeline column column Beggs & Duckler, 21.3 10.0 1.6 60.1 Brill Eaton & Flanigan Duns & Ros Dukler, 4.3 8.6 2.8 51.7 Eaton & Flanigan Hagedom & Dukler, 12.8 16.8 5.3 55.2 Brown Eaton & Flanigan Orkiszewski Dukler, 7.5 18.8 10.5 50.0 Eaton & Flanigan

[0036] Adjustment of the flow model was performed for the same set of correlations of the calibration step. Table 5 shows the adjustment factors calculated using Equations (12) to (15) in each iteration until convergence. A maximum admissible variation of less than 0.001 between two successive iterations was established as a criterion. The methodology proved to be stable. On average, 5 to 6 iterations were required to reach convergence of each of the analyzed cases. Comparison of the newly simulated values after adjustment with the respective operational readings is illustrated in FIG. 2. Note that dispersion of the results was drastically reduced compared to the previous comparison. The same conclusion is obtained by analyzing the average percentage errors of the adjusted model, as shown in Table 6. It is also noted that all four sets of correlations outperformed the best non-adjusted set.

TABLE-US-00005 TABLE 5 Calculation of factors in Equations (9) to (12), from the first iteration until adjustment convergence Vertical Horizontal flow flow Adjustment factors correlation correlation Iteration ({circumflex over (b)}.sub.ΔP.sub.data).sup.−1 ({circumflex over (b)}.sub.ΔP.sub.data).sup.−1 ({circumflex over (b)}.sub.ΔP.sub.column).sup.−1 ({circumflex over (b)}.sub.ΔT.sub.column).sup.−1 Beggs & Dukler, 1 0.8245 1.0736 0.9891 2.4113 Brill Eaton & 2 0.8041 1.2023 0.9966 2.2942 Flanigan 3 0.8006 1.2163 0.9982 2.2477 4 0.8004 1.2173 0.9985 2.2366 5 0.8004 1.2171 0.9986 2.2348 6 0.8004 1.2171 0.9986 2.2348 Duns & Ros Dukler, 1 0.9629 1.0602 0.9754 1.9971 Eaton & 2 0.9539 1.1065 0.9746 2.1171 Flanigan 3 0.9529 1.1160 0.9747 2.1173 4 0.9527 1.1168 0.9748 2.1160 5 0.9525 1.1169 0.9748 2.1159 Hagedorn & Dukler, 1 0.8879 0.8598 1.0483 2.1589 Brown Eaton & 2 0.8748 0.9135 1.0485 2.1013 Flanigan 3 0.8736 0.9187 1.0499 2.0780 4 0.8736 0.9190 1.0501 2.0731 5 0.8737 0.9188 1.0501 2.0725 Orkiszewski Dukler, 1 1.0774 1.2040 1.1155 1.9367 Eaton & 2 1.0575 1.2025 1.0915 2.1320 Flanigan 3 1.0550 1.2189 1.0943 2.1511 4 1.0541 1.2215 1.0943 2.1482 5 1.0541 1.2220 1.0945 2.1469 6 1.0541 1.2220 1.0945 2.1466

TABLE-US-00006 TABLE 6 Average percentage errors of the flow model in Marlim II simulator, adjusted. Average percentage error Vertical Horizontal ΔP in the ΔT in the ΔP in the ΔT in the flow flow production production production production correlation correlation pipeline pipeline column column Beggs & Duckler, 1.7 5.4 1.2 11.6 Brill Eaton & Flanigan Duns & Ros Dukler, 1.5 5.2 1.2 11.2 Eaton & Flanigan Hagedom & Dukler, 2.1 3.5 2.7 9.8 Brown Eaton & Flanigan Orkiszewski Dukler, 1.4 6.0 1.5 11.1 Eaton & Flanigan

[0037] The method of the present invention is implemented in optimization systems of multiphase flow models in the company. As an example, the BR-SiOP system and the PYEE repository can be mentioned, which contain Python codes with algorithms that are useful for Lift and Flow Engineers. The method has been available internally at Petrobras since 2015. It is estimated that it has been successfully used hundreds of times since then. To the authors' knowledge, the method has never been published in conferences or journals external to Petrobras.

[0038] The method can be used to optimize adjustment factors of flow simulation models in onshore and offshore production and injection wellbores.

[0039] The proposed adjustment methodology is equivalent to displaying, in a scatter plot, values of variables referring to the actual system on the x-axis, and those obtained by the flow model on the y-axis (see FIG. 3A).

[0040] The relationship between measured (x) and simulated (y) data is assumed to be linear for the purpose of simplifying the adjustment, being written as:

y=bx  (3)

where the slope b is estimated by {circumflex over (b)} from a sample of (x, y) pairs. Thus, the Sample Regression Function (SRF) will be given by:

ŷ={circumflex over (b)}x  (4)

and application of the principle of the least squares leads to imposing the condition

[00001]$\begin{array}{cc}\min \underset{i=1}{\overset{n}{.Math.}}{d}_i^2=\min \underset{i=1}{\overset{n}{.Math.}}{\left(y_i-\hat{b}{x}_i\right)}^2& \left(5\right)\end{array}$

Canceling the derivative with respect to b:

[00002]$\begin{array}{cc}\frac{d}{db}\underset{i=1}{\overset{n}{.Math.}}{\left(y_i-\hat{b}{x}_i\right)}^2=0& \left(6\right)\end{array}$$\begin{array}{cc}-2\underset{i=1}{\overset{n}{.Math.}}{x}_i\left(y_i-\hat{b}{x}_i\right)=0& \left(7\right)\end{array}$$\begin{array}{cc}\underset{i=1}{\overset{n}{.Math.}}{x}_i{y}_i=\hat{b}\underset{i=1}{\overset{n}{.Math.}}{x}_i^2& \left(8\right)\end{array}$$\begin{array}{cc}\hat{b}=\frac{{.Math.}_{i=1}^n{x}_i{y}_i}{{.Math.}_{i=1}^n{x}_i^2}& \left(9\right)\end{array}$

Finally, the adjustment factor is defined as

[00003]$\begin{array}{cc}\phi =\frac{1}{\hat{b}}& \left(10\right)\end{array}$

The systematic error correction step can be interpreted as a displacement of the SRF so that it coincides with the ideal regression line, as illustrated in FIG. 3B. This is achieved in a simple manner by

y.sub.i,aj=φy.sub.i, ∀ 1≤i≤n  (11)

Adjustment of multiphase flow models must be applied simultaneously in the calculation of pressure and temperature drop in the pipeline and production column, thus totaling four variables if the TPT and PDG sensors are both functional. After simulating the n operational conditions of the real system, the slopes of the regression line are given, respectively, by

[00004]$\begin{array}{cc}{\hat{b}}_{\Delta P_{\mathrm{pipe}}}=\frac{{.Math.}_{i=1}^n\Delta P_{\mathrm{pipe},i}^{\mathrm{meas}}\Delta P_{\mathrm{pipe},i}^{\mathrm{sim}}}{{.Math.}_{i=1}^n{\left(\Delta P_{\mathrm{pipe},i}^{meas}\right)}^2}& \left(12\right)\end{array}$$\begin{array}{cc}{\hat{b}}_{\Delta T_{\mathrm{pipe}}}=\frac{{.Math.}_{i=1}^n\Delta T_{\mathrm{pipe},i}^{\mathrm{meas}}\Delta T_{\mathrm{pipe},i}^{\mathrm{sim}}}{{.Math.}_{i=1}^n{\left(\Delta T_{\mathrm{pipe},i}^{meas}\right)}^2}& \left(13\right)\end{array}$$\begin{array}{cc}{\hat{b}}_{\Delta P_{column}}=\frac{{.Math.}_{i=1}^n\Delta P_{\mathrm{column},i}^{\mathrm{meas}}\Delta P_{\mathrm{column},i}^{\mathrm{sim}}}{{.Math.}_{i=1}^n{\left(\Delta P_{\mathrm{column},i}^{\mathrm{meas}}\right)}^2}& \left(14\right)\end{array}$$\begin{array}{cc}{\hat{b}}_{\Delta T_{column}}=\frac{{.Math.}_{i=1}^n\Delta T_{\mathrm{column},i}^{\mathrm{meas}}\Delta T_{\mathrm{column},i}^{\mathrm{sim}}}{{.Math.}_{i=1}^n{\left(\Delta T_{\mathrm{column},i}^{\mathrm{meas}}\right)}^2}& \left(15\right)\end{array}$

Then, these values must be inverted and passed on to the flow simulator, so that, for each model section the pressure and temperature gradients calculated thereby are adjusted. It can be expressed mathematically as

[00005]$\begin{array}{cc}\frac{\mathrm{dP}}{dL}{.Math.}_{\mathrm{pipe},i}^{aj}=\frac{1}{{\hat{b}}_{\Delta P_{\mathrm{pipe}}}}\frac{\mathrm{dP}}{dL}{.Math.}_{\mathrm{pipe},i}^{sim}={\phi }_{\Delta P_{\mathrm{pipe}}}\frac{\mathrm{dP}}{dL}{|}_{\mathrm{pipe},i}^{sim}& \left(16\right)\end{array}$$\begin{array}{cc}\frac{\mathrm{dT}}{dL}{.Math.}_{\mathrm{pipe},i}^{aj}=\frac{1}{{\hat{b}}_{\Delta T_{\mathrm{pipe}}}}\frac{\mathrm{dT}}{dL}{.Math.}_{\mathrm{pipe},i}^{sim}={\phi }_{\Delta T_{\mathrm{pipe}}}\frac{\mathrm{dT}}{dL}{|}_{\mathrm{pipe},i}^{sim}& \left(17\right)\end{array}$$\begin{array}{cc}\frac{\mathrm{dP}}{dL}{|}_{co\mathrm{lumn},i}^{aj}=\frac{1}{{\hat{b}}_{\Delta P_{column}}}\frac{\mathrm{dP}}{dL}{|}_{co\mathrm{lumn},i}^{sim}={\phi }_{\Delta P_{column}}\frac{\mathrm{dP}}{dL}{|}_{co\mathrm{lumn},i}^{sim}& \left(18\right)\end{array}$$\begin{array}{cc}\frac{\mathrm{dT}}{dL}{|}_{co\mathrm{lumn},i}^{aj}=\frac{1}{{\hat{b}}_{\Delta T_{column}}}\frac{dT}{dL}{|}_{co\mathrm{lumn},i}^{sim}={\phi }_{\Delta T_{\mathrm{pipe}}}\frac{dT}{dL}{|}_{co\mathrm{lumn},i}^{sim}& \left(19\right)\end{array}$

[0041] Due to the influence of pressure drop models on temperature drop models and vice versa, non-linearities arise, which transform Equations (12) to (15) into approximate expressions. Some iterations are necessary for variation of the coefficients be sufficiently small, when the adjustment convergence is finally verified.

[0042] It should be noted that if the TPT sensor is non-functional it will be impossible to distinguish errors in the pipeline from errors in the production column, and the calculated coefficients are only two. Also, if PDG is non-functional, the coefficients in the column cannot be calculated, only those in the pipe. Finally, if both sensors are non-functional, the error calculation (and subsequently, the model adjustment) cannot be performed.

[0043] Although not illustrated in the present invention, adjustment of pressure drop models in the service pipeline and IPR can be performed using this same methodology.