The Choke as a Brainbox for Smart Wellhead Control

DOI: http://dx.doi.org/10.24018/ejers.2021.6.1.2346 Vol 6 | Issue 1 | January 2021 114 Abstract — This project uses production data to generate well-specific correlations for GLR, BSW and sand concentration which are used for predictions. A software has been developed to effect a smart control algorithm. This results in a bean up or bean down operation depending on the current flowing conditions and constraints. Excel programming environment was used to write a code that constantly takes in measured data points, models the behavior of the individual data sets with bean size and controls the choke if the parameters of interest go above a predetermined cut-off. The software was also equipped with an inverse matrix solving algorithm that enables it to determine the choke performance constants for any set of initialization data. A set of data from field X were supplied and the choke performance constants; A, B, C, D and E, were found to be 10, 0.546, 0.0, 1.89 and 1.0 respectively. In addition to that, data from subsequent production operations were entered and the software was able to control the choke size to ensure that production stays below set constraints of 500, 80 and 10 in field units for GLR, BSW and sand concentration respectively. From this, it can be concluded that the software can effectively maintain the production of unwanted well effluents below their cut-offs, thereby improving oil production and the overall Net Profit Value (NPV) of a project.


I. INTRODUCTION
Smart control involves monitoring well behavior, measuring some parameters, and taking a decision to impose a profile on the well without having to intervene physically. This requires prior installation of some necessary hardware [2]. The goal, which is to make profit means that application of this technology should be factored in at the field development stage to ensure an optimized project.
The wellhead, together with its members, aid in controlling the volume of oil which a well can produce. The rate at which a well is allowed to produce is usually being stipulated by local authorities. However, when production commences, the proposed rate may become sub-optimal. This may be due to a possible well impairment or other well problems.
Well problems make the well inefficient; causing a reduction in oil production, damaging of surface facilities, Published  increased costs from treatment of undesired fluids, and as such should be dealt with early enough. Well problems such as excessive sand, water and gas production are common in many wells. Allowing such problems to persist would negatively affect the overall Net Profit Value (NPV) of a project. The wellhead here will operate with a feedback control process where surface data is continuously measured [2]. Furthermore, based on a generated model for measured variables (Basic Sediment and Water "BSW", Gas Liquid/Oil Ratio "GLR" and sand concentration) and set constraints, flow is adjusted. Available models are either too theoretical or require some parameters to be determined offfield [3], [4]. The method used here employs already acquired production data to generate and/or improve models.
The objectives of this study are: 1. To develop an algorithm for controlling the bean size so as to produce within acceptable limits. 2. To employ a cognitive technique to determine choke performance constants. 3. To develop a user-friendly software to execute the codes.

A. Research Design
As earlier stated, the method of operation of the wellhead would be a feedback control process of constant measurement, modelling and control. However, the measurement component of the process would also be able to predict the measured variables, after which a comparison of the measured and predicted values can be made. Modelling is carried out using a developed algorithm, into which the measured/predicted values are supplied. Finally, @ @ @ @ The Choke as a Brainbox for Smart Wellhead Control Stanley I. Okafor, Azubuike H. Amadi, Mobolaji A. Abegunde, and Joseph A. Ajienka the control is effected by the choke using a reactive strategy. An algorithm is developed, and a set of programs are written to execute the process.

B. Data Source
The information needed for the project was gotten from a producing well in field X. These data are used to initialize the software and to show the effect of subsequently added data on choke correlations respectively. Production cut-offs for GLR, BSW and sand concentration used in this study are listed in Table I.

C. Choke Parameter Determination
The multiphase flow equation relates production variables (GLR, Specific Gravity 'S.G' etc.) with bean size. Using the data from the production test, several relationships of different coefficients can be generated from the multiphase relationship in Equation 1 [1].
where, q is production rate, Pwh is wellhead pressure, S is bean size, Rp is producing GLR and γo is oil specific gravity. A, B, C, D and E are choke performance constants. From this, we can solve for the constants of the multiphase flow equation. However, in order to solve for these constants, they should be treated as variables in the equation. To write a program to solve such a problem, linearization of the multiphase equation is necessary. The objective here was to solve five equations simultaneously for the variables (A, B, C, D and E). The linear equations for obtaining the choke performance constants are of the form: A matrix of equations can be generated to solve for the desired variables.

D. Model Generation and Selection
This project utilizes the production data to develop a model for each variable to be controlled, as a function of an independent variable (i.e., the bean size).
In determining the best model, several things are taken into consideration. Firstly, the ability of the model to predict the condition of the shut in well (without the aid of the algorithm), which is zero production of effluents. For gas-oil ratio prediction, the viable models should be able to predict unstable flow conditions. Furthermore, the correlation coefficient of the relationship developed is displayed to guide the engineer in making reasonable decisions. A correlation coefficient of unity means that the model gives accurate predictions of the dependent variable while that of zero means that no correlation exists between variables.

E. Method of Data Analysis
To test the viability of the software to control the flow rates of effluents, the calculated/estimated parameters after adjusting the bean size are compared to their corresponding actual/measured values using tables, plots etc. The standard deviation, error margins and other important statistical parameters are computed. Also, the correlation coefficients of the generated correlations are computed. The relationships for calculating the statistical parameters are displayed from Equations 3 to 7.
where PD is percentage difference, APD is average percentage difference, AAPD is average absolute percentage difference, SD is standard deviation, 'R' is correlation coefficient, 'E' means expected, 'M' means measured and 'n' represents number of PD values.

F. Algorithm
The program works by following the instructions shown in the flowchart in Fig. 2.

A. Research and Analysis
The steps described previously have been implemented and a software has been developed. The software is capable of executing the algorithm, performing arithmetic calculations, displaying plots and predicting choke performance constants.
The choke performance constants for a set of data are presented in this section. This is followed by an analysis of the possible models for describing the production parameters of interest. Also, the effect of the smart control on production (which is simulated by inputting a new set of data) is analyzed. Finally, the transition of selected models (i.e., model update) as a result of inputting new data (Big data) is shown.

B. Choke Performance Constants
The multiphase flow parameters for the data from field X were used to solve for the choke performance constants. The results are presented in Table II.

C. Models
The graphical plots for the different models are displayed in Appendix A.

Model Selection for GLR
The polynomial model was selected for predicting the GLR values since the statistical parameters show that it has the least deviation and since it gives the best representation for both stable and unstable flow. The results of the statistical parameters are displayed in Table A-1.

Model Selection for BSW
The power model was selected for predicting the BSW values since the statistical parameters show that it has the least deviation and since it gives a good representation of the shut in condition of the well. The results of the statistical parameters are displayed in Table A-2.

Model Selection for Sand Concentration
The power model was selected for predicting the BSW values since it has good statistical parameters gives a good representation of the shut in condition of the well. The results of the statistical parameters are displayed in Table A-3.

D. Effect of Automation
The effect of the smart wellhead software on production was tested on new set of data points, assuming cut-off values of 500 cf/bbl., 80% and 10 ppb for GLR, BSW and sand concentrations respectively. The graphical plots for the new set of data are displayed in Appendix B.

Effect on GLR
To test for GLR, it was assumed that the BSW and sand concentrations were below their cut-offs so as to see the effect of automation on the choke size due to GLR constraints. The results for the operation are shown in Table  B-1.

Effect on BSW
To test for BSW, it was assumed that the GLR and sand concentrations were below their cut-offs so as to see the effect of automation on the choke size due to BSW constraints. The results for the operation are shown in Table  B-2.

Effect on Sand Concentration
To test for sand concentration, it was assumed that the BSW and GLR were below their cut-offs so as to see the effect of automation on the choke size due to sand constraints. The results for the operation are shown in Table  B-3.

E. Effect of New Data on Models
The new data point was entirely entered into the software and this resulted in an alteration of the previous initialization model. This alteration is reflected in the constants of the mathematical model of the parameters of interest. The results are displayed in Table III.

IV. CONCLUSION
As a result of the work carried out in this project, an excel-based software capable of controlling well fluids without any need for physical intervention (of choke) was developed. The software was equipped with some minor calculative prowess including the determination of choke performance constants. The link for understanding the framework of the software build-up is shown below: https://drive.google.com/drive/folders/1mkqHEXMKpM7o4 UqmIMgXU-t3j8MVrxgp?usp=sharing.
The software proved that it is possible to control well effluents using a smart and remote approach for well production.

FUNDING
This project was partially funded by the World Bank Centre of Excellence, University of Portharcourt, Rivers, Nigeria.