RU2715593C1 - Method of operative control of water flooding of formations - Google Patents

Abstract

FIELD: oil and gas industry.

SUBSTANCE: invention relates to oil industry, namely, to control of water flooding of oil formations. Method includes extraction of oil through production wells and pumping of working agent through injection wells, wherein for determination of optimum pickups of injection wells and production rate of production well fluids, mathematical model of deposit is used, in which, as initial data for each production well and injection wells potentially influencing it, metrics are taken in form of a measurement date, intake capacity, fluid flow rate and oil fraction in the production, pressure at the bottom of injection and production wells, dynamic level of liquid in annular space of production well. Mathematical model used is combined with artificial neural network volumetric-resistive method of CRM (Capacitance-Resistive Models), which enables to obtain dependence of liquid flow rate, bottomhole pressure, dynamic level and oil production wells fraction from injection wells intake capacity current value, wherein mathematical model is adapted by obtaining minimum discrepancy between actual and design data of fluid flow rate, oil ratio, bottomhole pressure and dynamic level of each operating production well, using the neural network values of hydrodynamic resistance coefficients between wells in the CRM model are determined, recovering field of oil saturation of formation and obtaining functional dependence of daily production of liquid and oil of production well depending on intake capacity of injection wells surrounding it and production of liquid of surrounding production wells, then, total oil production is maximized in the field as a whole by reallocation of intake capacity of injection wells and control fluid recovery of production wells with imposing restrictions on injection volumes for efficient arrangement of oil displacement system with water and maintaining formation pressure.

EFFECT: technical result consists in provision of efficient organization of oil displacement system with water and formation pressure maintenance system.

1 cl, 5 dwg, 2 tbl

Description

The invention relates to the oil industry, namely, to control the injection of water into an injection well, depending on the operating conditions and the proportion of water in the production of nearby production wells.

The described method for controlling the water flooding of the companies of the Tyumen Oil Research Center LLC (method 1) is known, including redistributing the injection between injection wells based on the analysis of correlation coefficients constructed by the CRM method [Ruchkin A.A., Stepanov S.V., Knyazev A.V. Stepanov A.V., Korytov A.V., Avsyanko I.N. “Study of the features of assessing the mutual influence of wells using the CRM model as an example” // Bulletin of the Tyumen State University. Physical and mathematical modeling. Oil, gas, energy. - 2018 - T. 4 - No. 4 - S. 148-168]. The method allows to determine, according to the analytical formula, the mutual influence coefficients, on the basis of which it is possible to develop recommendations for regulating the waterflooding system. The coefficients are selected by exhaustive search to achieve a minimum in the deviation between the model and actual indicators.

The disadvantage of the described method is its inapplicability to determine the functional relationship between injection of the injection well and the water cut of the producer and, as a result, the impossibility of accurate prediction of oil production. Another disadvantage is the impossibility of applying the method in the event of a shutdown of wells, transfer of production wells to injection wells and in the presence of geological and technological measures in the wells under study (reperforation of formations, hydraulic fracturing, etc.).

The described method of controlling water flooding of the company Gazpromneft STC LLC (method 2) is known, [Khatmullin I.F., Andrianova AM, Margarit A.S., Simonov M.V., Peretz D.S., Tsanda A.P., Budyonny S.A., Lushpeev V.A. “Semi-analytical models of well interference calculation based on the class of CRM models” // Oil Industry - 2018 - No. 12]. The method allows to build a functional relationship between the flow rate of production wells and injection wells, taking into account stops and transfers. At its core, method 2 is an improvement on method 1.

The disadvantage of the described method is a multiple increase in the variation parameters. In the initial analytical model, there appear coefficients for taking into account unsteadiness of the inflow, coefficients characterizing the rate of decrease in productivity and distribution of injectivity from injection to production wells, complicating the possibility of practical application of the method. The disadvantage with the definition of water cut inherent in method 1 remains.

A known system of statistical and neural network analysis of telemetry data of oilfield objects [patent RU 2598785 LLC TatASU, authors: Bespalov Aleksey Petrovich, Akhmetzyanov Rustam Rasimovich, Ekimtsov Sergey Alexandrovich, Denisov Oleg Vladimirovich, Lazareva Regina Gennadevna], including a block for calculating production parameters based on neural network analysis and calculation unit for the interaction vectors of oil refining systems for the regulation of technological processes. However, this system is not intended to determine the optimal injection into injection wells and controls the technological processes in the oil collection and treatment system based on telemetry data of surface collection and preparation networks.

The set of features closest to the set of essential features of the claimed invention is inherent in the known method of controlling water flooding [RF patent 2614338], including determining the mutual influence of production and injection wells and forming recommendations for redistributing the injection. The method allows to determine the functional dependence of the operation of production wells on the injectivity of the fluid injection wells based on the processing of primary production information by artificial neural networks, which helps to achieve a technical result.

The disadvantages of the known method adopted for the prototype are:

- firstly, the inability to take into account the influence of production wells on each other and, as a consequence, the inability to work with a whole class of field development systems in the mode of maintaining reservoir pressure by water injection, with a tight grouping of production wells, for example three-row, in-line or focal, with grouped in space producing wells surrounded by injection.

- secondly, the lack of training of the neural network on changes in bottomhole pressure in production and injection wells, which reduces the quality of adaptation of the proxy model and the effectiveness of the method.

These disadvantages are due to the fact that the prototype of the method does not include a physical model of the fluid filtration process and to build a functional relationship between the injectivity of injection wells and fluid production in a production well, a logistic function is used that does not allow for the change in fluid production in a production well due to a change in bottomhole pressure in this well.

The task to which the claimed technical solution is directed is the effective organization of a reservoir pressure maintenance system (RPM).

In the implementation of the technical solution, the problem is solved by achieving a technical result, which consists in providing a targeted impact on the formation by injecting water in order to maintain reservoir pressure, organizing a system for oil displacement by water and controlling bottom-hole pressure of production and injection wells and production of liquid in production wells.

The specified technical result is achieved by the fact that the method of operational control of waterflooding involves the creation of a mathematical model of the field (proxy model) in which the actual and calculated production rates of the production well’s fluid are adapted depending on the injectivity affecting the injection wells and the bottom hole pressure or associated with it a functionally dynamic level of fluid in the annulus of the producing well, adaptation of the actual and estimated change in the proportion of oil to of existing wells on the basis of an increase or a decrease in injectivity, injection wells affecting it and changes in bottomhole pressure or a functionally dynamic level associated with it. At the same time, they compose the function of daily oil production of a producing well with the tuning parameters determined during adaptation of the proxy model and the functions of the fluid flow rate and changing the oil fraction to maximize the total daily oil production of the field as a whole when redistributing injection volumes of injection wells and selecting bottom-hole pressure of production wells.

Determination of optimal operating modes of injection wells contributes to obtaining maximum oil production in the field as a whole. As initial data, it uses primary production information (measurements of fluid flow rate, bottomhole pressure, dynamic fluid level, oil share of producing wells, injectivity and wellhead pressure of injection wells), which helps to achieve a technical result with minimal errors. It is important to note that in the indicated set of initial information, data on pressures and dynamic levels are not necessary and serve to improve the quality of the proxy model.

The method illustrates materials where:

in FIG. 1 presents an algorithm for implementing the method in the form of a flowchart,

in FIG. 2 - schematically shows the location of the wells, the numbers indicate: 1, 2, 4, 5, 6, 8, 9, 11, 12, 13, 15, 16, 18, 19, 21, 22, 23, 25, 26 - production wells , 3, 7, 10, 14, 17, 20, 24 - injection wells, (x, y) - well coordinates,

in FIG. 3 shows a view of a fluid production response from a change in injection,

in FIG. 4 is a schematic view of a well with a designation of pressure measurement locations,

in FIG. 5- an example of a field site for testing an algorithm.

The method is carried out by performing the following sequential actions:

search for influencing injection and production wells for each production well;

creation of a training sample for adapting the proxy model to the flow rate and the proportion of oil;

creating a training sample for bottomhole pressure or dynamic level in the wellbore to refine and adjust the proxy model;

adaptation of the proxy model to the flow rate using the coefficients of the CRM model (Capacitance-Resistive Models) and the neural network;

adaptation of the proxy model to the share of oil;

obtaining the function of the daily oil production of the producing well with the addition of the adjustment parameters of the fluid flow rate and the proportion of oil determined at the stage of adaptation of the proxy model;

setting restrictions on the injectivity of injection wells (equality equality - redistribution of injection volumes, inequality restrictions - on the minimum and maximum injection values) and restrictions on fluid production from producing wells;

determination of the optimal value of fluid production from producing wells;

determination of optimal operating modes of injection wells to ensure maximum daily oil production in the field as a whole (maximization of oil production).

The dynamic level is the absolute elevation or depth from the wellhead (in m) at which the fluid level and the space between the tubing and the casing in the well are held for one or another fluid withdrawal value. [Geological Dictionary: in 2 volumes. - M .: Subsoil. Edited by K.N. Paffengoltz et al., 1978.]

The proxy model is a mathematical model of the field, is an alternative to 3-D hydrodynamic models, allows you to get the relationship between the daily oil production of the producing well from the injectivity of the injection wells surrounding it. The proxy model likewise reproduces and allows predicting the performance of wells, it takes into account the patterns in the responses of production wells to disturbances of the injection wells surrounding it, identified empirically.

The search for impacting wells is carried out on the basis of the existing system for placing wells in the field. Potentially affecting injection wells are determined by the principle of falling into the radius R from the producing well (the radius is determined by the distance between the producing and injection wells according to the development system). That is, in the radius R13 it is required to find all the injection wells from the calculation area from each production well 13. If the distance between the production well 13 (x13, y13) and the injection well 10 (x10, y10) is greater than the radius R, then well 10 does not affect operating mode of the producing well 13. In FIG. 2 shows an example of a search for influencing injection wells for an producing well 13.

A training sample is a table of initial values of injectivity of injection wells, correlated by date with fluid flow rate, bottom-hole pressure, dynamic fluid levels in the wellbore and the proportion of oil in the producing well needed to adapt the proxy model [Transl. from Polish I.D. Rudinsky, Neural networks for information processing. - M .: Finance and statistics - 2002 - 344 p.].

For each production and injection wells affecting it, the following indicators must be compared:

- date of measurement;

- throttle response (m ³ / day);

- fluid flow rate (t / day);

- the proportion of oil (unit);

- bottomhole pressure or dynamic fluid levels in the wellbore;

- well condition (in operation / inaction).

In this case, the dynamics of the producing well (fluid flow rate, oil fraction, bottomhole pressure, or dynamic fluid levels in the wellbore of the producing well) is shifted forward by the amount of time lag that characterizes the delay in the response of the producing well when the mode of the surrounding injection well changes. The time lag is calculated based on the adaptation of the proxy model to historical indicators: that is, the time that elapses from the moment the regime in the injection well changes and the corresponding response to this change in the production well is determined.

The training sample should contain representative data on injection, fluid and oil flow rates, bottom-hole pressure, or dynamic levels. All data should be cleaned of random noise and "emissions". Unlike the prototype (patent 2614338), omissions in the operation of the injection well need not be filled with zeros or the last values.

Adaptation of the proxy model to the fluid flow rate using the coefficients of the CRM model (Capacitance-Resistive Models) and the neural network is an automatic adjustment of the fluid flow rate calculated in the proxy model to the actual fluid flow rate. An optimization problem is compiled to minimize the standard deviation of the actual and calculated fluid flow rates.

The functional dependence that describes the behavior of the production rate of a producing well’s fluid from the injectivity affecting it of an injection well is a CRM (Capacitance-Resistive Models) function; the form of the functional dependence of the production rate of a fluid on time, injectivity, and pressure change is represented by equation (1).

Where:

j is the value belonging to the j-th production well;

j is the value belonging to the i-th injection well;

q - flow rate (m ³ / s);

t ₀ , t _k , t _n - respectively, the initial, intermediate and current time instants (sec);

τ is the time constant (sec);

f _ij is the coefficient of interaction between the i-th injection and j-th production wells;

- интенсивность аквифера (м³/сек);

- the intensity of aquifer (m ³ / s);

I - throttle response (m ³ / s);

- забойное давление (Па);

- bottomhole pressure (Pa);

Functional dependence (1) was previously mentioned in the work [Ruchkin A.A., Stepanov SV., Knyazev A.V., Stepanov A.V., Korytov A.V., Avsyanko I.N. Study of the features of the assessment of the mutual influence of wells on the example of the CRM model // Bulletin of the Tyumen State University. Physical and mathematical modeling. Oil, gas, energy - 2018 - T. 4 - No. 4 - S. 148-168]

- и коэффициент взаимовлияния между i-й нагнетательной и j-й добывающей скважинами f_ij являются параметрами адаптации и подбираются для достижения минимального расхождения между фактическими и расчетными показателями дебита жидкости в целевой скважине при помощи нейронной сети.Time constant τ, aquifer intensity

- and the coefficient of interaction between the i-th injection and j-th production wells f _ij are adaptation parameters and are selected to achieve the minimum discrepancy between the actual and estimated rates of fluid flow in the target well using a neural network.

загружается в зависимость (1) из сформированной обучающей выборки. В случае отсутствия представительных данных или прямых замеров, забойное давление пересчитывается из динамического уровня по зависимости (2).Bottomhole pressure

loaded into dependence (1) from the generated training set. In the absence of representative data or direct measurements, the bottomhole pressure is recalculated from the dynamic level according to dependence (2).

where: τ _W (N) is the weighted average density of the gas-liquid mixture in the well, N is the dynamic level (m); g - acceleration of gravity (9.81 m / s ² ); Rzatr - annular pressure (Pa) determined at the point of the well in FIG. 4. Dependence (2) is well known and is the formula for the pressure of a liquid column.

The CRM method allows you to take into account the effect on the fluid flow of neighboring production wells, for this it is enough to put the fluid flow rate in equation 1 instead of the injectivity, but with a negative sign, the method also allows you to take into account the effect on the flow rate of the bottomhole pressure. In FIG. Figure 3 schematically shows the response of a change in fluid flow rate to a change in injection into an injection well. Functional dependence is physically meaningful and, when used as an activation function in a neural network, will achieve the claimed technical result.

Adaptation of the proxy model for the share of oil

Adaptation of the proxy model to the oil share is an automatic adjustment of the change in the oil share calculated in the proxy model to the actual indicators of the oil share change. An optimization problem is compiled to minimize the standard deviation of the actual and calculated oil fraction.

There are two cases of a change in the proportion of oil in a producing well with a change in the injectivity of injection wells:

1) With an increase in injectivity of the injection well, the proportion of oil in the producing well decreases.

The functional dependence that describes the behavior of reducing the proportion of oil of a producing well from the injectivity affecting it of an injection well is an exponential function:

- доля нефти добывающей скважины j на дату обучения t.Where:

- the proportion of oil of the producing well j at the training date t.

- приемистость нагнетательной скважины i на дату обучения t.

- injectivity of injection well i at the training date t.

- приемистость нагнетательной скважины i на дату обучения t-1.

- injectivity of injection well i at the training date t-1.

- настроечные параметры.

- tuning parameters.

2) When the injectivity of the injection well decreases, the proportion of oil in the well increases. The functional dependence that describes the behavior of increasing the proportion of oil of a producing well from the injectivity affecting it of an injection well is an exponential function:

- доля нефти добывающей скважины j на дату обучения t.Where:

- the proportion of oil of the producing well j at the training date t.

- приемистость нагнетательной скважины i на дату обучения t.

- injectivity of injection well i at the training date t.

- приемистость нагнетательной скважины i на дату обучения t-1.

- injectivity of injection well i at the training date t-1.

- настроечные параметры.

- tuning parameters.

Functional dependencies (2), (3) were previously mentioned in patent RU 2614338 C1.

In general terms, the objective function, which must be optimized to minimize the discrepancy between the actual and estimated oil shares, has the form of a standard error:

Where:

date ^train_begin - start date of the training set.

date ^train_end - end date of the training set.

N is the number of elements in the training set.

M^react _j - количество нагнетательных скважин действующих на одну добывающую скважину j.

M ^react _j is the number of injection wells operating on one production well j.

- доля нефти добывающей скважины j на дату t-1, д.ед.

- the proportion of oil of the producing well j at the date t-1, units

таким образом, чтобы значение функции стремилось к минимуму.The optimization problem is solved by the gradient descent method. Or by any other alternative method, it is necessary to change the settings

so that the value of the function tends to a minimum.

As a result of minimization, the optimal values of the tuning parameters are determined so that the standard deviation between the calculated and actual oil fractions tends to a minimum. Thus, the proxy model is adapted to the historical watering indicators of each producing well of the field, and with it it becomes possible to forecast the change in the proportion of oil of producing wells depending on the operating mode of the injection wells surrounding them.

Obtaining the function of daily oil production of a producing well

Adaptation of the proxy model of the field to the fluid flow rate and oil fraction depending on the injectivity of injection wells is necessary in order to obtain the function of daily oil production in the future.

According to the well-known formula, the daily oil production of a producing well is nothing more than [Yu.P. Zheltov, Oil Field Development: Textbook. for universities. - 2nd ed., Revised. and add. - M .: Nedra Publishing House OJSC, 1998, 365 pp.]:

Where:

- суточная добыча нефти добывающей скважины j, м³/сут;

- daily oil production of the producing well j, m ³ / day;

- дебит жидкости добывающей скважины j, м³/сут;

- flow rate of the producing well j, m ³ / day;

- доля нефти добывающей скважины j, д.ед.

- the share of oil of the producing well j, d.ed.

Setting limits for injectivity of injection wells is possible according to several scenarios:

1. equality constraint - redistribution of a fixed injection volume between wells;

2. limitations inequality - the injectivity of individual injection wells should not exceed the maximum injectivity determined for these wells

Setting limits on fluid production from production wells.

The minimum and maximum fluid rates are set based on the limitations of the operation of pumping equipment. In case of their absence, the minimum and maximum fluid flow rate, which was encountered in the well during its actual operation, is taken. A restriction is set on the total fluid withdrawal for all producing wells, taking into account the possibility of field infrastructure, for example, the throughput of pipelines and separators.

The optimal operating modes of injection wells are such volumes of injection into wells, taking into account the restrictions, which provide the greatest oil withdrawals at the optimum flow rate taking into account the restrictions. Therefore, to effectively organize the system of oil displacement by water and maintaining reservoir pressure, the necessary injection volumes of injection wells at the field and the necessary liquid withdrawals from the well are determined using a proxy model.

The combined sequence of actions allows us to solve the problem of effectively organizing an oil displacement system with water and maintaining reservoir pressure.

The developed method of operational control and management of oilflooding of oil reservoirs allows to solve a whole range of problems:

1. adapt to the actual operating modes of production wells;

2. build maps of the mutual influence of wells;

3. to predict the daily oil production of the producing well, depending on the operating modes of the injection wells;

4. redistribute injection and production between wells to ensure the most efficient oil displacement.

Example: a formation is drilled by one production and one injection well, through which oil is sampled and the working agent-water is injected (for example, production - 1, injection - 7, Fig. 5). Initial data on the production and injection wells are shown in table 1.

Influencing injection wells are determined for production well 1. In the case under consideration, there is only one injection well (7).

A training sample is drawn up for adapting the mathematical model of the field to the fluid flow rate, the proportion of oil, in this case it is all the data from table 1.

The time lag value is determined, for this, the dynamics of the producing well (fluid flow rate, oil fraction) are shifted by a certain amount and the mutual correlation is studied, the time lag will correspond to the maximum correlation value. The value of the time lag for this deposit is 10 days.

Based on the training sample, the proxy model is adapted for the flow rate of the liquid, that is, the tuning parameters are selected:

τ is the time constant;

e _w ^k - aquifer intensity;

f ₁₇ and the coefficient of interaction between the 7th injection and 1st production wells.

The tuning parameters are selected by the neural network so that the discrepancy between the calculated and the actual flow rate of the fluid is minimal. Table 2 presents the values of the actual and calculated by the proxy model fluid flow rate of the producing well 1 depending on the injectivity of the well 7.

Adapt the proxy model according to the proportion of oil, that is, select the tuning parameters

таким образом, чтобы расхождение между рассчитанной и фактической долей нефти было минимальным. В таблице 2 представлены значения фактической и рассчитанной по прокси-модели доли нефти добывающей скважины 1.

so that the discrepancy between the calculated and the actual share of oil is minimal. Table 2 presents the values of the actual and calculated by the proxy model of the oil share of the producing well 1.

Compose the function of the daily oil production of the producing well 1, depending on the injectivity of the well 7.

To determine the optimal operating conditions of injection well 7, and to obtain maximum oil production in well 1, the injectivity is limited: injectivity must be positive, less than the maximum injectivity according to the history of the well. To effectively organize the system of oil displacement by water and maintaining reservoir pressure in well 1, using the proxy model, the required injection volumes of injection well 2 are determined, and fluid production in production well 1 is varied.

Optimal well operation modes:

production well 1 - fluid flow rate - 680 m ³ / day.,

injection well 7 - injectivity - 750 m ³ / day.

These modes provide maximum oil withdrawal in well 1.

Claims (1)

A method of operational control of waterflooding, characterized in that it includes the selection of oil through production wells and the injection of a working agent through injection wells, assessing the effect on the production rate and water cut of production of neighboring production wells and the injectivity of neighboring injection wells, recommendations for redistributing the injection, while determining optimal values injectivity of injection wells using a mathematical model of the field, in which the initial data for each production wells and potential injection and production wells potentially affecting it, take indicators in the form of a measurement date, injectivity, fluid flow rate and oil fraction, bottomhole pressure or dynamic fluid level in the wellbore, and a combination of the analytical solution of the material balance equation is used as a mathematical model and Darcy’s Law, reflecting a change in the production rate of a producing well’s fluid with a change in the injectivity of the injection wells affecting it and the surrounding fluid’s production producing, while adapting the mathematical model by obtaining the minimum discrepancy between the actual and calculated data of fluid flow rate, bottomhole pressure and oil fraction of each producing well, determine the optimal values of the model’s tuning parameters, then maximize the total oil production over the field as a whole by redistributing the injectivity of injection wells and changes in fluid flow rates in production wells with restrictions on injection and production volumes for effective organization of oil displacement system and water to maintain reservoir pressure.

Images