A method for pumping an oil well without stopping the pumping

Abstract

This invention proposes a method for controlling intermittent pumping in oil wells without shutting down the pump. Taking oil well productivity as the target, it determines a reasonable production range by combining pump efficiency and historical production data. Through the established IPR production curve, it calculates a reasonable bottom-hole flowing pressure range and dynamic fluid level fluctuation range, and formulates different intermittent pumping operation cycles without shutting down the pump. Combined with dynamic fluid level recovery curve testing, it determines the optimal continuous pumping and oscillation time. Through continuous correction of the intermittent pumping operation cycle, it provides the optimal intermittent pumping oil production operation cycle. By analyzing the influencing factors of the intermittent pumping operation system, and selecting nine independent factors, it employs a BPNN neural network algorithm, using the normal operating time of intermittent pumping without shutting down the pump as the analysis object, to determine the optimal operating time of the oil well.

Description

Technical Field

[0001] This invention relates to the field of oilfield pumping well production modes, specifically a method for formulating an optimal control strategy for pumping oil wells without shutting down the pumping station. Background Technology

[0002] As major oilfields in China enter the mid-to-late stages of production, the number of wells with low production and low pump efficiency is increasing year by year. These wells generally exhibit high energy consumption per 100 meters of fluid. Therefore, to improve the production of these wells and reduce energy consumption, an intermittent pumping control strategy is adopted. This involves controlling the well opening and closing time to ensure stable downhole supply and drainage, thereby achieving the goal of energy saving and increased production. However, in actual production, conventional intermittent pumping control often leads to pump jamming, rod breakage, and other malfunctions due to long well shutdown times, thus affecting production. Related studies show that traditional intermittent pumping control technology affects the average production by 0.1 m³ / s. 3 / d, the potential of some wells has not been fully realized.

[0003] Therefore, considering the shortcomings of traditional intermittent pumping control technology, a non-stop intermittent pumping control technology has been proposed in China. This technology involves a combination of continuous full-cycle operation and low-energy oscillating operation of the pumping unit crank, transforming long-cycle centralized intermittent oil production into multiple short-cycle decentralized intermittent oil production. It replaces conventional intermittent pumping shutdown operation with low-power, small-angle oscillating operation of the crank, achieving non-stop short-cycle pumping without downhole pumping. Compared with conventional intermittent pumping control technology, the energy consumption per 100 meters of fluid is significantly reduced, showing broad application prospects. In the non-stop intermittent pumping control process, the operating regime includes the pumping well operating cycle and operating time; therefore, determining the optimal operating regime is crucial. Summary of the Invention

[0004] To address the shortcomings of traditional intermittent pumping control technology, this invention proposes a method for formulating an optimal control strategy for intermittent pumping without shutting down the pumping unit. By establishing an IRP (Intermittent Pumping Reproduction) production curve, aiming for maximum well productivity, and combining well pump efficiency and historical production data, a reasonable production range is determined. The bottom fluid level depth is determined based on the production level. Based on the reasonable fluid level fluctuation range, on-site fluid level curve tests are conducted to determine the optimal continuous pumping time and oscillation time. By continuously observing fluid level changes, the optimal non-shutdown production cycle is finally determined. That is, while ensuring stable production, the fluid level fluctuation is kept within a certain range to maximize productivity; this can be considered the optimal non-shutdown intermittent pumping cycle. By analyzing the influencing factors of the non-shutdown intermittent pumping operation system, nine independent factors are selected. A BPNN (Backpropagation Neural Network) algorithm is used, with the normal operating time of non-shutdown intermittent pumping as the analysis object, to determine the optimal operating time of the pumping unit.

[0005] This invention adopts the following technical solution: a method for controlling intermittent pumping of oil wells without shutting down the pump, comprising the following steps:

[0006] The dynamic fluid level range is determined by establishing IRP production curves and combining them with historical data from pumping wells.

[0007] By testing the operating cycle of oil wells, the operating cycle of oil wells is optimized, and the optimized operating cycle of oil wells is made to meet the dynamic fluid level range.

[0008] Data is collected from the pumping unit based on the optimized pumping well operating cycle. The pumping unit variables containing the collected data are used as input to the neural network model. The non-stop pumping operation time of the pumping unit is obtained through the neural network model.

[0009] The establishment of the IRP production curve includes the following steps:

[0010] With production rate q as the x-axis, bottom hole flowing pressure p wf Using the vertical axis as the ordinate, construct a curve showing the relationship between well production and bottom hole flowing pressure:

[0011]

[0012] In the formula, q max For maximum output, p r Let q be the formation pressure, q be the well production rate, and p be the oil well production rate. wf This refers to the bottom-hole flowing pressure.

[0013] The maximum output q max :

[0014]

[0015] In the formula, q test p represents the daily oil production at the test point. wf(test) For the bottom hole flowing pressure at the test point, p r For formation pressure, q max This represents the maximum production of the oil well.

[0016] The bottom-hole flowing pressure at the test point is obtained by the following formula:

[0017] P wf(test) =P r -(H mid -L f )×ρ×g×10 -6 -P t

[0018] In the formula, p wf(test) The bottom-hole flowing pressure at the test point; p r Formation pressure; H mid It is a medium-deep reservoir; L f ρ is the dynamic fluid level; ρ is the density of the oil in the wellbore; P t For sleeve pressure.

[0019] The optimization of the pumping well operating cycle through pumping well operating cycle testing includes the following steps:

[0020] 1) The fluctuation range of the dynamic liquid level is as follows:

[0021]

[0022] Where, L gf This refers to the range of fluctuations in the dynamic liquid surface. This represents the dynamic liquid level value at maximum production.

[0023] 2) Conduct field tests on oil wells according to different oil well operating cycles to obtain the fluctuation range of dynamic fluid level under different oil well operating cycles;

[0024] 3) Select the pumping well operating cycle corresponding to the fluctuation range that conforms to the fluctuation range of the dynamic fluid level as the optimal pumping well operating cycle.

[0025] The determination of the dynamic liquid level range satisfies the following conditions:

[0026] q hismin ≤q≤q hismax η > 70%

[0027] In the formula, q is the oil well production rate, q hismin This represents the lowest production rate in the history of oil wells, q hismax η represents the highest historical production of the oil well, and η is the pump efficiency.

[0028] The variables of the pumping unit include: dynamic liquid level, production volume, power consumption, water content, oil pressure, stroke, number of strokes, pump diameter, and pump efficiency.

[0029] The construction of the neural network model includes the following steps:

[0030] There are n training samples, each with 10 parameter variables. Therefore, the expression for a training sample is:

[0031]

[0032] In the formula, x i As a learning sample, x is the predictor variable for the training samples. i1 ,x i2 ,...,x i9 Nine parameters for each training sample;

[0033] Based on training data including learning samples and corresponding non-stop pumping operation time, a neural network model is trained to obtain a neural network model for predicting non-stop pumping operation time of the pumping unit.

[0034] The present invention has the following effects and advantages:

[0035] 1. Under current technological conditions, the acquisition of surface dynamometer cards, casing pressure, and oil pressure of oil wells can be achieved in real time. Therefore, analysis based on these real-time acquired data can ensure the accuracy of optimization results.

[0036] 2. By adopting a combination of mechanistic and data models, the previous work system based on manual experience was replaced, further reducing the energy consumption of oil wells.

[0037] 3. Based on the BPNN neural network, the optimal working cycle of the oil well can be quickly predicted at the next moment, reducing the lag caused by human intervention and ensuring the real-time and efficient operation of the oil well. Attached Figure Description

[0038] Figure 1 This is the IRP curve diagram of the present invention;

[0039] Figure 2 This is a schematic diagram of the optimal production range and stroke count calculated based on the IPR curve;

[0040] Figure 3 This is a schematic diagram of the dynamic liquid level recovery curve test;

[0041] Figure 4 This is a schematic diagram of a BPNN neural network model;

[0042] Figure 5 This is a diagram showing the actual results after training. Detailed Implementation

[0043] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0044] This invention addresses the technical requirements of intermittent pumping control in oil wells by designing a novel optimal control strategy method for non-stop intermittent pumping. The invention establishes an IRP (Intake-Related Production Curve) curve, aiming to maximize well productivity, and determines a reasonable production range. The production rate determines the bottom hole dynamic fluid level depth. Based on the reasonable dynamic fluid level fluctuation range, the optimal continuous pumping time and oscillation time are determined. By continuously observing dynamic fluid level changes, the optimal non-stop pumping cycle is finally determined. This cycle maximizes productivity while ensuring stable production and keeping dynamic fluid level fluctuations within a certain range. Furthermore, by analyzing the influencing factors of non-stop intermittent pumping operations, nine independent factors are selected. A BPNN (Backpropagation Neural Network) algorithm is used, focusing on the normal operating time of non-stop intermittent pumping, to determine the optimal operating time of the oil well. This invention enables the determination of the optimal cycle and time for non-stop intermittent pumping, thereby achieving energy saving, consumption reduction, and refined management of oil wells.

[0045] 1. Establish IPR production curve

[0046] The IPR curve is the relationship between well production and bottom hole flowing pressure. When saturation pressure and other test point data are missing, the method for plotting a gas-driven IPR curve based on a single test point is as follows: production q is plotted on the x-axis, and bottom hole flowing pressure p... wf Plot a curve with the vertical axis as the ordinate:

[0047]

[0048] In the formula, q max For maximum output, p r Let q be the formation pressure, q be the well production rate, and p be the oil well production rate. wf This refers to the bottom-hole flowing pressure.

[0049] Among them, the maximum output q max The calculation method is as follows:

[0050]

[0051] In the formula, q test p represents the daily oil production at the test point. wf(test) For the bottom hole flowing pressure at the test point, p r For formation pressure, q max This represents the maximum production of the oil well.

[0052] The calculation method for the bottom-hole flowing pressure at the test point is as follows:

[0053] P wf(test) =P r -(H mid -L f )×ρ×g×10 -6 -P t

[0054] In the formula, p wf(test) The bottom-hole flowing pressure at the test point; p r H represents reservoir pressure. mid It is a medium-deep reservoir; L f ρ is the dynamic fluid level; ρ is the density of the oil in the wellbore; P t For sleeve pressure.

[0055] 2. Determine the optimal operating cycle for non-stop operation.

[0056] By plotting the IPR curve, a reasonable production range is determined, thereby identifying a reasonable bottomhole flowing pressure range. The formula for calculating bottomhole flowing pressure shows that the main factor affecting it is the dynamic fluid level. Therefore, based on the bottomhole flowing pressure range, a reasonable dynamic fluid level fluctuation range can be determined. The determination of a reasonable production range satisfies the following conditions:

[0057] q hismin ≤q≤q hismax η > 70%

[0058] In the formula, q hismin This represents the lowest production rate in the history of oil wells, q hismax η represents the highest historical production of the oil well, and η is the pump efficiency.

[0059] The determination of a reasonable range for dynamic liquid level fluctuations should aim to maximize production output. Therefore, the range of dynamic liquid level fluctuations can be defined as follows:

[0060]

[0061] Where, L gf To ensure a reasonable range of dynamic liquid level fluctuations, This represents the dynamic liquid level value at maximum production.

[0062] When the pumping unit operates at different cycles, the dynamic fluid level eventually tends to stabilize. Therefore, different operating cycles for pumping wells are established, and the optimal operating cycle for the non-stop pumping operation is determined through field testing. The operating cycles for this system are shown in Table 1.

[0063] Table 1

[0064]

[0065] Based on the IPR curve, the range of dynamic liquid level fluctuation is given under the premise of ensuring minimal change in production volume. The operating regime that corresponds to the stable dynamic liquid level in the above test results falls within this fluctuation range. The operating time plus the oscillation time is then taken as the optimal operating regime. If none of the above test schemes meet the requirements, the oscillation time can be further refined to 30 min, 35 min, 40 min... 85 min, 90 min, etc., and the test can be repeated. When the final dynamic liquid level fluctuation range is within the required range, the optimal operating cycle is finally obtained.

[0066] 3. Determine the optimal running time for non-stop operation.

[0067] Determine the factors affecting the non-stop operation time of the pumping unit.

[0068] In addition to conventional geological characteristics, factors affecting the productivity of low-yield and inefficient wells should also include oil pressure, dynamic fluid level, and production regime. Through systematic analysis of actual oil well data, the influencing factors are summarized into nine independent variables, including: dynamic fluid level, fluid production, power consumption, water cut, oil pressure, stroke, stroke frequency, pump diameter, and pump efficiency.

[0069] Establish a BPNN neural network prediction model

[0070] Establish a BPNN neural network model, such as Figure 4 As shown, the input parameters, such as oil pressure, stroke, and stroke count, are data collected by digital equipment with a collection cycle of 30 minutes. Here, the 48 data points collected in one day are weighted and averaged to obtain the average value for the day. The dynamic fluid level and water cut are data measured by oilfield field staff with a collection cycle of one day. The production volume and power consumption are calculated from the collected data and basic data. The pump diameter is oilfield basic data obtained from the oilfield A2 database. The output parameter, "non-stop pumping operation time," is data collected by the frequency converter.

[0071] Assuming there are n training samples, each with 10 parameter variables, the expression for a training sample is:

[0072]

[0073] In the formula, x i As a learning sample, x is the predictor variable for the training samples. i1 ,x i2 ,...,x i9 Nine influencing parameters for each learning sample.

[0074] Based on the training data, the BPNN neural network algorithm flow is as follows:

[0075] (1) Initialize network weights and biases: During the initialization phase, each network connection weight is given a small random number. At the same time, each neuron has a bias (the bias can be regarded as the weight of each neuron itself), which is also initialized to a random number.

[0076] (2) Forward propagation: Input training samples and calculate the output of each neuron. The calculation method for each neuron is the same, which is obtained by a linear combination of its inputs.

[0077] (3) Calculate the error and perform backpropagation: After one forward propagation, calculate the error between the actual output and the neural network output. Then, redistribute the weights and biases of each neuron using the gradient descent method to adjust the network weights and neural network element biases.

[0078] (5) Training ends: Repeat the above steps. For each sample, determine if its error is less than the set threshold or if the number of iterations has been reached, then end the training. Otherwise, proceed to the second step to continue training.

[0079] Example:

[0080] like Figure 1 , is the plotted IPR production curve.

[0081] The optimal production range and stroke count were calculated for it, and the results are as follows: Figure 2 .

[0082] Based on the IPR curve, the reasonable range of dynamic liquid level fluctuation can be determined to be 850m ± 5m.

[0083] The dynamic fluid level recovery curve test was performed, and the specific results are as follows: Figure 3 As shown in the figure, under the operating cycle of 30 minutes of operation followed by 30 minutes of oscillation, the dynamic fluid level stabilizes within the range of 847m to 853m, which meets the reasonable dynamic fluid level fluctuation range calculated above. Therefore, the optimal operating cycle for this well is 30 minutes of operation followed by 30 minutes of oscillation.

[0084] After fine-tuning the optimal operating cycle for one month, 30 test data points were obtained, as shown in Table 2.

[0085] Table 2

[0086]

[0087]

[0088] A neural network model was built based on BPNN, with a learning rate of 0.01, 5000 iterations, 10 hidden layers, and a minimum error of <0.1. The actual results after training are as follows: Figure 5 The training error of the calculation model is calculated, and the results meet the requirements of actual field use.

Claims

1. A method for controlling intermittent pumping without shutting down the pumping well, characterized in that, Includes the following steps: By establishing an IPR production curve, with production rate q as the abscissa and bottom hole flowing pressure p as the ordinate, wf Using the vertical axis as the ordinate, the bottom-hole flowing pressure range is determined by combining historical production data of the pumping well, and then the dynamic fluid level range is determined based on the bottom-hole flowing pressure range. The operating cycle of the oil well was optimized through field testing. The operating cycle includes the running time and swing time of the pumping unit crank, and the optimized operating cycle meets the dynamic fluid level range. Data is collected from the pumping unit based on the optimized pumping well operating cycle. The data includes dynamic fluid level, production rate, power consumption, water cut, oil pressure, stroke, number of strokes, pump diameter, and pump efficiency. The pumping unit variables containing the collected data are used as input to a BP neural network model, and the pumping operation time without shutting down the pumping unit is obtained through the BP neural network model.

2. The method for controlling intermittent pumping of an oil well without shutting down the pump, as described in claim 1, is characterized in that... The establishment of the IPR production curve includes the following steps: With production rate q as the x-axis, bottom hole flowing pressure p wf Using the vertical axis as the ordinate, construct a curve showing the relationship between well production and bottom hole flowing pressure: In the formula, q max For maximum output, p r Let q be the formation pressure, q be the well production rate, and p be the oil well production rate. wf This refers to the bottom-hole flowing pressure.

3. The method for controlling intermittent pumping of an oil well without shutting down the pump, as described in claim 2, is characterized in that... The maximum output q max : In the formula, q test p represents the daily oil production at the test point. wf(test) For the bottom hole flowing pressure at the test point, p r For formation pressure, q max This represents the maximum production of the oil well.

4. The method for controlling intermittent pumping of an oil well without shutting down the pump, as described in claim 3, is characterized in that... The bottom-hole flowing pressure at the test point is obtained by the following formula: P wf(test) =P r -(H mid -L f )×ρ×g×10 -6 -P t In the formula, p wf(test) The bottom-hole flowing pressure at the test point; p r Formation pressure; H mid It is a medium-deep reservoir; L f ρ is the dynamic fluid level; ρ is the density of the oil in the wellbore; P t For sleeve pressure.

5. The method for controlling intermittent pumping of an oil well without shutting down the pump, as described in claim 1, is characterized in that... The optimization of the pumping well operating cycle through pumping well operating cycle testing includes the following steps: 1) The fluctuation range of the dynamic liquid level is as follows: Where, L gf This refers to the range of fluctuations in the dynamic liquid surface. This represents the dynamic liquid level value at maximum production. 2) Conduct field tests on oil wells according to different oil well operating cycles to obtain the fluctuation range of dynamic fluid level under different oil well operating cycles; 3) Select the pumping well operating cycle corresponding to the fluctuation range that conforms to the fluctuation range of the dynamic fluid level as the optimal pumping well operating cycle.

6. The method for controlling intermittent pumping of an oil well without shutting down the pump, as described in claim 5, is characterized in that... The determination of the dynamic liquid level range satisfies the following conditions: q hismin ≤q≤q hismax ,h>70% In the formula, q is the oil well production rate, q hismin This represents the lowest production rate in the history of oil wells, q hismax η represents the highest historical production of the oil well, and η is the pump efficiency.

7. The method for controlling intermittent pumping of an oil well without shutting down the pump, as described in claim 1, is characterized in that... The variables of the pumping unit include: dynamic liquid level, production volume, power consumption, water content, oil pressure, stroke, number of strokes, pump diameter, and pump efficiency.

8. The method for controlling intermittent pumping of an oil well without shutting down the pump, as described in claim 1, is characterized in that... The construction of the neural network model includes the following steps: There are n training samples, each with 10 parameter variables. Therefore, the expression for a training sample is: In the formula, x i As a learning sample, x is the predictor variable for the training samples. i1 ,x i2 ,...,x i9 Nine parameters for each training sample; Based on training data including learning samples and corresponding non-stop pumping operation time, a neural network model is trained to obtain a neural network model for predicting non-stop pumping operation time of the pumping unit.

Images