Fracturing curve intelligent classification method
By using a data correction method based on multi-sensor fusion and variational physical constraints, the problems of time-consuming, labor-intensive, and unreliable traditional fracturing curve classification are solved, enabling real-time, accurate classification and reliable estimation of fracturing curves.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- WUHAN SHENGHUAWEIYE TECHNOLGY CO LTD
- Filing Date
- 2026-04-02
- Publication Date
- 2026-05-15
AI Technical Summary
Traditional fracturing curve classification methods are time-consuming and labor-intensive, easily affected by subjective human experience, and the reliability of single-point measurement is insufficient, making it difficult to meet the needs of real-time monitoring.
By integrating multiple sensors and quantifying uncertainty, a reliability weight model is introduced, and data correction is performed using variational physical constraints. The pressure curve is then decomposed to obtain estimated values for displacement and sand ratio, enabling real-time classification of fracturing curves.
It enables real-time and accurate classification of fracturing curves, reduces the impact of faults, improves the confidence of data correction and the reliability of displacement estimation, ensures that the inversion results conform to physical conditions, and supports timely early warning and proactive intervention.
Smart Images
Figure CN122045954A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of drilling monitoring technology, and in particular to an intelligent classification method for fracturing curves. Background Technology
[0002] Every year, there are tens of thousands of hydraulic fracturing operation curves. The traditional classification method is to classify them manually after the operation, taking into account the reservoir characteristics and the changes in various parameters during the operation. This method is not only time-consuming and labor-intensive, but the classification results are also easily affected by the subjective experience of the classifiers.
[0003] Sensors are commonly used to obtain fracturing operation curves, but single-point, isolated measurements are not only unreliable but may also produce unreasonable output results due to the lack of physical constraints. Estimating the displacement during fracturing operations is difficult, and the classification of overall fracturing curves is mainly based on offline, manual analysis, which cannot meet real-time requirements.
[0004] Therefore, it is essential to provide an intelligent classification method for fracturing curves. By using multi-sensor fusion, uncertainty quantification, and physical constraint data correction, this method can facilitate real-time monitoring and curve classification of fracturing operations, enabling timely early warning and prompting for proactive intervention. Summary of the Invention
[0005] In view of this, the present invention proposes an intelligent classification method for fracturing curves that can integrate data from different types of sensors, add physical constraints to enhance the rationality of the data, and obtain displacement information through pressure inversion estimation.
[0006] This invention provides an intelligent classification method for fracturing curves, comprising the following steps: S1: Configure multiple sensors for pressure operation, fuse and quantify the measurement values obtained by each sensor, perform data correction based on variational physical constraints, and output the corrected pressure curve; S2: Perform inversion decomposition on the corrected pressure curve, define the inversion optimization model and inversion constraints, and solve to obtain the decomposed corrected pressure measurement value; based on the corrected pressure measurement value, further obtain the estimated sand ratio and displacement value for pressure operation, and verify the displacement estimate value; S3: Based on the obtained corrected pressure measurement value after decomposition, the displacement estimate value and the sand ratio estimate value after verification, identify the displacement estimate value after it reaches the stable period, analyze the pressure ratio characteristics and sand ratio dynamic characteristics of the corrected pressure measurement value after decomposition during the stable period, and classify the fracturing curve.
[0007] Based on the above technical solution, preferably, the fusion and uncertainty quantification of the measured values acquired by each sensor in step S1 specifically includes acquiring the original measured value of pumping pressure acquired by the pumping pressure sensor, the original measured value of casing pressure acquired by the casing pressure sensor, and the original measured value of tubing pressure acquired by the tubing pressure sensor, and letting the original measured value of the sensor be... , subscript p , c , t These represent the pump injection pressure, casing pressure, and tubing pressure, respectively. (The text in parentheses is incomplete and cannot be translated.) t Represents a time variable. , T For time variables t The upper limit of the value, i and j Let [1, 2, 3] be the index of the sensor; based on the original measured values, the variance of the original measured values, and the predicted value of the sensor's measured pressure, a reliability weighting model for the sensor is established. The raw sensor measurements are weighted by reliability and then summed to obtain the fused pressure estimate. Combined with fusion pressure estimation Sensor reliability weight model The standard deviation of the original measurements and the correlation coefficient between the sensors are used to obtain the fusion uncertainty. The standard deviation of the original measurements is the square root of the variance of the original measurements.
[0008] Preferably, the data correction based on variational physical constraints described in step S1, and the output of the correction pressure curve, specifically includes defining a measurement pressure function. Predicting pressure function Correction pressure curve Measure pressure function The pressure estimation is derived from multi-sensor fusion; an adaptive weighting function, a total variation regularization term, and a physical constraint term are defined; an objective function is established based on the adaptive weighting function, the total variation regularization term, and the physical constraint term; and the corrected pressure curve is obtained through gradient descent. .
[0009] Further preferably, step S2 involves inverting the correction pressure curve, defining an inversion optimization model and inversion constraints, and solving for the decomposed corrected measured values. Specifically, this includes setting the actual pumping pressure, casing pressure, and tubing pressure to be determined as follows: An inversion optimization model is established with constraints set to obtain the frictional pressure drop inside the oil pipe at this point. Local pressure drop in a real perforation .
[0010] More preferably, in step S2, the displacement estimate for pressure operation based on the calibrated measurement value is obtained by acquiring the frictional pressure drop in the oil pipe at this time. Then, combining the Darcy-Weisbach equation with the formula for calculating the frictional pressure drop in the oil pipe, the Darcy-Weisbach equation was rewritten to obtain the displacement estimate. Q ( t The calculation formula for displacement is used to obtain the estimated displacement value. Q ( t ).
[0011] Further optimization involves verifying the displacement estimate to obtain the displacement estimate. Q ( t After that, based on the displacement estimate Q ( t The estimated value of the local pressure drop of the perforation is obtained by using the formula for calculating the local pressure drop of the perforation. Estimate the local pressure drop of the perforation. Is it due to local pressure drop in a real perforation? Within the deviation range: if the estimated value of the local pressure drop of the perforation is... Local pressure drop not exceeding the actual perforation depth If the displacement estimate is ±15%, then the displacement estimate is considered to be... Q ( t Reasonable; if the estimated local pressure drop of the perforation is... Local pressure drop exceeding that of a true perforation If the value is ±15%, then perform any of the following operations until the estimated local pressure drop of the perforation is met. Local pressure drop not exceeding the actual perforation depth Up to ±15%: 1) Keep the calculation formula for the estimated value of the local pressure drop of the perforation unchanged, and adjust the estimated value of the local pressure drop of the perforation. The coefficients in the formula; 2) Keep the calculation formula for the estimated local pressure drop of the perforation unchanged, and adjust the estimated displacement value. Q ( t ) The coefficients in the expression; 3) Use polynomial fitting to replace the calculation formula for the current estimate of the local pressure drop of the perforation.
[0012] More preferably, obtaining the estimated sand ratio in step S2 involves defining a pressure fluctuation index and fluctuation frequency characteristics; and constructing a linear model for sand ratio estimation based on the pressure fluctuation index and fluctuation frequency characteristics. SR(t) Calculate the estimated value of the sand ratio.
[0013] Furthermore, step S3 specifically includes: S31: Define time periodW Set the conditions for determining displacement stability; S32: Set feature thresholds and calculate features, and make classification decisions based on the relationship between features and feature thresholds.
[0014] Preferably, step S31 specifically includes: displacement stability needs to simultaneously meet the following conditions: 1) Obtaining displacement estimate. Q ( t The corresponding time-varying sequence [ Q ( t ], given time period W Defined by time t The sliding window centered on the given time period W Inside, all displacement estimates within the sliding window. Q ( t The mean and standard deviation of ) are and The stability of the coefficient of variation is defined as follows: Divide by And not exceeding the mutation threshold ; 2) To stabilize the trend ,in During the time period W Internally, through the time-varying sequence [ Q ( t The trend slope obtained by performing linear regression; 3) Let the sequence change over time [ Q ( t The autocorrelation function of ] is , representing a sequence [ Q ( t )] and the sequence in delay The following sequence The correlation, For maximum delay, It is a time-varying sequence. Q ( t During the time period W variance within, autocorrelation function Divide by variance Autocorrelation stability is obtained; 4) Define the stable period W sp To ensure that the stability of the coefficient of variation, trend stability, and autocorrelation stability all do not exceed a set threshold within a certain time period, let the stability period be... W sp Maximum duration With minimum duration The difference is the stable duration. .
[0015] Further preferably, step S32 includes the following specific contents: Based on the estimated sand ratio SR ( t During the stable period W sp Within the sequence, the sand ratio estimate is obtained by using the trend slope obtained through linear regression. SR ( t The slope of the sand ratio trend corresponding to the sequence of ) Define the first sand ratio trend threshold. Second sand ratio trend threshold and the third sand ratio trend threshold During the stable period W sp The actual pumping pressure is obtained using the least squares linear regression slope formula. Pressure trend slope characteristics For the actual pumping pressure Pressure trend slope characteristics The standardized slope is obtained by dimensionless processing. ,; will stabilize W sp Segmentation U The pressure trend slope characteristics of the actual pumping pressure are obtained by dividing the time slices into equal-length time slices. ; When the sand ratio trend slope Less than the third sand ratio trend threshold When the sand ratio trend slope is 0%, it indicates that the dynamic characteristics of the sand ratio are decreasing, and no fracturing curve classification is performed at this time; when the sand ratio trend slope is 0%, it indicates that the dynamic characteristics of the sand ratio are decreasing, and no fracturing curve classification is performed at this time. Greater than the first sand ratio trend threshold Or the slope of the sand ratio trend The absolute value does not exceed the second sand ratio trend threshold. At that time, further judgment is made on the pressure trend slope characteristics and pressure trend slope tolerance of the actual pumping pressure. The relationship between the output fracturing curves determines the category of the output curve.
[0016] The intelligent classification method for fracturing curves provided by this invention has the following advantages compared with the prior art: 1. This invention obtains independent measurement values from multiple pressure sensors, introduces a reliability weight model, and evaluates the quality of the sensors in real time to reduce the impact of failures; it clarifies and quantifies the uncertainties in the data correction process, providing confidence assessment for subsequent analysis; then, it further performs data correction under physical constraints, balances historical data with current data through an adaptive weight mechanism, and effectively removes noise while maintaining signal characteristics through total variational regularization.
[0017] 2. The true pumping pressure, casing pressure, and tubing pressure are obtained by inversion decomposition from the corrected comprehensive pressure curve. Constraints such as pressure balance equations are used to ensure that the inversion results meet the physical constraints. During the inversion process, the true local pressure drop of the perforation is obtained, providing key parameters for subsequent displacement estimation.
[0018] 3. After obtaining the displacement estimate, combine it with the calculation formula for the local pressure drop of the perforation to obtain the estimated value of the local pressure drop of the perforation. By comparing the estimated value of the local pressure drop of the perforation with the actual local pressure drop of the perforation obtained in the previous step, verify whether the displacement estimate is reasonable and whether the formula parameters need to be optimized, thereby improving the reliability of the displacement estimate.
[0019] 4. After obtaining the displacement, sand ratio, and actual pumping pressure, the stability of the displacement is evaluated through three dimensions: coefficient of variation, trend stability, and autocorrelation. Only when the sand ratio trend slope is confirmed to be stable or in the above-mentioned stage can the slope change of the actual pumping pressure be analyzed, thereby achieving accurate fracturing curve classification. Attached Figure Description
[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0021] Figure 1 This is a flowchart of an intelligent classification method for fracturing curves according to the present invention; Figure 2 This is a schematic diagram of a curve category in the intelligent classification method for fracturing curves of the present invention; Figure 3 This is a schematic diagram of another curve category in the intelligent classification method for fracturing curves of the present invention; Figure 4 This is a schematic diagram of another curve category in the intelligent classification method for fracturing curves of the present invention; Figure 5 This is a schematic diagram illustrating another value category of the intelligent classification method for fracturing curves according to the present invention. Detailed Implementation
[0022] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0023] like Figure 1 As shown, this invention provides an intelligent classification method for fracturing curves, comprising the following steps: S1: Configure multiple sensors for pressure operation, fuse and quantify the measurement values obtained by each sensor, perform data correction based on variational physical constraints, and output the corrected pressure curve.
[0024] The measured values acquired by each sensor are fused and their uncertainties are quantified. Specifically, this includes acquiring the raw measured values of pump injection pressure from the pump injection pressure sensor, casing pressure from the casing pressure sensor, and tubing pressure from the tubing pressure sensor, and letting the raw measured values of the sensors be... , subscript p , c , t These represent the pump injection pressure, casing pressure, and tubing pressure, respectively. (The text in parentheses is incomplete and cannot be translated.) t Represents a time variable. , T For time variables t The upper limit of the value, i and j Here is the index of the sensor, with a value range of [1, 2, 3]; the variance of the sensor's raw measurements is... Establish a reliability weight model for the sensor: , For reliability function, , The predicted value of the pressure measured by the sensor. The integral interval represents the deviation between the original measured value and the predicted value. Indicates in t A certain time window before the moment. This is the reliability attenuation coefficient. The communication quality factor is represented by the fusion pressure estimate. , integrating uncertainty , The correlation coefficient between the sensors; Uncertainty propagation provides the expected value and variance of the output when the fused stress estimate is input into the downstream stress correlation function, approximated using Taylor expansion, and confidence intervals are constructed. Specific content includes: Uncertainty Propagation: , The output of the downstream model is a random variable. For the expectation operator, the second term on the right-hand side of the formula is a correction term that explains the systematic shift in the output mean caused by the nonlinearity of the downstream model and fusion uncertainty. The output variance is approximated as: Given a confidence level , Construct confidence intervals for significance levels. , These are the quantiles of the standard normal distribution.
[0025] Uncertainty propagation is used to estimate the probability distribution characteristics of the input fusion pressure. Based on Taylor expansion, the expectation and variance of the downstream function output are analytically approximated. Through forward estimation, a corrected expectation is obtained. Output variance and confidence interval to assess the extent to which fusion stress will ultimately worsen when there is fusion uncertainty.
[0026] Step S1, which involves data correction based on variational physical constraints and outputting a corrected pressure curve, specifically includes defining a measurement pressure function. Predicting pressure function Correction pressure curve Measure pressure function The fusion pressure estimation representation derived from multiple sensors; the adaptive weighting function is... ,in To measure the error scale parameter, For time decay weight, The time decay constant; the total variation regularization term , Here is the regularization strength parameter. For smoothing parameters, These are the gradient coefficients. For gradient weights, For time gradient, For Huber-type regularization parameter functions, x The independent variable is ; the physical constraint term is , These are the physical constraint weight parameters. This represents the sequence number of the physical constraint weights. For the first l A physical constraint operator; an objective function is established based on the weighted sum of an adaptive weight function, a total variational regularization term, and physical constraint terms. J | P corr ( tThe corrected pressure curve is obtained by solving the objective function using the gradient descent method. .
[0027] , where λ2 is the physical constraint strength parameter.
[0028] By objective function J | P corr ( t )| Convert to Given a scalar function, use gradient descent to calculate the objective function with respect to... gradient▽ J Then iterate and update until... Convergence occurs, achieving the required accuracy. Gradient descent is a commonly used technique in this field and will not be elaborated upon here.
[0029] S2: Perform inversion decomposition on the corrected pressure curve, define the inversion optimization model and inversion constraints, and solve to obtain the decomposed corrected pressure measurement value; based on the corrected pressure measurement value, further obtain the estimated sand ratio and displacement value for pressure operation, and verify the displacement estimate value.
[0030] Step S2 involves inverting and decomposing the corrected pressure curve, defining an inversion optimization model and inversion constraints, and solving for the decomposed corrected measured values. Specifically, this includes setting the actual pumping pressure, casing pressure, and tubing pressure to be determined as follows: Establish an inversion optimization model for , The regularization coefficient is . For the regularization function, set the following constraints: , , , This refers to the frictional pressure drop within the oil pipe. For gravity pressure drop, , For elevation changes, The sine value of the well inclination angle; For the actual local pressure drop of the perforation, The pressure difference between the tubing and the annulus. For fusion function, For the parameters of the fusion function, The composite pressure is calculated by re-integrating the actual pumping pressure, casing pressure, and tubing pressure obtained through inversion. , The allowable error tolerance is defined; the frictional pressure drop within the oil pipe at this point is obtained by combining the aforementioned constraints. Local pressure drop in a real perforation .
[0031] Based on the corrected measurements, the displacement estimate for the pressure operation is further obtained, which is to obtain the frictional pressure drop in the tubing at this time. Then, the frictional pressure drop inside the oil pipe was calculated using the Darcy-Weisbach equation. The formula, ,in f D The Darcy friction coefficient is based on time-varying parameters. L 0 represents the pipe segment length. D The diameter of the pipe section is... ρ For fluid density, v For fluid velocity, , A For cross-sectional area, Q ( t The displacement estimate is obtained by rewriting the Darcy-Weisbach equation. Q ( t The expression for ) is: , This is a sign function based on the value of frictional pressure drop. The expression for the time-varying Darcy friction coefficient is as follows: , τ For integration variables, b The attenuation coefficient is... c The growth coefficient, Estimated displacement Q ( t Regarding the instantaneous rate of change of the integral variable, Indicates in t The fluid density within the pipe section where friction is constantly being calculated. The initial friction coefficient, This represents the final asymptotic value of the friction coefficient; the displacement estimate is then calculated. Q ( t ).
[0032] The process of verifying the displacement estimate is to obtain the displacement estimate. Q ( t Then, combining the estimated value of the local pressure drop of the perforation with the calculation formula, the estimated value of the local pressure drop of the perforation is obtained. Estimate the local pressure drop of the perforation. Is it due to local pressure drop in a real perforation? Within the deviation range, the estimated value of the local pressure drop of the perforation is calculated using the following formula. ,in The cross-sectional area of a single aperture. The number of holes that can be effectively opened. The orifice flow coefficient; if the estimated value of the local pressure drop in the perforation... Local pressure drop not exceeding the actual perforation depth If the displacement estimate is ±15%, then the displacement estimate is considered to be... Q ( t Reasonable; if the estimated local pressure drop of the perforation is... Local pressure drop exceeding that of a true perforation If the value is ±15%, then perform any of the following operations until the estimated local pressure drop of the perforation is met. Local pressure drop not exceeding the actual perforation depth Up to ±15%: 1) Keep the calculation formula for the estimated local pressure drop of the perforation unchanged, and adjust the number of effectively opened orifices and the orifice flow coefficient; 2) Keep the calculation formula for the estimated local pressure drop of the perforation unchanged, and adjust the attenuation coefficient, growth coefficient and final asymptotic value of the friction coefficient in the expression based on the time-varying Darcy friction coefficient; 3) Replace the current calculation formula for the estimated local pressure drop of the perforation with a polynomial fitting method, such as a quadratic function or a cubic function.
[0033] A more reliable displacement estimate can be provided by back-calculating the displacement value from the local pressure drop of the perforation. Q ( t This provides more accurate data, which can improve the reliability of subsequent classification.
[0034] Obtaining further estimates of the sand ratio is essential for defining the pressure fluctuation index. , for The wellhead pumping pressure at any given time, The length of the sliding time window. In the time window The average value of the wellhead pumping pressure is used to calculate the degree to which the wellhead pumping pressure deviates from the average value within a time window; fluctuation frequency characteristics. , The Fourier transform operator is used to transform time-domain signals. Convert to frequency domain signal The function is used to obtain the frequency at which the amplitude reaches its maximum in the frequency domain. This is used to extract the dominant frequency of pressure fluctuations within the current time window; a linear model for sand ratio estimation is constructed. SR(t) : , , , , These are, respectively, the baseline sand ratio, the sensitivity coefficient of wave capacity to sand ratio, the weighting coefficient of dominant frequency to sand ratio, and the correction coefficient of wave capacity change rate to sand ratio. , , , To obtain the estimated sand ratio, which is a dimensionless sand volume concentration ratio.
[0035] S3: Based on the obtained decomposed and corrected pressure measurements, the verified displacement estimate, and the sand ratio estimate, identify the period after the displacement estimate reaches a stable state. During this stable period, analyze the pressure ratio characteristics and sand ratio dynamic characteristics of the decomposed and corrected pressure measurements to classify the fracturing curves. Specifically, this includes the following steps:
[0036] S31: Define time period W Set the conditions for determining displacement stability.
[0037] Displacement stability requires the following conditions to be met simultaneously: 1) Obtaining displacement estimates Q ( t The corresponding time-varying sequence [ Q ( t ], given time period W Defined by time t The sliding window centered on the given time period W Inside, all displacement estimates within the sliding window. Q ( t The mean and standard deviation of ) are and Define the stability of the coefficient of variation , The mutation threshold is set to 0.05. 2) To stabilize the trend ,in During the time period W Internally, through the time-varying sequence [ Q ( t The trend slope obtained by performing linear regression. , The trend slope threshold. , Time period W All displacement estimates within Q ( t The average value of ) 3) Let the sequence change over time [ Q ( t The autocorrelation function of ] is , representing a sequence [ Q ( t )] and the sequence in delay The following sequence The correlation, For maximum delay, It is a time-varying sequence. Q ( tDuring the time period W within variance, Autocorrelation threshold, autocorrelation stability , ; 4) Define the stable period W sp To simultaneously satisfy the inequalities of stability of the coefficient of variation, the trend stability, and the autocorrelation stability, let the steady period be... W sp Maximum duration With minimum duration The difference is the stable duration. ,and , For the minimum stable duration, .
[0038] Confirming the stability of the discharge rate is essential for identifying the type of fracturing curve based on subsequent trends in pump injection pressure and sand ratio. The unit of discharge rate is cubic meters per second.
[0039] S32: Set feature thresholds and calculate features, and make classification decisions based on the relationship between features and feature thresholds.
[0040] Based on the estimated sand ratio SR ( t During the stable period W sp Within the sequence, the sand ratio estimate is obtained by using the trend slope obtained through linear regression. SR ( t The slope of the sand ratio trend corresponding to the sequence of ) Define the first sand ratio trend threshold. Second sand ratio trend threshold and the third sand ratio trend threshold The values are respectively , , .
[0041] During the stable period W sp The actual pumping pressure is obtained using the least squares linear regression slope formula. Pressure trend slope characteristics For the actual pumping pressure Pressure trend slope characteristics The standardized slope is obtained by dimensionless processing. , It is the actual pumping pressure. During the stable period W sp The maximum and minimum values within the period; the stable periodW sp Segmentation U The pressure trend slope characteristics of the actual pumping pressure are obtained by dividing the time slices into equal-length time slices. , Define the sequence number of the time slice; define a pressure trend slope tolerance. , Indicates the period of stability W sp The actual pumping pressure inside The average value; When the sand ratio trend slope When the sand ratio trend slope is 0%, it indicates that the dynamic characteristics of the sand ratio are decreasing, and no fracturing curve classification is performed at this time; when the sand ratio trend slope is 0%, it indicates that the dynamic characteristics of the sand ratio are decreasing, and no fracturing curve classification is performed at this time. or At that time, further judgment is made on the pressure trend slope characteristics and pressure trend slope tolerance of the actual pumping pressure. The relationship between the output fracturing curves determines the category of the output curve.
[0042] Reference Appendix Figure 2 , Figure 3 , Figure 4 and Figure 5 In step S31, the corresponding type in the attached diagram is confirmed. For example... Figure 2 As shown, after the blue displacement stability meets the requirements, further [further action is needed] during the stable period. W sp Internal judgment of sand ratio trend slope After confirming that the sand ratio is located in the red platform or step-like rising area in the graph until it reaches zero, the sand ratio trend slope is... Greater than the third sand ratio trend threshold The non-decreasing trend was then used to further determine the actual pumping pressure. Pressure trend slope characteristics If it's a continuous downward trend, then it corresponds to... Figure 2 The type of fracturing curve that descends; if the pressure trend slope characteristics It shows a downward trend, but gradually levels off until it stabilizes, which corresponds to... Figure 3 The type of fracturing curve that indicates a decrease in stability; if the pressure trend slope characteristics If it remains unchanged continuously, then it corresponds to Figure 4 The stability of the fracturing curve type; if the pressure trend slope characteristics An upward trend corresponds to Figure 5 The type of rising fracturing curve, Figures 2 to 5 In the middle, the slope characteristics of the pressure trend The unit is psi / s.
[0043] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for intelligent classification of fracturing curves, characterized in that, Includes the following steps: S1: Configure multiple sensors for pressure operation, fuse and quantify the measurement values obtained by each sensor, perform data correction based on variational physical constraints, and output the corrected pressure curve; S2: Perform inversion decomposition on the corrected pressure curve, define the inversion optimization model and inversion constraints, and solve to obtain the decomposed corrected pressure measurement value; based on the corrected pressure measurement value, further obtain the estimated sand ratio and displacement value for pressure operation, and verify the displacement estimate value; S3: Based on the obtained corrected pressure measurement value after decomposition, the displacement estimate value and the sand ratio estimate value after verification, identify the displacement estimate value after it reaches the stable period, analyze the pressure ratio characteristics and sand ratio dynamic characteristics of the corrected pressure measurement value after decomposition during the stable period, and classify the fracturing curve.
2. The intelligent classification method for fracturing curves according to claim 1, characterized in that, The fusion and uncertainty quantification of the measured values acquired by each sensor in step S1 specifically includes acquiring the original measured values of pump injection pressure, casing pressure, and tubing pressure from the pump injection pressure sensor, the original measured values of casing pressure, and the original measured values of tubing pressure from the tubing pressure sensor, and setting the original measured values of the sensors as follows: , subscript p , c , t These represent the pump injection pressure, casing pressure, and tubing pressure, respectively. (The text in parentheses is incomplete and cannot be translated.) t Represents a time variable. , T For time variables t The upper limit of the value, i This is the index of the sensor, with a value range of [1, 2, 3]. Based on the raw measurements of each sensor, the variance of the raw measurements, and the predicted pressure measured by the sensor, a reliability weighting model for the sensor is established. The raw sensor measurements are weighted by reliability and then summed to obtain the fused pressure estimate. Combined with fusion pressure estimation Sensor reliability weight model The standard deviation of the original measurements and the correlation coefficient between the sensors are used to obtain the fusion uncertainty. The standard deviation of the original measurements is the square root of the variance of the original measurements.
3. The intelligent classification method for fracturing curves according to claim 2, characterized in that, Step S1, which involves data correction based on variational physical constraints and outputting a corrected pressure curve, specifically includes defining a measurement pressure function. Predicting pressure function Correction pressure curve Measure pressure function The pressure estimation is derived from multi-sensor fusion; an adaptive weighting function, a total variation regularization term, and a physical constraint term are defined; an objective function is established based on the adaptive weighting function, the total variation regularization term, and the physical constraint term; and the corrected pressure curve is obtained through gradient descent. .
4. The intelligent classification method for fracturing curves according to claim 3, characterized in that, Step S2 involves inverting and decomposing the corrected pressure curve, defining an inversion optimization model and inversion constraints, and solving for the decomposed corrected measured values. Specifically, this includes setting the actual pumping pressure, casing pressure, and tubing pressure to be determined as follows: An inversion optimization model was established with constraints set to obtain the frictional pressure drop inside the tubing at this point. Local pressure drop in a real perforation .
5. The intelligent classification method for fracturing curves according to claim 4, characterized in that, The step S2, which involves obtaining a displacement estimate for pressure operation based on the calibrated measurement value, involves obtaining the frictional pressure drop within the tubing at this point. Then, combining the Darcy-Weisbach equation with the formula for calculating the frictional pressure drop in the oil pipe, the Darcy-Weisbach equation was rewritten to obtain the displacement estimate. Q ( t The calculation formula for displacement is used to obtain the estimated displacement value. Q ( t ).
6. The intelligent classification method for fracturing curves according to claim 5, characterized in that, Verifying the displacement estimate yields the displacement estimate. Q ( t After that, based on the displacement estimate Q ( t The estimated value of the local pressure drop of the perforation is obtained by using the formula for calculating the local pressure drop of the perforation. Estimate the local pressure drop of the perforation. Is it due to local pressure drop in a real perforation? Within the deviation range: if the estimated value of the local pressure drop of the perforation is... Local pressure drop not exceeding the actual perforation depth If the displacement estimate is ±15%, then the displacement estimate is considered to be... Q ( t Reasonable; if the estimated local pressure drop of the perforation is... Local pressure drop exceeding that of a true perforation If the value is ±15%, then perform any of the following operations until the estimated local pressure drop of the perforation is met. Local pressure drop not exceeding the actual perforation depth Up to ±15%: 1) Keep the calculation formula for the estimated value of the local pressure drop of the perforation unchanged, and adjust the estimated value of the local pressure drop of the perforation. The coefficients in the formula; 2) Keep the calculation formula for the estimated local pressure drop of the perforation unchanged, and adjust the estimated displacement value. Q ( t ) The coefficients in the expression; 3) Use polynomial fitting to replace the calculation formula for the current estimate of the local pressure drop of the perforation.
7. The intelligent classification method for fracturing curves according to claim 5, characterized in that, The step S2 described in obtaining the estimated sand ratio involves defining pressure fluctuation indicators and fluctuation frequency characteristics; and constructing a linear model for sand ratio estimation based on these indicators. SR(t) Calculate the estimated value of the sand ratio.
8. The intelligent classification method for fracturing curves according to claim 7, characterized in that, Step S3 specifically includes: S31: Define time period W Set the conditions for determining displacement stability; S32: Set feature thresholds and calculate features, and make classification decisions based on the relationship between features and feature thresholds.
9. The intelligent classification method for fracturing curves according to claim 8, characterized in that, Step S31 specifically includes: Displacement stability needs to simultaneously meet the following conditions: 1) Obtain the displacement estimate. Q ( t The corresponding time-varying sequence [ Q ( t ], given time period W Defined by time t The sliding window centered on the given time period W Inside, all displacement estimates within the sliding window. Q ( t The mean and standard deviation of ) are and The stability of the coefficient of variation is defined as follows: Divide by And not exceeding the mutation threshold ; 2) To stabilize the trend ,in During the time period W Internally, through the time-varying sequence [ Q ( t The trend slope obtained by performing linear regression; 3) Let the sequence change over time [ Q ( t The autocorrelation function of ] is , representing a sequence [ Q ( t )] and the sequence in delay The following sequence The correlation, For maximum delay, It is a time-varying sequence. Q ( t During the time period W variance within, autocorrelation function Divide by variance Autocorrelation stability is obtained; 4) Define the stable period W sp To ensure that the stability of the coefficient of variation, trend stability, and autocorrelation stability all do not exceed a set threshold within a certain time period, let the stability period be... W sp Maximum duration With minimum duration The difference is the stable duration. .
10. The intelligent classification method for fracturing curves according to claim 9, characterized in that, The specific content of step S32 includes: Based on the estimated sand ratio SR ( t During the stable period W sp Within the sequence, the sand ratio estimate is obtained by using the trend slope obtained through linear regression. SR ( t The slope of the sand ratio trend corresponding to the sequence. Define the first sand ratio trend threshold. Second sand ratio trend threshold and the third sand ratio trend threshold During the stable period W sp The actual pumping pressure is obtained using the least squares linear regression slope formula. Pressure trend slope characteristics For the actual pumping pressure Pressure trend slope characteristics The standardized slope is obtained by dimensionless processing. ; will stabilize W sp Segmentation U The pressure trend slope characteristics of the actual pumping pressure are obtained by dividing the time slices into equal-length time slices. ; When the sand ratio trend slope Less than the third sand ratio trend threshold When the sand ratio trend slope is [value missing], it indicates that the dynamic characteristics of the sand ratio are decreasing, and fracturing curve classification is not performed at this time; when the sand ratio trend slope is [value missing], it indicates that the dynamic characteristics of the sand ratio are decreasing, and fracturing curve classification is not performed at this time. Greater than the first sand ratio trend threshold Or the slope of the sand ratio trend The absolute value does not exceed the second sand ratio trend threshold. At that time, further judgment is made on the pressure trend slope characteristics and pressure trend slope tolerance of the actual pumping pressure. The relationship between the output fracturing curves determines the category of the output curve.