Fracturing sand plugging risk early warning method based on deep autoregression model and reverse slope method

Abstract

The invention discloses a fracturing sand plugging risk early warning method based on a deep autoregression model and an inverse slope method. The method comprises the steps that time series data of oil pressure, casing pressure, sand ratio, proppant concentration and discharge displacement in fracturing construction are collected respectively, then normalization processing is carried out, and a training set and a verification set are divided; inputting the training set into a deep autoregression model for training, and then verifying by using the verification set to obtain predicted time sequence data of oil pressure and casing pressure; and based on the predicted time sequence data of the oil pressure and the casing pressure, judging whether sand blockage occurs or not by using an inverse slope method. The method has the advantages that oil pressure data and casing pressure data which are most sensitive to sand plugging are analyzed and calculated, multivariate time series prediction is carried out through the deep autoregression model, the numerical values of the oil pressure and the casing pressure are predicted respectively, and early warning is carried out on sand plugging through the reverse slope method. By combining the neural network and the reverse slope method, the early warning effect on sand blockage can be greatly improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of oilfield fracturing sand plugging risk early warning, and in particular relates to a fracturing sand plugging risk early warning method based on a deep autoregressive model and an inverse slope method. Background Art

[0002] The actual recovery rate of oil and gas in oil fields depends on effective fracturing measures. However, due to the complex geological and stratigraphic environment, defects and failures of supporting equipment and facilities, human operation and management errors during the exploitation process, fracturing construction is often accompanied by various abnormal working conditions. During construction, due to the rupture of the formation and the cracks generated, some formation mineral particles may enter the cracks. As the fracturing fluid is reduced or stopped, these mineral particles may be deposited inside the cracks to form sand plugs. Sand plugging will seriously affect the fracturing effect and even cause major safety problems. Therefore, if the sand plugging during fracturing construction can be warned, it is of great significance to prompt on-site technicians to take effective countermeasures in time to avoid sand plugging.

[0003] Current fracturing operations have the following problems: (1) Research focuses on optimizing intelligent fracturing parameters to increase production and discharge, but there is relatively little research on early warning; (2) There is little research on feature engineering in the prediction of fracturing construction pressure, and there is a lack of utilization of multiple construction parameters related to production; (3) The net pressure prediction method for sand plugging mainly focuses on predicting the trend of downhole pressure changes, but due to the inability to obtain accurate parameters such as rock closure pressure and fracture entrance friction, there is a problem of large errors in net pressure calculation, which limits the actual application of the net pressure prediction method on site.

[0004] At present, the research on sand plugging during fracturing construction is to analyze and summarize the causes of sand plugging from the perspectives of geology, construction, personnel and equipment, but most of them are remedial measures after the accident, and the site still relies on manual judgment. The research on fracturing construction parameters considers more economic goals and increasing production. The correlation between pressure parameters and other parameters has not been studied, and the possibility of fracturing construction accidents is still high.

[0005] The mainstream sand plugging judgment methods currently available include the net pressure fitting method, the double logarithmic curve method and the inverse slope method. The net pressure fitting method is to fit the bottom hole fracture pressure monitored during hydraulic fracturing construction with the fracture pressure simulated and calculated by the three-dimensional fracturing software, and to judge sand plugging by analyzing the net pressure change and finding signs of sand shedding. The construction pressure double logarithmic curve analysis method can also predict sand plugging, and the main method is to observe the double logarithmic coordinate curve of construction pressure and time. The slope reversal method is an empirical judgment method proposed on the basis of summarizing the fracturing construction curve.

[0006] The main warning models currently used include parameter monitoring warning models, machine learning warning models, and risk warning models. Among them, the parameter monitoring model mainly uses real-time on-site monitoring to observe whether the parameters exceed the preset thresholds to warn of sand plugging. However, its threshold is greatly affected by various factors, has weak universality, is prone to false alarms and missed alarms, and is not forward-looking. The machine learning warning model is forward-looking, can warn in advance, and does not rely on manual experience. However, it requires a large amount of monitoring data for training. The risk warning model analyzes accident risks from the perspective of uncertainty, but the actual application effect of this model is not good. In summary, the fracturing sand plugging warning system still needs further research. Summary of the invention

[0007] In order to solve the above problems, the object of the present invention is to provide a method for early warning of fracturing sand plugging risk based on a deep autoregressive model and an inverse slope method.

[0008] In order to achieve the above object, the method for early warning of fracturing sand plugging risk based on a deep autoregressive model and an inverse slope method provided by the present invention comprises the following steps performed in sequence:

[0009] 1) Collect the time series data of parameters such as oil pressure, casing pressure, sand ratio, proppant concentration, and discharge volume during fracturing construction, and then normalize them and divide them into training set and validation set;

[0010] 2) Input the above training set into the deep autoregressive model for training, and then use the validation set for validation to obtain the predicted time series data of oil pressure and casing pressure;

[0011] 3) Based on the predicted time series data of the oil pressure and casing pressure, the inverse slope method is used to determine whether sand plugging occurs.

[0012] In step 1), the method of respectively collecting time series data of parameters such as oil pressure, casing pressure, sand ratio, proppant concentration, and discharge displacement during fracturing construction, and then normalizing and dividing them into a training set and a validation set is:

[0013] Collect the real-time original oil pressure and casing pressure data during fracturing construction and convert them into time series data for analysis; collect the original sand ratio, proppant concentration, and discharge displacement data and convert them into time series data;

[0014] The above time series data are normalized to solve the problem of different value ranges. The normalization formula is:

[0015]

[0016] where Y i represents the normalized data of the i-th parameter, X iRepresents the original data of the i-th parameter, X imax and X imin Respectively represent the maximum and minimum values ​​of the i-th parameter;

[0017] The first 80% of the normalized time series data were used as the training set, and the last 20% as the validation set; the oil pressure and casing pressure were used as the target variables whose future changes needed to be predicted, and the sand ratio, proppant concentration, and discharge displacement were used as covariates that helped predict the target variables.

[0018] In step 2), the method of inputting the training set into the deep autoregressive model for training and then using the validation set for validation to obtain the predicted time series data of oil pressure and casing pressure is:

[0019] The deep autoregressive model is built with the long short-term memory network as the core;

[0020] All the time series data in the above training set are input into the deep autoregressive model for training. The trained deep autoregressive model takes the historical data of the target variable and the historical data of the covariate and the data within the next minute as input to obtain the predicted time series data of the target variable within the next minute. The trained deep autoregressive model is verified with the above validation set, and the hyperparameter range and change step size are set through the grid search method. Continuous attempts are made to find suitable model parameters to optimize the model performance and ultimately obtain higher prediction accuracy.

[0021] In step 3), the method of judging whether sand plugging occurs by using the inverse slope method based on the predicted time series data of the oil pressure and casing pressure is:

[0022] The slope calculated from the oil pressure and casing pressure data of the historical stable sand adding stage before the start of deep autoregressive model prediction is used as the alarm threshold. The slopes of each point of the oil pressure and casing pressure in the predicted time series data of the above oil pressure and casing pressure are calculated. When the calculated absolute value of the slope exceeds the alarm threshold, it indicates that sand blockage is about to occur, reminding the construction personnel to immediately take measures such as reducing the displacement and reducing the sand ratio to avoid sand blockage.

[0023] The method for early warning of sand plugging risk in fracturing based on deep autoregressive model and inverse slope method provided by the present invention has the following beneficial effects: analyzing and calculating the oil pressure and casing pressure data that are most sensitive to sand plugging, using the deep autoregressive model to perform multivariate time series prediction, respectively predicting the values ​​of oil pressure and casing pressure, and using the inverse slope method to provide early warning of sand plugging. The present invention can greatly improve the early warning effect of sand plugging by combining neural network and inverse slope method, and to a certain extent solve the problems of lag, accidental and mechanical problems existing in the existing sand plugging identification by manual observation of curves and manual observation. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A flow chart of the method for early warning of fracturing sand plugging risk based on a deep autoregressive model and an inverse slope method provided by the present invention. DETAILED DESCRIPTION

[0025] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0026] like Figure 1 As shown, the fracturing sand plugging risk early warning method based on the deep autoregressive model and the inverse slope method provided by the present invention includes the following steps performed in sequence:

[0027] 1) Collect the time series data of parameters such as oil pressure, casing pressure, sand ratio, proppant concentration, and discharge volume during fracturing construction, and then normalize them and divide them into training set and validation set;

[0028] By analyzing the changes in sand plugging parameters, it is found that the most obvious manifestation of sand plugging is the increase in tubing pressure and casing pressure. In order to facilitate research and analysis, it is necessary to collect real-time original oil pressure and casing pressure data during fracturing construction and convert them into time series data for analysis. Time series data is a data sequence arranged in chronological order, in which each data point is associated with a specific time point or time period. In addition, the three parameters of sand ratio, proppant concentration, and discharge displacement also have obvious reactions to sand plugging, so it is also necessary to collect the original sand ratio, proppant concentration, and discharge displacement data and convert them into time series data;

[0029] The above time series data are normalized to solve the problem of different value ranges. The normalization formula is:

[0030]

[0031] where Y i represents the normalized data of the i-th parameter, X i Represents the original data of the i-th parameter, X imax and X imin Respectively represent the maximum and minimum values ​​of the i-th parameter;

[0032] The first 80% of each normalized time series data is used as the training set, and the last 20% is used as the validation set. The oil pressure and casing pressure are used as the target variables whose future changes need to be predicted, and the sand ratio, proppant concentration, and discharge displacement are used as covariates that help predict the target variables.

[0033] 2) Input the above training set into the deep autoregressive model for training, and then use the validation set for validation to obtain the predicted time series data of oil pressure and casing pressure;

[0034] Deep Autoregressive Model (DeepAR) is a deep learning model for time series prediction proposed by Amazon. The main feature of the deep autoregressive model is that it can adaptively learn the seasonal and trend characteristics in time series data, thereby achieving more accurate predictions. It captures the complex relationship between data by introducing historical data of the target variable and historical data of related covariates in the input, thereby predicting the future value of the target variable. The deep autoregressive model used in the present invention is built with the long short-term memory network (LSTM) as the core.

[0035] All the time series data in the above training set are input into the deep autoregressive model for training. The trained deep autoregressive model can use the historical data of the target variable and the historical data of the covariate and the data within the next minute as input to obtain the predicted time series data of the target variable within the next minute. Since the covariates are all artificially controlled quantities in fracturing construction, their future values ​​can also be set in advance in actual work. By predicting the future target variables, sand plugging can be warned in advance by analyzing the future changes in oil pressure and casing pressure. In order to ensure the accuracy of the prediction, the trained deep autoregressive model is verified with the above validation set. The hyperparameter range and change step size are set by the grid search method. Continuous attempts are made to find suitable model parameters to optimize the model performance and ultimately obtain a higher prediction accuracy.

[0036] 3) Based on the above predicted time series data of oil pressure and casing pressure, the inverse slope method is used to determine whether sand plugging occurs;

[0037] The reverse slope method is an empirical method that can be used to predict sand plugging in real time. The principle is: after the pre-fluid is pumped into the wellbore of the oil well, proppant is added, which will cause the liquid column pressure in the wellbore to gradually increase, while the oil pressure and casing pressure gradually decrease at a relatively constant slope. Once the entire wellbore is filled with sand-carrying fluid, the liquid column pressure reaches the maximum, and the oil pressure and casing pressure stop decreasing. This process is reflected in the construction curve as a negative slope, that is, it shows a trend of pressure decreasing over time. As the construction continues, the oil pressure and casing pressure remain unchanged for a short time, or begin to increase gradually. When the oil pressure and casing pressure begin to gradually rise, the slope after the reversal is almost the same as the slope when the wellbore was previously loaded, which is reflected in the construction curve as a positive slope. The positive slope begins to deviate from the reverse slope, which is an obvious sign of sand plugging in the near-wellbore area.

[0038] The slope calculated from the oil pressure and casing pressure data of the historical stable sand adding stage before the start of deep autoregressive model prediction is used as the alarm threshold. The slopes of each point of the oil pressure and casing pressure in the predicted time series data of the above oil pressure and casing pressure are calculated. When the calculated absolute value of the slope exceeds the alarm threshold, it indicates that sand blockage is about to occur, reminding the construction personnel to immediately take measures such as reducing the displacement and reducing the sand ratio to avoid sand blockage.

Claims

1. A method for early warning of fracturing sand plugging risk based on a deep autoregressive model and an inverse slope method, characterized in that: The method comprises the following steps performed in sequence: 1) Collect the time series data of parameters such as oil pressure, casing pressure, sand ratio, proppant concentration, and discharge volume during fracturing construction, and then normalize them and divide them into training set and validation set; 2) Input the above training set into the deep autoregressive model for training, and then use the validation set for validation to obtain the predicted time series data of oil pressure and casing pressure; 3) Based on the predicted time series data of the oil pressure and casing pressure, the inverse slope method is used to determine whether sand plugging occurs.

2. The method for early warning of fracturing sand plugging risk based on deep autoregressive model and inverse slope method according to claim 1 is characterized by: In step 1), the method of respectively collecting time series data of parameters such as oil pressure, casing pressure, sand ratio, proppant concentration, and discharge displacement during fracturing construction, and then normalizing and dividing them into a training set and a validation set is: Collect the real-time original oil pressure and casing pressure data during fracturing construction and convert them into time series data for analysis; collect the original sand ratio, proppant concentration, and discharge displacement data and convert them into time series data; The above time series data are normalized to solve the problem of different value ranges. The normalization formula is: where Y i represents the normalized data of the i-th parameter, X i Represents the original data of the i-th parameter, X imax and X imin Respectively represent the maximum and minimum values ​​of the i-th parameter; The first 80% of the normalized time series data were used as the training set, and the last 20% as the validation set; the oil pressure and casing pressure were used as the target variables whose future changes needed to be predicted, and the sand ratio, proppant concentration, and discharge displacement were used as covariates that helped predict the target variables.

3. The method for early warning of fracturing sand plugging risk based on deep autoregressive model and inverse slope method according to claim 1, characterized in that: In step 2), the method of inputting the training set into the deep autoregressive model for training and then using the validation set for validation to obtain the predicted time series data of oil pressure and casing pressure is: The deep autoregressive model is built with the long short-term memory network as the core; All the time series data in the above training set are input into the deep autoregressive model for training. The trained deep autoregressive model takes the historical data of the target variable and the historical data of the covariate and the data within the next minute as input to obtain the predicted time series data of the target variable within the next minute. The trained deep autoregressive model is verified with the above validation set, and the hyperparameter range and change step size are set through the grid search method. Continuous attempts are made to find suitable model parameters to optimize the model performance and ultimately obtain higher prediction accuracy.

4. The method for early warning of fracturing sand plugging risk based on deep autoregressive model and inverse slope method according to claim 1, characterized in that: In step 3), the method of judging whether sand plugging occurs by using the inverse slope method based on the predicted time series data of the oil pressure and casing pressure is: The slope calculated from the oil pressure and casing pressure data of the historical stable sand adding stage before the start of deep autoregressive model prediction is used as the alarm threshold. The slopes of each point of the oil pressure and casing pressure in the predicted time series data of the above oil pressure and casing pressure are calculated. When the calculated absolute value of the slope exceeds the alarm threshold, it indicates that sand blockage is about to occur, reminding the construction personnel to immediately take measures such as reducing the displacement and reducing the sand ratio to avoid sand blockage.

Images