An oil production simulation system based on digital twin technology and a processing method thereof

Abstract

The present application relates to the technical field of oil collection, in particular to an oil collection simulation system based on digital twin technology and a processing method thereof, comprising: a multi-dimensional dynamic data acquisition module, used for real-time acquisition of formation characteristic dynamic data and equipment state feature data, and dynamic adjustment of acquisition frequency according to fluctuation amplitude of the acquisition data; a data preprocessing and change law modeling module, connected with the multi-dimensional dynamic data acquisition module, used for preprocessing of the formation characteristic dynamic data and the equipment state feature data acquired by the multi-dimensional dynamic data acquisition module, and modeling based on fusion of oil exploitation physical mechanism and machine learning. Through the cooperative work of multiple modules, the present application effectively solves the problem that the existing oil collection simulation system only initially imports data, and subsequent twin model parameters are difficult to adjust in real time according to actual data, resulting in increased deviation and inconsistency between simulation results and actual conditions.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of oil collection, in particular to an oil collection simulation system based on digital twin technology and a processing method thereof. BACKGROUND

[0002] As an important energy resource, the collection process of oil involves multiple complex links such as wellhead equipment operation, formation fluid flow, pipeline transportation, etc. Due to the harsh environment and variable working conditions of the oil collection site, directly optimizing the collection scheme through field tests not only has high cost, but also may face safety risks. Therefore, the industry has begun to introduce digital twin technology to build an oil collection simulation system. This technology collects various data in the actual oil collection scene, such as wellhead pressure, fluid flow, equipment temperature, etc., and builds a highly consistent twin model in the virtual space. With the model, the collection process under different working conditions is simulated to provide support for optimizing collection parameters and predicting equipment failures for workers, thereby reducing the cost of field tests and improving the safety and efficiency of the collection process.

[0003] Currently, in the oil collection process, the formation characteristics will gradually change over time, such as the slight change in formation porosity due to fluid production, and the dynamic fluctuations in the operating parameters of wellhead equipment due to factors such as wear and aging. These changes will directly affect the actual state of the collection process. However, most existing simulation systems only import actual data at the initial modeling stage, and subsequent parameter updates to the twin model are difficult to automatically adjust according to the real-time variation of actual collection data, resulting in a gradual increase in the parameter deviation between the twin model and the actual collection scene. This makes the simulation results of the production process output by the simulation system inconsistent with the actual situation, and unable to provide accurate decision-making basis for workers, which may lead to unreasonable adjustment of the collection scheme, affecting the efficiency of oil collection and even causing equipment abnormalities. SUMMARY

[0004] To overcome the deficiencies of the prior art, the present application provides an oil collection simulation system based on digital twin technology and a processing method thereof, which solves the problem of the prior art that the twin model only imports data initially and the subsequent parameters are difficult to adjust in real time with actual data, resulting in an increase in parameter deviation, inconsistency between simulation results and reality, and inability to provide accurate decision-making basis.

[0005] To achieve the above purpose, the present application realizes the following technical scheme: an oil collection simulation system based on digital twin technology, comprising:

[0006] A multi-dimensional dynamic data acquisition module is used to acquire formation characteristic dynamic data and equipment state feature data in real time, and dynamically adjust the acquisition frequency according to the fluctuation amplitude of the acquisition data.

[0007] The data preprocessing and change law modeling module is connected with the multi-dimensional dynamic data acquisition module, is used for preprocessing the formation property dynamic data and the equipment state characteristic data collected by the multi-dimensional dynamic data acquisition module, and is based on the fusion modeling of the physical mechanism of oil exploitation and machine learning to dynamically train the formation property change model and the equipment parameter fluctuation model, so that the future short-term change trend of the formation property dynamic data and the equipment state characteristic data is predicted.

[0008] The twin model parameter adaptive updating module is connected with the data preprocessing and change law modeling module, is used for dividing the priority of the twin model parameters according to the influence weight of the parameters on the acquisition process, and is used for comparing the real-time data collected by the multi-dimensional dynamic data acquisition module with the deviation of the current parameters of the twin model in real time according to the hierarchical incremental updating mechanism of the parameter priority, and when the deviation exceeds the preset accuracy threshold, calculating the parameter correction amount based on the formation property change model and the equipment parameter fluctuation model that have been constructed, and incrementally updating the corresponding formation property parameters and equipment state parameters in the twin model.

[0009] The simulation result feedback optimization module is connected with the twin model parameter adaptive updating module and the multi-dimensional dynamic data acquisition module, is used for constructing a deviation analysis engine, comparing the simulation output of the twin model with the field data collected at the same period, and decomposing and attributing the deviation, and based on the deviation attribution result, reversely optimizing the acquisition frequency of the multi-dimensional dynamic data acquisition module or the model parameters of the data preprocessing and change law modeling module, and providing a parameter sensitivity analysis function.

[0010] Further, the multi-dimensional dynamic data acquisition module comprises:

[0011] The formation sensing unit is used for being deployed in the formation of an oil well, and is used for collecting at least one of formation property dynamic data such as formation porosity, permeability and formation pressure gradient in real time.

[0012] The micro-electromechanical sensing module is used for being deployed on a key component of a wellhead equipment, and is used for collecting at least one of equipment state characteristic data such as equipment vibration frequency, component temperature fluctuation and sealing performance attenuation in real time.

[0013] The adaptive acquisition controller is connected with the formation sensing unit and the micro-electromechanical sensing module, is used for dynamically adjusting the acquisition frequency of the formation sensing unit and the micro-electromechanical sensing module according to the fluctuation amplitude of the formation property dynamic data or the equipment state characteristic data, and when the parameter fluctuation is less than a preset stable threshold, the acquisition frequency is reduced, and when the parameter fluctuation exceeds the preset stable threshold, the acquisition frequency is automatically increased.

[0014] Further, the data preprocessing and variation law modeling module comprises:

[0015] A multi-source data preprocessing unit is configured to remove interference values in the formation property dynamic data and the equipment state feature data by an outlier identification algorithm, and fill in occasional collection gaps by a data completion algorithm;

[0016] A dynamic law modeling unit is connected with the multi-source data preprocessing unit, configured to take the preprocessed formation property dynamic data and equipment state feature data as input, construct a basic law framework based on a physical mechanism of oil extraction, and introduce a time series prediction algorithm to dynamically train the formation property variation model and the equipment parameter fluctuation model.

[0017] Further, the formation property variation model trained by the dynamic law modeling unit is a formation porosity change curve model with respect to the amount of extraction, and the equipment parameter fluctuation model is a vibration frequency decay model with respect to the running time.

[0018] Further, the twin model parameter adaptive updating module divides the twin model parameters into core influence layer parameters and auxiliary influence layer parameters.

[0019] The twin model parameter adaptive updating module comprises:

[0020] A core parameter updating unit is configured to adopt a real-time comparison and instant updating mode, and when a deviation between the collected real-time data and the current core influence layer parameters of the twin model exceeds a preset accuracy threshold, a parameter correction amount is calculated based on the formation property variation model and the equipment parameter fluctuation model, and the corresponding core influence layer parameters in the twin model are incrementally updated.

[0021] An auxiliary parameter updating unit is configured to adopt a periodic verification and batch optimization mode, and a deviation of the collected data is summarized according to a preset period, and the corresponding auxiliary influence layer parameters in the twin model are batch fine-tuned in combination with long-term trend prediction results of the formation property variation model and the equipment parameter fluctuation model.

[0022] Further, the twin model parameter adaptive updating module further comprises:

[0023] An updating effectiveness verification unit is connected with the core parameter updating unit and the auxiliary parameter updating unit, configured to, after each parameter update, compare the simulation output of the updated twin model with the actual collected field data of the same period by virtual actual comparison verification, and if the deviation is less than a set threshold, the update is confirmed to be effective.

[0024] Further, when the update validity verification unit judges that the deviation is out of tolerance, a backtracking mechanism is automatically triggered, and the parameters of the corresponding change law model in the data preprocessing and change law modeling module are recalibrated by the twin model parameter adaptive updating module.

[0025] Further, the simulation result feedback optimization module comprises:

[0026] a deviation analysis engine for decomposing the deviation between the twin model simulation output and the actual collected data, locating the source of the deviation as insufficient collection frequency of the multi-dimensional dynamic data collection module or poor adaptability of the data preprocessing and change law modeling module to specific working conditions, and generating a deviation attribution report;

[0027] and a closed-loop optimization strategy unit connected with the deviation analysis engine, for adjusting the collection frequency of the corresponding parameters in the multi-dimensional dynamic data collection module, or updating the training samples of the machine learning algorithm in the data preprocessing and change law modeling module, or adjusting the correlation coefficient of the physical mechanism model, based on the deviation attribution result.

[0028] Further, the simulation result feedback optimization module further comprises:

[0029] a decision support enhancement unit connected with the closed-loop optimization strategy unit, for simulating the influence of different collection parameters on collection efficiency or equipment life based on the updated twin model, and outputting quantitative optimization suggestions.

[0030] The application also provides a petroleum collection simulation method based on digital twin technology, which is applied to the petroleum collection simulation system based on digital twin technology.

[0031] real-time collection of formation property dynamic data and equipment state characteristic data, and dynamic adjustment of the collection frequency according to the fluctuation amplitude of the collected data;

[0032] preprocessing of the collected formation property dynamic data and equipment state characteristic data, and dynamic training of a formation property change model and an equipment parameter fluctuation model based on fusion modeling of petroleum exploitation physical mechanism and machine learning, so as to realize prediction of the future short-term change trend of the formation property dynamic data and the equipment state characteristic data;

[0033] According to the influence weight of parameters on the acquisition process, the priority of the twin model parameters is divided, and according to the parameter priority, a hierarchical incremental updating mechanism is adopted, the deviation of the collected real-time data from the current parameters of the twin model is compared in real time, when the deviation exceeds the preset accuracy threshold, the parameter correction amount is calculated based on the constructed stratum characteristic change model and the equipment parameter fluctuation model, and the corresponding stratum characteristic parameters and equipment state parameters in the twin model are incrementally updated;

[0034] A deviation analysis engine is constructed to compare the simulation output of the twin model with the actual collected field data at the same period, and the deviation is decomposed and attributed;

[0035] Based on the deviation attribution result, the data acquisition frequency or the model parameters are optimized in reverse, and the parameter sensitivity analysis is provided.

[0036] Compared with the prior art, the beneficial effects of the present application are:

[0037] The present application solves the problem that the existing petroleum collection simulation system only imports initial data, and the subsequent twin model parameters are difficult to adjust in real time according to the actual data, resulting in increased deviation and inconsistent simulation results with actual data. Among them, the multi-dimensional dynamic data acquisition module dynamically adjusts the acquisition frequency according to the data fluctuation amplitude, which ensures accurate data acquisition when the fluctuation is intense, and avoids redundant acquisition when it is stable; the data preprocessing and change rule modeling module eliminates outliers, fills in data gaps, and integrates petroleum production physical mechanism and machine learning modeling to predict short-term data trends; the twin model parameter adaptive updating module updates the parameters in layers according to the parameter influence weight, the core parameters are updated in real time, the auxiliary parameters are periodically fine-tuned in batches, and the model is calibrated through effectiveness verification and backtracking mechanism to ensure that the parameters are consistent with the actual situation; the simulation result feedback optimization module decomposes and attributes the deviation and optimizes the acquisition frequency and model parameters in reverse, provides parameter sensitivity analysis and quantitative decision suggestions, and finally makes the twin model accurately reflect the actual scene, reduces the simulation deviation, provides reliable basis for collection scheme optimization, and improves the safety and efficiency of petroleum collection. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 The system structure diagram of the present application is shown in the figure;

[0039] Figure 2 The method flowchart of the present application is shown in the figure. DETAILED DESCRIPTION

[0040] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0041] Please refer to Figure 1 The present application provides a petroleum production simulation system based on digital twin technology, comprising:

[0042] A multi-dimensional dynamic data acquisition module is configured to acquire real-time formation property dynamic data and equipment state characteristic data, and dynamically adjust the acquisition frequency according to the fluctuation amplitude of the acquired data;

[0043] A data preprocessing and change law modeling module is connected with the multi-dimensional dynamic data acquisition module, configured to preprocess the formation property dynamic data and equipment state characteristic data acquired by the multi-dimensional dynamic data acquisition module, and dynamically train a formation property change model and an equipment parameter fluctuation model based on the fusion modeling of the physical mechanism of oil production and machine learning, so as to predict the future short-term change trend of the formation property dynamic data and equipment state characteristic data;

[0044] A twin model parameter self-adaptive updating module is connected with the data preprocessing and change law modeling module, configured to divide the priority of the twin model parameters according to the influence weight of the parameters on the acquisition process, and use a hierarchical incremental updating mechanism according to the parameter priority to compare the deviation between the real-time data acquired by the multi-dimensional dynamic data acquisition module and the current parameters of the twin model in real time, and when the deviation exceeds a preset accuracy threshold, calculate the parameter correction amount based on the formation property change model and the equipment parameter fluctuation model that have been constructed, and incrementally update the corresponding formation property parameters and equipment state parameters in the twin model;

[0045] A simulation result feedback optimization module is connected with the twin model parameter self-adaptive updating module and the multi-dimensional dynamic data acquisition module, configured to construct a deviation analysis engine, compare the simulation output of the twin model with the field data actually acquired at the same period, and decompose and attribute the deviation, and based on the deviation attribution result, reversely optimize the acquisition frequency of the multi-dimensional dynamic data acquisition module or the model parameters of the data preprocessing and change law modeling module, and provide a parameter sensitivity analysis function.

[0046] Specifically, the present embodiment aims at the problem in the background art that the existing simulation system only initially imports data, and it is difficult to adjust the subsequent parameter update with real-time changes, resulting in increased deviation, and realizes dynamic adaptation and accurate simulation through the cooperation of four modules, as follows:

[0047] When the multi-dimensional dynamic data acquisition module is deployed, first, the sensing devices are arranged at different formation depths of the oil well, such as the middle and bottom of the oil layer, to acquire real-time dynamic data of formation characteristics, such as formation porosity and permeability. At the same time, sensing components are installed on key components such as the crank of the wellhead pumping unit and the wellhead valve to acquire equipment state characteristic data such as equipment vibration frequency and component temperature. The module dynamically adjusts the acquisition frequency through a preset fluctuation judgment formula: let the real-time fluctuation amplitude of a certain parameter be , , where is the parameter value at the current time, is the parameter value at the previous acquisition time; the preset stable threshold is . When , the acquisition frequency is reduced from the initial 1 time / 2 minutes to 1 time / 5 minutes; when , the acquisition frequency is automatically increased to 1 time / 30 seconds, ensuring that data acquisition is more intensive when fluctuations are intense, and providing accurate data basis for subsequent modeling.

[0048] The data preprocessing and change rule modeling module first preprocesses the collected data: 3σ outlier identification algorithm is used to eliminate interference values, and the formula is , where is a single data, is the mean value of the parameter data set, is the standard deviation, and the data meeting this condition is determined as an outlier and is eliminated; for occasional collection gaps, linear interpolation method is used to complete, and the formula is , where is the data of at the missing time , are the data of before and after the adjacent collection time , . After preprocessing, a basic framework is constructed based on the physical mechanism of oil exploitation, such as Darcy's law, and an LSTM time series prediction algorithm is introduced for integrated modeling to train the formation characteristic change model and the equipment parameter fluctuation model. Taking the formation characteristic change model (formation porosity change model) as an example, the formula is , where is the formation porosity at time t, is the initial porosity, is the porosity decay coefficient, with the unit of is the oil production at time τ, with the unit of m³ / h, is the cumulative production from 0 to t, with the unit of m³; in this formula, the product of and the cumulative production has the dimension of %, and The dimension of the quantity is consistent, and normalization is not required. The model can predict the trend of the formation porosity change in the next 12 hours, providing a basis for parameter updating.

[0049] The twin model parameter adaptive updating module first divides the parameter priority: based on the analytic hierarchy process, the influence weight of each parameter on the acquisition process is calculated, for example, the influence weight of formation pressure and wellhead flow on acquisition efficiency is more than 0.6, which is divided into core influence layer parameters; the influence weight of environmental temperature and pipeline humidity is less than 0.3, which is divided into auxiliary influence layer parameters. A hierarchical incremental updating mechanism is adopted: for the core influence layer parameters, the deviation is calculated in real time is the real-time acquisition data, is the current parameter value of the twin model; the preset precision threshold is set as 0.5%. When , the correction amount is calculated based on the formation characteristic change model and the equipment parameter fluctuation model is the reasonable value of the parameter predicted by the model, and the core parameter is updated incrementally in real time. For the auxiliary influence layer parameters, the deviation is summarized every 1 hour, and the long-term trend prediction result of the model is combined, such as the trend of environmental temperature change in the next 24 hours, to perform batch fine-tuning to avoid frequent updating of system resources.

[0050] The simulation result feedback optimization module constructs a deviation analysis engine: compares the production and equipment energy consumption data simulated by the twin model with the actual acquisition data at the same period, and decomposes the deviation through the deviation decomposition formula decomposes the deviation, wherein is the total deviation, is the data acquisition deviation, is the model modeling deviation, so as to locate the source of the deviation. For example, if accounts for more than 60%, it is determined that the acquisition frequency is insufficient; if accounts for more than 70%, it is determined that the model has poor adaptability to high sand content conditions. Based on the attribution result, the reverse optimization is performed: when the acquisition frequency is insufficient, the stable threshold of the corresponding parameter is lowered from 0.5% to 0.3%, triggering higher frequency acquisition; when the model adaptability is poor, training samples under high sand content conditions are supplemented, for example, 200 groups of historical data with sand content exceeding 15% are added. At the same time, the parameter sensitivity analysis function is provided, and the change rate of the simulation result when the parameter changes by 1% is calculated to clearly determine the key optimization direction.

[0051] In this embodiment, through dynamic acquisition, fusion modeling, hierarchical updating and closed-loop optimization, the parameters of the twin model are always consistent with the actual scene, the deviation between the simulation result and the actual situation is greatly reduced, and accurate support is provided for the acquisition scheme optimization.

[0052] In the embodiment, the multi-dimensional dynamic data acquisition module comprises:

[0053] a formation sensing unit for being deployed in the formation of the oil well to acquire real-time dynamic data of at least one formation characteristic such as formation porosity, permeability and formation pressure gradient;

[0054] a micro-electromechanical sensing module for being deployed on key components of the wellhead equipment to acquire real-time equipment state characteristic data such as equipment vibration frequency, component temperature fluctuation and sealing performance attenuation;

[0055] and an adaptive acquisition controller connected with the formation sensing unit and the micro-electromechanical sensing module, for dynamically adjusting the acquisition frequency of the formation sensing unit and the micro-electromechanical sensing module according to the fluctuation amplitude of the formation characteristic dynamic data or the equipment state characteristic data, reducing the acquisition frequency when the parameter fluctuation is less than a preset stable threshold, and automatically increasing the acquisition frequency when the parameter fluctuation exceeds the preset stable threshold.

[0056] Specifically, the formation sensing unit adopts a distributed optical fiber sensor, which is deployed along the outer wall of the casing of the oil well, for example, extending from the wellhead to 10 meters below the bottom of the oil layer, to acquire real-time formation porosity, permeability and formation pressure gradient data; the formation porosity is converted from the formation sound wave propagation speed detected by the sensor, and the permeability is calculated based on the correlation between the porosity and the viscosity of the formation fluid.

[0057] The micro-electromechanical sensing module selects MEMS vibration sensors and temperature sensors, which are respectively installed on key components such as the crank pin of the wellhead pumping unit, the reduction gearbox and the sealing surface of the wellhead valve; for example, a vibration sensor is installed at the crank pin, with a sampling rate of 1000 Hz to acquire real-time equipment vibration frequency; a temperature sensor is installed at the sealing surface, with an accuracy of ±0.1℃ to acquire real-time component temperature fluctuation, and a pressure sensor is used to monitor the sealing performance attenuation, which is reflected by the difference between the leakage pressure of the sealing medium and the initial sealing pressure.

[0058] The adaptive acquisition controller has a built-in data processing chip, which first calculates the acquired formation characteristic dynamic data and equipment state characteristic data in real time to obtain the fluctuation amplitude of each parameter . The preset stable threshold : for example, the formation porosity stable threshold is set to 0.5%, and the equipment vibration frequency stable threshold is set to 2 Hz. When the fluctuation amplitude of a certain parameter , the controller sends an instruction to the corresponding sensor to reduce the acquisition frequency; for example, when the formation porosity fluctuation is 0.3%, the acquisition frequency of the formation sensing unit is reduced from 1 time per minute to 1 time per 5 minutes. When the vibration frequency of the device fluctuates at 3Hz, the controller automatically increases the acquisition frequency of the vibration data from the microelectromechanical sensing module from 1 time / 30 seconds to 1 time / 5 seconds to ensure that critical fluctuation data is not lost, while avoiding redundant data acquisition in a stable state and reducing data transmission and storage costs.

[0059] In this embodiment, the data preprocessing and change pattern modeling module includes:

[0060] The multi-source data preprocessing unit is used to remove interference values ​​from dynamic data of formation characteristics and equipment status characteristic data through outlier identification algorithms, and to fill occasional acquisition gaps through data completion algorithms.

[0061] It also includes a dynamic law modeling unit, which is connected to a multi-source data preprocessing unit. This unit uses preprocessed dynamic data of formation characteristics and equipment status characteristics as input to construct a basic law framework based on the physical mechanism of oil extraction. It also introduces a time-series prediction algorithm to dynamically train the formation characteristic change model and the equipment parameter fluctuation model.

[0062] Specifically, the multi-source data preprocessing unit first identifies outliers: for dynamic formation characteristic data such as formation pressure and equipment status characteristic data such as equipment temperature, a 3σ algorithm is used to remove interfering values; for example, in a set of wellhead temperature data, the temperature value at a certain moment is 85℃, while the average value of the dataset is... Standard deviation Calculated The data was identified as an outlier and removed. For occasional data gaps, such as a 10-minute data loss due to a temporary power outage of the sensor, linear interpolation was used to fill the gap; for example, if there was a 10:10 gap between 10:00 (data at 25Hz) and 10:20 (data at 28Hz) in the equipment vibration frequency range, the calculated... Complete the data completion.

[0063] The dynamic law modeling unit takes preprocessed data as input and first constructs a basic law framework based on the physical mechanism of oil extraction; for example, for formation characteristics, it constructs correlations based on Darcy's law, the formula of which is... ,in For fluid flow rate, For penetration rate, For the flow cross-sectional area, For pressure difference, For fluid viscosity, The flow length is used to determine the physical relationship between formation parameters using this law. Then, the ARIMA time-series prediction algorithm is introduced to dynamically train the formation characteristic variation model and the equipment parameter fluctuation model; for example, when training the equipment parameter fluctuation model, preprocessed equipment vibration frequency data is used as input, and the ARIMA model formula is... wherein is d-th order difference for making data stationary, is a constant term, is an autoregressive coefficient, is a moving average coefficient, is white noise error. Through the model, the device vibration frequency change in the next 6 hours can be predicted, providing a prediction basis for twin model parameter updating, and improving the forward-looking nature of parameter updating.

[0064] In the embodiment, the formation property change model trained by the dynamic law modeling unit is a formation porosity change curve model with the change of the production volume, and the device parameter fluctuation model is a device vibration frequency decay model with the running time.

[0065] Specifically, when training the formation property change model (the formation porosity change curve model with the change of the production volume), the preprocessed formation porosity data and the cumulative production volume data at the corresponding time are taken as samples, and an exponential decay model is used for construction, and the formula is wherein is the formation porosity when the cumulative production volume is Q, and the unit is %, is the formation porosity when the initial cumulative production volume is 0, and the unit is %, is the porosity decay coefficient, and the unit is m⁻³, is the cumulative production volume, and the unit is m³. In the formula, is a dimensionless quantity, and the calculation method is m⁻³×m³=1, which is consistent with the requirement of the exponential function and does not need to be normalized; for example, the initial porosity of a certain oil well is , the decay coefficient is , when the cumulative production volume is , the porosity is , which accurately reflects the decay trend of the porosity with the production volume.

[0066] When training the device parameter fluctuation model (the device vibration frequency decay model with the running time), the preprocessed device vibration frequency data and the device cumulative running time data are taken as samples, and a linear decay model is used for construction, and the formula is wherein is the device vibration frequency when the cumulative running time is T, and the unit is Hz, is the vibration frequency when the initial running time is 0, and the unit is Hz, is the vibration frequency decay coefficient, and the unit is Hz / h, is the device cumulative running time, and the unit is h. In the formula, the dimension of is Hz, and the calculation method is Hz / h×h=Hz, which is consistent with The dimension of the frequency is Hz; for example, the initial vibration frequency of a wellhead valve The decay coefficient When the cumulative running time The vibration frequency Accurately reflects the decay law of the vibration frequency with the running time, providing support for equipment state prediction.

[0067] In the embodiment, the twin model parameter adaptive updating module divides the twin model parameters into core influence layer parameters and auxiliary influence layer parameters.

[0068] The twin model parameter adaptive updating module comprises:

[0069] The core parameter updating unit is configured to adopt a real-time comparison and instant updating mode, calculate a parameter correction amount based on the formation property change model and the equipment parameter fluctuation model when the deviation of the collected real-time data from the current core influence layer parameters of the twin model exceeds a preset accuracy threshold, and perform incremental updating on the corresponding core influence layer parameters in the twin model.

[0070] The auxiliary parameter updating unit is configured to adopt a periodic checking and batch optimization mode, aggregate the deviations of the collected data according to a preset period, combine the long-term trend prediction results of the formation property change model and the equipment parameter fluctuation model, and perform batch fine-tuning on the corresponding auxiliary influence layer parameters in the twin model.

[0071] Specifically, the twin model parameter adaptive updating module first divides the parameter priority by the analytic hierarchy process: five experts in the fields of oil exploitation and digital twin are invited to score the influence weights of parameters such as formation pressure, permeability, equipment temperature, and environmental humidity, weight values are calculated based on the scoring results, parameters with a weight value greater than or equal to 0.5 are divided into core influence layer parameters, such as formation pressure and wellhead flow rate, and parameters with a weight value less than 0.5 are divided into auxiliary influence layer parameters, such as environmental humidity and pipe outer wall temperature.

[0072] The core parameter updating unit adopts a real-time comparison and instant updating mode: real-time collection of actual data of the core influence layer parameters Comparison with the current core parameter value of the twin model Calculation of the deviation The preset accuracy threshold When The parameter correction amount is calculated based on the formation property change model and the equipment parameter fluctuation model , wherein is the reasonable value of the parameter predicted by the model; for example, the actual value of the formation pressure , the current value of the model , the deviation The reasonable pressure value is predicted by the formation property change model , and the correction amount The formation pressure parameters in the twin model are incrementally updated to ensure real-time accuracy of the core parameters.

[0073] The auxiliary parameter updating unit adopts a periodic verification batch optimization mode: the preset update period is 2 hours, and every 2 hours, the actual acquisition data of the auxiliary influence layer parameters are summarized, the average deviation is calculated , wherein is the amount of data collected in the period, is the deviation of a single data. Combined with the long-term trend prediction results of the device parameter fluctuation model, such as the future 12-hour environmental humidity change trend, the auxiliary parameters are batch fine-tuned; for example, the average deviation of environmental humidity , the model predicts that the future humidity will slowly rise, and the environmental humidity parameter in the twin model will be batch adjusted by 0.3%, which ensures the accuracy of the auxiliary parameters and avoids frequent updates from occupying system computing power.

[0074] In the present embodiment, the twin model parameter self-adaptive updating module further comprises:

[0075] The update effectiveness verification unit is connected with the core parameter updating unit and the auxiliary parameter updating unit, and is used to compare the simulation output of the updated twin model with the actual field data collected at the same period after each parameter update, and if the deviation is less than the set threshold, the update is confirmed to be effective.

[0076] Specifically, the update effectiveness verification unit is connected with the core parameter updating unit and the auxiliary parameter updating unit through a data interface, and triggers a virtual actual comparison verification process immediately after each parameter update is completed. The parameter update includes immediate updating of the core parameters and batch updating of the auxiliary parameters.

[0077] During verification, the simulation data output by the updated twin model, such as simulated production and simulated device temperature, is first obtained, and then the actual field data collected at the same period is retrieved. The same period refers to within 10 minutes after the update, and the average deviation of the two is calculated , wherein is the amount of verification data, is the jth simulation data, is the jth actual data. The preset set threshold : for example, the set threshold of simulated production is 0.5 m³ / h, and the set threshold of simulated device temperature is 0.3℃. If the calculated , a verification report of “update effective” is generated and stored in the system database; for example, the average deviation of the simulated production after the update is 0.3 m³ / h<0.5 m³ / h, which is determined to be effective, ensuring that the updated twin model can accurately reflect the actual scene and provide a reliable basis for subsequent simulation.

[0078] In this embodiment, when the error of the update validity verification unit exceeds the standard, the backtracking mechanism is automatically triggered, and the parameters of the corresponding change pattern model in the data preprocessing and change pattern modeling module are recalibrated by the twin model parameter adaptive update module.

[0079] Specifically, when updating the average deviation calculated by the validity verification unit In other words, when the deviation exceeds the standard, a "deviation exceeds the standard" signal is automatically sent to the twin model parameter adaptive update module to trigger the backtracking mechanism.

[0080] After the backtracking mechanism is activated, the twin model parameter adaptive update module first retrieves the basic data on which this parameter update is based. This basic data includes the collected real-time data, the output data of the formation characteristic change model and the equipment parameter fluctuation model. By tracing the data source, the root cause of the deviation is located: for example, if it is found that the predicted value output by the model deviates greatly from the actual data, it is determined that the change law model parameters in the data preprocessing and change law modeling module are inaccurate.

[0081] Subsequently, the twin model parameter adaptive update module sends a parameter calibration command to the data preprocessing and variation pattern modeling module to recalibrate the parameters of the corresponding variation pattern model. For example, for the stratigraphic characteristic variation model, the formula is: If the porosity predicted by the model deviates significantly from the actual porosity, adjust the attenuation coefficient. The value was adjusted from the initial 0.002 m⁻³ to 0.0018 m⁻³; for the equipment parameter fluctuation model, the formula is as follows: If the predicted vibration frequency deviates significantly, adjust the attenuation coefficient. The value was adjusted from 0.02 Hz / h to 0.019 Hz / h. After calibration, the parameter update and validity verification process was repeated until the deviation met the requirements, forming a closed loop of "update-verification-backtracking-calibration" to ensure that the model always has high accuracy.

[0082] In this embodiment, the simulation result feedback optimization module includes:

[0083] The deviation analysis engine is used to break down the deviation between the simulated output of the twin model and the actual collected data, locate the source of the deviation as insufficient acquisition frequency of the multi-dimensional dynamic data acquisition module or poor adaptability of the data preprocessing and change law modeling module to specific working conditions, and generate a deviation attribution report.

[0084] It also includes a closed-loop optimization strategy unit, which is connected to the deviation analysis engine. This unit is used to adjust the acquisition frequency of corresponding parameters in the multi-dimensional dynamic data acquisition module based on the deviation attribution results, or to update the training samples of the machine learning algorithm in the data preprocessing and change pattern modeling module, or to adjust the correlation coefficient of the physical mechanism model.

[0085] Specifically, the deviation analysis engine compares the simulation data output by the twin model, such as simulated formation pressure and simulated equipment energy consumption, with the actual field data collected at the same time through a data comparison algorithm, calculates the total deviation , and locates the source of the deviation through a deviation decomposition model: if the data collection deviation accounts for more than 50%, of which is delayed data, it is determined that the collection frequency of the multi-dimensional dynamic data collection module is insufficient; if the modeling deviation accounts for more than 60%, of which is the model calculation value, it is determined that the data preprocessing and change law modeling module has poor adaptability to specific working conditions such as high sand content and high pressure working conditions. The deviation analysis engine generates a deviation attribution report based on the positioning results, clearly indicating the type, proportion and improvement direction of the deviation.

[0086] After receiving the deviation attribution report, the closed-loop optimization strategy unit formulates an optimization scheme: if the attribution result is insufficient collection frequency, the stable threshold of the corresponding parameter in the multi-dimensional dynamic data collection module is adjusted , for example, the stable threshold of the formation pressure is lowered from 0.8 MPa to 0.5 MPa, triggering higher frequency collection; if the attribution result is poor model adaptability, the training samples of the machine learning algorithm in the data preprocessing and change law modeling module are updated, for example, 300 groups of historical data of high sand content working conditions are supplemented, or the correlation coefficient of the physical mechanism model is adjusted, for example, the correction coefficient of the viscosity coefficient in Darcy's law is adjusted from 1.0 to 1.2, improving the model's adaptability to special working conditions. Through targeted optimization, the simulation deviation is continuously reduced, and the system reliability is improved.

[0087] In this embodiment, the simulation result feedback optimization module further comprises:

[0088] A decision support enhancement unit is connected to the closed-loop optimization strategy unit and is used to simulate the impact of different collection parameters on collection efficiency or equipment life based on the updated twin model, and output quantitative optimization suggestions.

[0089] Specifically, the decision support enhancement unit is connected to the closed-loop optimization strategy unit through a data bus, and obtains the updated twin model data after the closed-loop optimization strategy unit completes the data collection frequency or model parameter optimization.

[0090] The unit constructs a multi-scenario simulation environment based on the updated twin model, simulates the influence of different acquisition parameters, such as wellhead flow rate and production pressure, on the acquisition efficiency or the service life of the equipment, for example, simulates the production efficiency and the service life loss of key components of the equipment, such as the crank of the oil pumping unit, when the wellhead flow rate is 15 m³ / h, 20 m³ / h and 25 m³ / h respectively, and simulates the flow state of the formation fluid and the degree of pipeline wear when the production pressure is 12 MPa, 15 MPa and 18 MPa respectively.

[0091] Subsequently, the unit calculates quantitative indicators such as the acquisition efficiency improvement rate and the equipment life extension time in each scenario through a quantitative analysis algorithm, and generates an optimization suggestion report; for example, the report clearly states that "when the wellhead flow rate is adjusted to 20 m³ / h, the acquisition efficiency is improved and the equipment life consumption rate is reduced", providing a quantitative basis for staff to develop an acquisition plan and helping staff find the optimal balance between acquisition efficiency and equipment protection, thereby improving the overall efficiency of oil acquisition.

[0092] Please refer to Figure 2 The application also provides an oil acquisition simulation method based on digital twin technology, which is applied to the oil acquisition simulation system based on digital twin technology in any of the above aspects, and comprises the following steps:

[0093] Real-time acquisition of formation property dynamic data and equipment state feature data, and dynamic adjustment of acquisition frequency according to the fluctuation amplitude of the acquisition data;

[0094] Pretreatment of the acquired formation property dynamic data and equipment state feature data, and dynamic training of a formation property change model and an equipment parameter fluctuation model based on the fusion of oil production physical mechanism and machine learning modeling, to realize the prediction of the future short-term change trend of the formation property dynamic data and the equipment state feature data;

[0095] According to the influence weight of the parameters on the acquisition process, the priority of the twin model parameters is divided, and a hierarchical incremental updating mechanism is adopted according to the parameter priority to compare the deviation between the real-time data acquired and the current parameters of the twin model in real time, and when the deviation exceeds the preset accuracy threshold, the parameter correction amount is calculated based on the formation property change model and the equipment parameter fluctuation model that have been constructed, and the corresponding formation property parameters and equipment state parameters in the twin model are incrementally updated;

[0096] A deviation analysis engine is constructed to compare the simulation output of the twin model with the actual field data collected at the same period, and to decompose and attribute the deviation;

[0097] Based on the deviation attribution result, the data acquisition frequency or the model parameters are optimized in reverse, and parameter sensitivity analysis is provided.

[0098] Specifically, step 1: real-time data acquisition and dynamic frequency adjustment. Deploy distributed optical fiber sensors in the oil well formation to collect dynamic data such as formation porosity, permeability, and other formation characteristics; deploy MEMS sensors in key components of wellhead equipment, such as oil pumping unit reducer, wellhead valve, to collect equipment vibration frequency, component temperature fluctuation and other equipment state characteristic data. Through the adaptive acquisition controller, calculate the parameter fluctuation amplitude , preset stable threshold , when , reduce the acquisition frequency, when , increase the acquisition frequency, to ensure the accuracy and economy of data acquisition.

[0099] Step 2: data preprocessing and rule modeling. Use 3σ algorithm to eliminate outliers in collected data, and use linear interpolation method to complete occasional data gaps; use the preprocessed data as input, based on the physical mechanism of oil production, build the basic framework such as Darcy's law, introduce LSTM or ARIMA time series prediction algorithm, train the formation characteristic change model and equipment parameter fluctuation model, such as porosity change model with production, equipment parameter fluctuation model such as vibration frequency decay model with running time, to predict the future short-term change trend of data.

[0100] Step 3: hierarchical incremental update of twin model parameters. Divide the parameter priority by AHP, into core influence layer and auxiliary influence layer, calculate the real-time deviation , when , exceed the preset accuracy threshold, calculate the correction amount based on the model built in step 2, update the core parameters in real time; for auxiliary influence layer parameters, aggregate the deviation according to the preset period, combine with the long-term trend prediction results of the model, and batch fine-tune the auxiliary parameters to ensure that the model parameters are consistent with the actual scene.

[0101] Step 4: deviation analysis and attribution. Build a deviation analysis engine to compare the simulation output of the twin model with the actual collected field data at the same period, such as simulated production and simulated equipment temperature, calculate the total deviation and decompose it into data collection deviation and modeling deviation, locate the deviation source, such as insufficient collection frequency and poor model adaptability, and generate a deviation attribution report.

[0102] Step 5: reverse optimization and sensitivity analysis. Based on the deviation attribution results, adjust the data collection frequency, such as lowering the stable threshold, or model parameters, such as supplementing training samples and adjusting physical mechanism model coefficients; at the same time, calculate the change rate of simulation results when each parameter changes by 1%, and carry out parameter sensitivity analysis to determine the key optimization direction and provide support for data collection scheme optimization.

[0103] The method solves the problems of parameter updating lag and large deviation of the existing simulation method through a closed-loop process of "collection-modeling-updating-optimization", and improves the accuracy of the simulation result.

[0104] In summary, the present application effectively solves the problem that the existing oil collection simulation system only initially imports data, and the subsequent twin model parameters are difficult to adjust in real time with actual data, resulting in increased deviation and inconsistency between simulation results and reality, by the cooperative work of multiple modules. Among them, the multi-dimensional dynamic data collection module dynamically adjusts the collection frequency according to the data fluctuation amplitude, which not only ensures accurate data collection when the fluctuation is intense, but also avoids redundant collection when it is stable; the data preprocessing and change law modeling module eliminates outliers, fills in data gaps, and combines oil production physical mechanism and machine learning modeling to predict short-term data trends; the twin model parameter adaptive updating module updates the parameters according to the parameter influence weight, and the core parameters are updated in real time, the auxiliary parameters are adjusted periodically, and the model is calibrated through effectiveness verification and backtracking mechanism to ensure that the parameters are consistent with the actual situation; the simulation result feedback optimization module decomposes the deviation attribution and reversely optimizes the collection frequency and model parameters, provides parameter sensitivity analysis and quantitative decision suggestions, and finally makes the twin model accurately reflect the actual scene, reduces the simulation deviation, provides a reliable basis for the optimization of the collection scheme, and improves the safety and efficiency of oil collection.

[0105] It should be noted that, in this text, relational terms such as first and second are used merely to distinguish one entity or action from another, without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variant thereof are intended to cover non-exclusive inclusions, so that a process, method, article, or apparatus that includes a list of elements does not only include those elements, but also includes other elements not expressly listed, or inherent to such a process, method, article, or apparatus.

[0106] Although embodiments of the present application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and alterations can be made thereto without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A petroleum production simulation system based on digital twin technology, characterized in that, Comprise: Multi-dimensional dynamic data acquisition module, for real-time acquisition of formation characteristics dynamic data and equipment state characteristic data, and dynamically adjusting the acquisition frequency according to the fluctuation amplitude of the acquisition data; The multi-dimensional dynamic data acquisition module comprises a formation sensing unit, a micro-electro-mechanical sensing module and a self-adaptive acquisition controller; the formation sensing unit is used for acquiring at least one formation characteristic dynamic data in formation porosity, formation permeability and formation pressure gradient; the micro-electro-mechanical sensing module is used for acquiring at least one equipment state characteristic data in equipment vibration frequency, component temperature fluctuation and sealing performance attenuation; the self-adaptive acquisition controller is used for adjusting the acquisition frequency through comparison of a parameter fluctuation amplitude with a preset stable threshold value , wherein, is a parameter value at a current moment, is a parameter value at a previous acquisition moment, when the acquisition frequency is reduced, and when the acquisition frequency is increased. Data preprocessing and change law modeling module, connected with the multi-dimensional dynamic data acquisition module, for preprocessing the formation characteristics dynamic data and the equipment state characteristic data collected by the multi-dimensional dynamic data acquisition module, and dynamically training the formation characteristics change model and the equipment parameter fluctuation model based on the fusion modeling of petroleum exploitation physical mechanism and machine learning; the preprocessing adopts 3σ outlier identification algorithm to eliminate interference values, and linear interpolation method to complete data gaps; the fusion modeling takes the petroleum exploitation physical mechanism as the basic framework, and introduces LSTM or ARIMA time series prediction algorithm; the formation characteristics change model is a formation porosity change curve model with production, and the equipment parameter fluctuation model is a device vibration frequency decay model with running time; The twin model parameter adaptive updating module is connected with the data preprocessing and variation law modeling module, and is used for dividing a twin model parameter priority according to an influence weight of a parameter on a collection process, and adopting a hierarchical incremental updating mechanism according to the parameter priority; the parameter priority is divided into a core influence layer parameter and an auxiliary influence layer parameter through an analytic hierarchy process, the core influence layer parameter adopts a real-time comparison instant updating mode, and the auxiliary influence layer parameter adopts a periodic verification batch fine-tuning mode; a deviation between real-time data collected in real time and a current parameter of the twin model is compared in real time, the deviation is calculated through a formula , wherein, is the real-time collected data, is a current parameter value of the twin model, and when the deviation exceeds a preset precision threshold , a parameter correction amount is calculated based on a constructed stratum characteristic variation model and a device parameter fluctuation model , a corresponding stratum characteristic parameter and a device state parameter in the twin model are incrementally updated, wherein, is a model predicted reasonable parameter value; the twin model parameter adaptive updating module further includes an updating validity verification unit, which is used for comparing simulation output with actual collected data after each parameter update, and calculating an average deviation through a formula , wherein, m is a verification data amount, is jth simulation data, is jth actual data, and if the average deviation is less than a set threshold , it is confirmed that the updating is valid, and if the average deviation is greater than the set threshold , a backtracking mechanism is triggered to recalibrate the variation law model parameter. The simulation result feedback optimization module is connected with the twin model parameter adaptive updating module and the multi-dimensional dynamic data acquisition module, is used for constructing a deviation analysis engine, and analyzes the deviation of the twin model simulation output and the actual collected field data in the same period through a deviation disassembly formula The deviation of the twin model simulation output and the actual collected field data in the same period is disassembled and attributed, The data acquisition deviation, The model modeling deviation; based on the deviation attribution result, the acquisition frequency of the multi-dimensional dynamic data acquisition module or the model parameters of the data preprocessing and change rule modeling module are reversely optimized; and a parameter sensitivity analysis function is provided by calculating the change rate of the simulation result when the parameter changes by 1%.

2. The digital twin technology based oil production simulation system of claim 1, wherein, The reverse optimization of the simulation result feedback optimization module includes adjusting the preset stable threshold of the acquisition frequency, updating the training samples of the machine learning algorithm, and adjusting the correlation coefficient of the physical mechanism model.

3. The digital twin technology based oil recovery simulation system of claim 1, wherein, The simulation result feedback optimization module further comprises a decision support enhancement unit for simulating the influence of different acquisition parameters on the acquisition efficiency or the equipment life based on the updated twin model, and outputting quantitative optimization suggestions.

4. The oil production simulation method based on the digital twin technology, applied to the oil production simulation system based on the digital twin technology in any one of claims 1-3, characterized in that, Comprise: Real-time acquisition of formation characteristics dynamic data and equipment state characteristic data, through calculation parameter fluctuation amplitude Comparison with the preset stable threshold Dynamic adjustment of acquisition frequency, wherein, The current time parameter value, The parameter value at the previous acquisition time, when Lower the acquisition frequency, when Increase the acquisition frequency; The collected formation characteristics dynamic data and equipment state characteristic data are preprocessed, 3σ outlier identification algorithm is used to eliminate interference values, and linear interpolation method is used to complete data gaps; the preprocessed data is input, fusion modeling is carried out based on the petroleum exploitation physical mechanism and LSTM or ARIMA time series prediction algorithm, dynamic training is carried out on the formation characteristics change model of formation porosity change with production and the equipment parameter fluctuation model of equipment vibration frequency decay with running time, and the future short-term change trend of data is predicted; The parameter priority of the twin model is divided into a core influence layer and an auxiliary influence layer by an analytic hierarchy process, a layered incremental updating mechanism is adopted, the core influence layer parameters adopt a real-time comparison and immediate updating mode, and the auxiliary influence layer parameters adopt a periodic checking and batch fine-tuning mode; the deviation between the real-time data collected in real time and the current parameters of the twin model is calculated in real time , wherein is the real-time collected data, is the current parameter value of the twin model, and when the deviation exceeds a preset precision threshold , a parameter correction amount is calculated based on a constructed change model , and the corresponding formation characteristic parameters and equipment state parameters are incrementally updated, wherein is a parameter reasonable value predicted by the model; after each parameter update, the effectiveness is verified, and the average deviation is calculated by a formula , wherein m is the checking data amount, is the jth simulation data, is the jth actual data, and if the average deviation is less than a set threshold , it is confirmed to be effective, and if the average deviation is greater than the set threshold , the change law model parameters are backtracked and calibrated; A bias analysis engine is constructed to decompose the bias by a bias decomposition formula The bias between the simulation output of the twin model and the actual field data collected at the same period is decomposed and attributed, wherein, the data collection bias, the model modeling bias; Based on the deviation attribution result, the data acquisition frequency or the model parameter is optimized in reverse, and the parameter sensitivity analysis is provided by calculating the change rate of the simulation result when the parameter changes by 1%.

Images