Shale oil fracturing fluid return displacement prediction method and device based on time sequence prediction
By combining time-series prediction methods with LSTM, GRU, and XGBoost models, the problems of data accuracy and universality in the prediction of shale oil fracturing fluid flowback volume in existing technologies have been solved, enabling more accurate flowback fluid volume prediction and equipment optimization, thereby improving production efficiency and economic benefits.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- PETROCHINA CO LTD
- Filing Date
- 2024-11-15
- Publication Date
- 2026-05-15
AI Technical Summary
Existing methods for predicting shale oil fracturing fluid flowback are insufficient in terms of high data accuracy requirements and complex geological conditions. Furthermore, modern artificial intelligence methods have limited model universality, making it difficult to achieve accurate predictions.
A time-series-based prediction method is adopted, combining LSTM, GRU and XGBoost models. By training historical time-series and non-time-series feature sample data, a backflow volume prediction model is constructed. The final prediction value is generated by weight summation, thereby improving the prediction accuracy.
It enables more accurate prediction of backflow volume, guides optimized equipment operation, reduces operating costs, and improves production efficiency and equipment utilization.
Smart Images
Figure CN122045798A_ABST
Abstract
Description
Technical Field
[0001] This specification relates to the field of oil and gas field exploration and development, and in particular to a method and apparatus for predicting the flowback volume of fracturing fluid in shale oil based on time-series prediction. Background Technology
[0002] The application of large-scale volumetric fracturing technology in the Southwest Oil and Gas Field has significantly increased oil and gas production, particularly in shale oil and gas development. By accurately predicting flowback fluid volume and processing real-time data, it guides the optimization of post-fracturing flowback regimes, the timing of artificial lift interventions, and process selection, thereby improving production efficiency and formation energy utilization, reducing operating costs, and maximizing economic benefits. Overall, accurate prediction of flowback fluid volume is a key step in achieving scientific management and efficient production of shale oil wells.
[0003] Currently, commonly used prediction methods are mainly divided into traditional mechanistic modeling analysis and modern artificial intelligence and machine learning methods. Traditional mechanistic modeling analysis includes single-factor and multi-factor analyses, which consider the influence of different factors on the flowback rate by establishing simple or complex prediction models and using reservoir engineering and geological modeling software for numerical simulation. Although these methods can provide useful predictions in some cases, they require high accuracy of well logging data and may have limitations when dealing with complex geological conditions. Statistical analysis and empirical formulas, such as linear regression, nonlinear regression, and Monte Carlo simulation-geomechanical data (inversion), also play important roles in flowback rate prediction, while modern artificial intelligence and machine learning methods offer new solutions. Although these methods perform well in data-driven situations, they still require high-quality data input, and the generalizability of the models is limited due to the complexity of the flowback mechanism. Summary of the Invention
[0004] To address the problems of existing technologies, this specification provides a method and apparatus for predicting shale oil fracturing fluid flowback based on time-series prediction.
[0005] This specification provides an embodiment of a method for predicting shale oil fracturing fluid flowback volume based on time-series prediction. The method includes: inputting monitoring data into a flowback fluid volume prediction model; wherein the flowback fluid volume prediction model includes: a first prediction model and a second prediction model, the first prediction model being obtained by training a time-series feature algorithm using historical time-series feature data and historical flowback fluid volumes, and the second prediction model being obtained by training a non-time-series feature algorithm using historical non-time-series feature data and historical flowback fluid volumes; assigning weights to the first flowback fluid volume prediction value output by the first prediction model and the second flowback fluid volume prediction value output by the second prediction model, and summing them to obtain the final flowback fluid volume prediction value.
[0006] According to one aspect of an embodiment of this specification, the backflow volume prediction model is trained as follows: A training sample dataset is determined, comprising: historical time-series feature sample data, historical non-time-series feature sample data, and corresponding historical backflow volumes; the historical time-series feature sample data is input into a first prediction model to obtain an initial prediction result output by the first prediction model; the non-time-series feature sample data is input into a second prediction model to obtain an initial prediction result output by the second prediction model; based on the loss function of the first prediction model, a loss value between the initial prediction result output by the first prediction model and the historical backflow volumes is calculated, and the first prediction model is iteratively trained based on the loss value; based on the loss function of the second prediction model, a loss value between the initial prediction result output by the second prediction model and the historical backflow volumes is calculated, and the second prediction model is iteratively trained based on the loss value to obtain the backflow volume prediction model.
[0007] According to one aspect of an embodiment of this specification, the method further includes: inputting the temporal feature sample data into an LSTM initial model and a GRU initial model to obtain a first prediction result output by the LSTM initial model and a second prediction result output by the GRU initial model; and inputting the non-temporal feature sample data into an XGBoost initial model to obtain a third prediction result output by the XGBoost initial model.
[0008] According to one aspect of the embodiments of this specification, calculating the loss value between the initial prediction result output by the first prediction model and the historical backflow volume based on the loss function of the first prediction model, and iteratively training the first prediction model based on the loss value, includes: using the first loss function to calculate the loss value between the first prediction result output by the LSTM initial model and the historical backflow volume, and the loss value between the second prediction result output by the GRU initial model and the historical backflow volume; if the loss value is greater than or equal to a preset threshold, iterating through the number of hidden layers, the number of nodes, and the number of iterations of the LSTM initial model and the GRU initial model until the loss value converges to the preset threshold.
[0009] According to one aspect of the embodiments of this specification, calculating the loss value between the initial prediction result output by the second prediction model and the historical backflow volume based on the loss function of the second prediction model, and iteratively training the second prediction model based on the loss value to obtain the backflow volume prediction model includes: calculating the loss value between the third prediction result output by the XGBoost initial model and the historical backflow volume based on the second loss function; if the loss value is greater than or equal to a preset threshold, iterating through the number of hidden layers, the number of nodes, and the number of iterations of the XGBoost initial model until the loss value converges to the preset threshold.
[0010] According to one aspect of the embodiments of this specification, the historical time-series feature sample data is determined by: determining the pressure changes, temperature changes, and flow changes within the historical time period based on the collected pressure, temperature, and flow data; determining the pressure changes, temperature changes, and flow changes within the historical time period to form a historical time-series feature sequence, wherein the historical time-series feature sequence is the historical time-series feature sample data.
[0011] According to one aspect of the embodiments of this specification, assigning weights to the first predicted backflow volume and the second predicted backflow volume, and summing them to obtain the final predicted backflow volume includes: assigning a first weight to the model trained by the LSTM initial model, assigning a second weight to the model trained by the GRU initial model, and assigning a third weight to the model trained by the XGBoost initial model; and determining the final predicted backflow volume based on the first weight, the second weight, the third weight, and the predicted values output by each initial model.
[0012] This specification provides an embodiment of a shale oil fracturing fluid flowback prediction device based on time-series prediction. The device includes: an input unit for inputting monitoring data into a flowback fluid prediction model; wherein the flowback fluid prediction model includes: a first prediction model and a second prediction model, the first prediction model being obtained by training a time-series feature algorithm using historical time-series feature sample data and historical flowback fluid volume, and the second prediction model being obtained by training a non-time-series feature algorithm using historical non-time-series feature sample data and historical flowback fluid volume; and a final flowback fluid volume prediction value determination unit for assigning weights to the first flowback fluid volume prediction value output by the first prediction model and the second flowback fluid volume prediction value output by the second prediction model, and summing them to obtain the final flowback fluid volume prediction value.
[0013] This specification also provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the time-series prediction method for predicting shale oil fracturing fluid flowback.
[0014] This specification also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the time-series prediction-based shale oil fracturing fluid flowback prediction method.
[0015] This manual utilizes a combination of XGBoost and LSTM models to more accurately predict the changing trends of return fluid volume. Based on the prediction results, it adjusts equipment load and operating parameters to ensure optimal equipment operation, avoiding overload or inefficient operation, improving equipment utilization and production efficiency, and thus reducing equipment maintenance and operating costs. It guides the optimization of equipment operating parameters to ensure efficient operation. Through the collaborative work of XGBoost, GRU, and LSTM models, it comprehensively captures the complex relationships between different feature data, ensuring the accuracy and comprehensiveness of predictions. This provides reliable decision-making support for production management, optimizing production decision-making processes and improving production efficiency. Attached Figure Description
[0016] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0017] Figure 1 The diagram shown is a flowchart of a time-series prediction method for predicting shale oil fracturing fluid flowback, as described in an embodiment of this specification.
[0018] Figure 2 The diagram shown is a flowchart of a method for training a flowback volume prediction model according to an embodiment of this specification.
[0019] Figure 3 The diagram shown is a flowchart of a method for iteratively training a first prediction model according to an embodiment of this specification.
[0020] Figure 4 The diagram shown is a flowchart of an iterative training method for a second prediction model according to an embodiment of this specification.
[0021] Figure 5 The diagram shown is a flowchart of a method for determining historical time-series feature sample data according to an embodiment of this specification.
[0022] Figure 6 The diagram shown is a flowchart of a method for determining the predicted final backflow volume according to an embodiment of this specification.
[0023] Figure 7 The diagram shown is a schematic representation of a shale oil fracturing fluid flowback prediction device based on time-series prediction, as described in an embodiment of this specification.
[0024] Figure 8 The diagram shown is a structural schematic of a computer device according to an embodiment of this specification.
[0025] Explanation of symbols in the attached drawings:
[0026] 701. Input Unit;
[0027] 702. Unit for determining the final predicted value of the backflow liquid volume;
[0028] 802. Computer equipment;
[0029] 804, Processor;
[0030] 806. Memory;
[0031] 808. Drive mechanism;
[0032] 810. Input / Output Module;
[0033] 812. Input devices;
[0034] 814. Output devices;
[0035] 816. Presentation equipment;
[0036] 818. Graphical User Interface;
[0037] 820. Network interface;
[0038] 822. Communication link;
[0039] 824. Communication bus. Detailed Implementation
[0040] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this specification.
[0041] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.
[0042] This specification provides the operational steps of the methods described in the embodiments or flowcharts, but based on conventional or non-inventive labor, more or fewer operational steps may be included. The order of steps listed in the embodiments is merely one possible execution order among many and does not represent the only possible execution order. In actual system or device products, the methods shown in the embodiments or drawings can be executed sequentially or in parallel.
[0043] It should be noted that the time-series prediction-based method and device for predicting shale oil fracturing fluid flowback in this manual can be used in the field of oil and gas field exploration and development. This manual does not limit the application field of the time-series prediction-based method and device for predicting shale oil fracturing fluid flowback.
[0044] Figure 1 The diagram shown is a flowchart of a method for predicting shale oil fracturing fluid flowback based on time-series prediction, according to an embodiment of this specification. The method includes the following steps:
[0045] Step 101: Input the real-time monitoring data into the backflow volume prediction model; wherein, the backflow volume prediction model includes: a first prediction model and a second prediction model, the first prediction model is obtained by training a time-series feature algorithm with historical backflow volume and historical backflow volume, and the second prediction model is obtained by training a non-time-series feature algorithm with historical non-time-series feature sample data and historical backflow volume.
[0046] In this step, real-time monitoring data includes one or any combination of current pressure, temperature, and flow rate data. The monitoring data is input into a pre-trained backflow volume prediction model to obtain the model's prediction output. In this specification, the time step of the backflow volume prediction model can be set to 5, i.e., using the historical load of the previous five days to predict the backflow load for the next day. A function is constructed based on the time-series data to perform feature extraction. The input feature set after function processing is a matrix of (200, 5), and the output label set is a matrix of (200, 1).
[0047] The backflow volume prediction model in this specification consists of a first prediction model and a second prediction model. Specifically, the first prediction model is obtained by training a time-series feature algorithm based on historical time-series feature sample data and historical backflow volume. The second prediction model is obtained by training a non-time-series feature algorithm based on historical non-time-series feature sample data and historical backflow volume. In the embodiments of this specification, the historical time-series feature sample data and the historical non-time-series feature sample data belong to the training sample dataset for training the backflow volume prediction model. Correspondingly, the first prediction model is a model that can identify the data features of the historical time-series feature sample data, and the second prediction model is a model that can identify the data features of the historical non-time-series feature sample data. During the training of the backflow volume prediction model, the historical time-series feature sample data and the historical non-time-series feature sample data are input together into the first prediction model and the second prediction model. The first prediction model and the second prediction model respectively identify sample data with different time-series features and are trained separately.
[0048] Step 102: Based on the first predicted value of the backflow volume output by the first prediction model and the second predicted value of the backflow volume output by the second prediction model, assign preset weights to the first predicted value of the backflow volume and the second predicted value of the backflow volume respectively, and sum them to obtain the final predicted value of the backflow volume.
[0049] In this step, the weights of historical time-series and non-time-series characteristic sample data on the amount of backflow are determined, and then the final predicted backflow volume is calculated. Historical time-series characteristic sample data has a more direct and significant impact on the amount of backflow, while historical non-time-series characteristic sample data has a smaller impact. Therefore, a relatively large weight can be assigned to the first predicted backflow volume, and a smaller weight can be assigned to the second predicted backflow volume.
[0050] Figure 2 The diagram shown is a flowchart of a method for training a flowback volume prediction model according to an embodiment of this specification, which specifically includes the following steps:
[0051] Step 201: Determine the training sample dataset, which includes: historical time-series feature sample data, historical non-time-series feature sample data, and corresponding historical flowback fluid volume. In the embodiments of this specification, historical data on flowback fluid volume, as well as multi-dimensional historical monitoring data such as pressure, temperature, and flow rate, and data such as formation characteristics and operating parameters are collected from sensors and monitoring equipment in oil and gas wells. Among them, data such as pressure, temperature, and flow rate change over time; therefore, these data are unified as historical time-series feature sample data. Formation characteristics, rock characteristics, and operating parameters, however, have no specific relationship with time changes; these parameters generally do not change over time. Therefore, these data are referred to as historical non-time-series feature sample data.
[0052] Step 202: Input the historical time series feature sample data into the first prediction model to obtain the initial prediction result output by the first prediction model.
[0053] Step 203: Input the historical non-time-series feature sample data into the second prediction model to obtain the initial prediction result output by the second prediction model. Specifically, the first prediction model can be either an LSTM model or a GRU model. The second prediction model can be either an XGBoost model.
[0054] LSTM, through its memory gating mechanism, excels at capturing long-term dependencies and handling complex changes in time series data. During model training, LSTM serves as the core component for capturing long-term series variations. GRU, a simplified version of LSTM, requires less computation and is primarily used to capture shorter-term time series dependencies, making it suitable for handling short-term fluctuations. Combining the two allows for a comprehensive capture of the complex dynamic features in time series data.
[0055] This specification demonstrates how the collaboration between GRU and LSTM models is effective. GRU is more efficient than LSTM in handling short-term dependencies, accelerating model training and inference, and further improving real-time performance and computational efficiency. Meanwhile, LSTM focuses on capturing long-term dependencies, ensuring the accuracy of predictions. Through efficient deep learning models and optimization algorithms, real-time data processing and prediction are achieved, providing timely guidance for production operations, optimizing processes, and improving efficiency. The combination of GRU and LSTM helps to quickly process time series with both short-term and long-term dependencies, resulting in more accurate predictions, avoiding unnecessary operational interventions in production, and reducing trial-and-error costs.
[0056] XGBoost is used to handle non-time-series features, such as formation data and downhole pressure. Through ensemble learning, XGBoost captures complex nonlinear relationships within the model, working in conjunction with time-series prediction models (LSTM and GRU) to make the overall model more robust. In this step, a collaborative model of LSTM, GRU, and XGBoost is trained using historical time-series and non-time-series feature sample data. LSTM and GRU handle time-series data, while XGBoost handles non-time-series features.
[0057] Step 204: Calculate the loss value between the initial prediction result output by the first prediction model and the historical backflow volume according to the loss function of the first prediction model, and iteratively train the first prediction model based on the loss value.
[0058] In this step, the parameters of the first prediction model are adjusted using the gradient descent optimization algorithm to ensure that the first prediction model can accurately predict the amount of backflowed liquid. The loss function can be the mean squared error (MSE).
[0059]
[0060] Among them, y i This is the historical backflow volume. is the predicted value output by the first prediction model, and n is the number of samples.
[0061] Step 205: Based on the loss function of the second prediction model, calculate the loss value between the initial prediction result output by the second prediction model and the historical backflow volume, and iteratively train the second prediction model based on the loss value to obtain the backflow volume prediction model.
[0062] In the embodiments of this specification, time-series data is processed by LSTM and GRU, while XGBoost is used to process the feature-enhanced static data. Therefore, the historical time-series feature sample data is input into the initial LSTM model and the initial GRU model to obtain the first prediction result output by the initial LSTM model and the second prediction result output by the initial GRU model; the non-time-series feature sample data is input into the initial XGBoost model to obtain the third prediction result output by the initial XGBoost model.
[0063] Figure 3 The diagram shown is a flowchart of a method for iteratively training a first prediction model according to an embodiment of this specification, which specifically includes the following steps:
[0064] Step 301: Use the first loss function to calculate the loss value of the first prediction result output by the LSTM initial model and the historical backflow volume, and the loss value of the second prediction result output by the GRU initial model and the historical backflow volume.
[0065] In this step, mean squared error can be used to simultaneously calculate the loss value of the prediction results output by the initial LSTM model and the initial GRU model and the historical backflow volume. Specifically, the loss value of the first prediction result output by the initial LSTM model after each iteration of training is calculated, which is the actual value of the historical backflow volume. The loss value of the second prediction result output by the initial GRU model after each iteration of training is calculated, which is the actual value of the historical backflow volume.
[0066] Step 302: If the loss value is greater than or equal to the preset threshold, iterate through the number of hidden layers, nodes and iterations of the initial LSTM model and the initial GRU model until the loss value converges to the preset threshold.
[0067] The initial models are trained iteratively by adjusting the parameters of the initial LSTM and GRU models each time. The loss value between the predicted output and historical flowback fluid volume of the updated initial model in the next iteration is then calculated. This process is repeated iteratively until the initial model's predicted output and historical flowback fluid volume loss value are below a preset range, at which point the initial model training is considered complete. Adjusting model parameters during training can employ techniques such as early stopping to reduce reliance on high-precision logging data.
[0068] By combining XGBoost and LSTM models, the changing trends of return fluid volume can be predicted more accurately. Based on the prediction results, equipment load and operating parameters can be adjusted to ensure optimal equipment operation, avoiding overload or inefficient operation, improving equipment utilization and production efficiency, and thus reducing equipment maintenance and operating costs. This guides the optimization of equipment operating parameters, ensuring efficient equipment operation. Through the collaborative work of XGBoost, GRU, and the initial LSTM model, the complex relationships between different feature data are comprehensively captured, ensuring the accuracy and comprehensiveness of predictions. This provides reliable decision-making support for production management, optimizing production decision-making processes and improving production efficiency.
[0069] Figure 4 The diagram shown is a flowchart of an iterative training method for a second prediction model according to an embodiment of this specification, which specifically includes the following steps:
[0070] Step 401: Calculate the loss value between the third prediction result output by the XGBoost initial model and the historical backflow volume based on the second loss function. In this step, the mean squared error can be used to calculate the loss value between the prediction result output by the XGBoost initial model and the historical backflow volume. Specifically, calculate the loss value between the third prediction result output by the XGBoost initial model after each iteration of training and the actual value of the historical backflow volume.
[0071] Step 402: If the loss value is greater than or equal to the preset threshold, iterate through the number of hidden layers, nodes and iterations of the initial XGBoost model until the loss value converges to the preset threshold.
[0072] The initial XGBoost model is trained iteratively by adjusting its parameters each time. The loss value between the updated model's predicted output and historical backflow volume is then calculated and repeated. This process continues until the initial model's predicted output and historical backflow volume loss value are below a preset range, at which point the initial model training is considered complete.
[0073] In this specification, during the iterative training of the first and second prediction models, different hyperparameter combinations are traversed to optimize the number of hidden layers, nodes, and iterations in the initial LSTM and GRU models. The optimal hyperparameter combination is then selected based on the XGBoost tree parameters. The reference values for each hyperparameter in the first and second prediction models are shown in Table 1 below.
[0074] Table 1 Model Hyperparameter Combination Table
[0075] Network hyperparameter name Reference value Number of iterations (epochs) 50,100,200,300 batch size 16,32,64,128 number of hidden layers 1,2,3,4 Number of hidden layer nodes (units) 8,16,32,48
[0076] During the training of the first and second prediction models, the hyperparameter data in Table 1 can be selected for training, respectively. Throughout the hyperparameter optimization process, MSE (Mean Sequence Estimation) is the evaluation criterion. Grid search or random search can minimize MSE to select the optimal combination of hyperparameters, ultimately improving the model's prediction performance.
[0077] Python code example:
[0078]
[0079]
[0080]
[0081] Figure 5 The diagram shown is a flowchart of a method for determining historical time-series feature sample data according to an embodiment of this specification, which specifically includes the following steps:
[0082] Step 501: Based on the collected pressure, temperature, and flow rate data for the historical period, determine the pressure, temperature, and flow rate changes within the historical period. This step can use sliding window and hysteresis feature methods to extract historical data on the backflow volume, and further obtain the changing trends of pressure, temperature, and flow rate through differential calculation and moving average method; at the same time, hysteresis features are used to capture dynamic changes in different time periods.
[0083] Step 502: The pressure changes, temperature changes, and flow rate changes within the historical time period are used to form a historical time series feature sequence, which is the historical time series feature sample data.
[0084] This step combines the pressure, temperature, and flow rate data obtained in the previous step and applies the data to the LSTM time-series prediction model to form historical time-series feature sequences, thereby improving prediction accuracy and model performance. These historical time-series feature sequences are the historical time-series feature sample data described in this specification.
[0085] In other embodiments of this specification, sliding windows and hysteresis features can be used to extract historical data on the backflow volume to determine the time series of the backflow volume. This specification not only considers historical data on the backflow volume but also comprehensively utilizes real-time monitoring data from multiple dimensions such as pressure, temperature, and flow rate. Through feature extraction and fusion, a more comprehensive and accurate prediction model is constructed, improving the reliability and accuracy of the prediction.
[0086] Figure 6 The diagram shown is a flowchart of a method for determining the predicted final backflow volume according to an embodiment of this specification, which specifically includes the following steps:
[0087] Step 601: Assign a first weight to the model trained from the LSTM initial model, a second weight to the model trained from the GRU initial model, and a third weight to the model trained from the XGBoost initial model. In this embodiment, the LSTM and GRU initial models have similar performance, and the time-series feature sample data processed by these two models has a greater impact on the final expected predicted backflow volume than the non-time-series feature sample data processed by the XGBoost initial model. Therefore, the first and second weight values are set greater than the third weight. In this embodiment, the first and second weight values can both be set to 0.4, and the third weight value can be set to 0.2. This embodiment does not limit the specific values of the first, second, and third weights.
[0088] Step 602: Determine the final predicted value of the backflow volume based on the first weight, the second weight, the third weight, and the predicted values output by each initial model. Multiply the first weight by the predicted value output by the model trained from the LSTM initial model, multiply the second weight by the predicted value output by the model trained from the GRU initial model, and multiply the third weight by the predicted value output by the second prediction model trained from the XGBoost initial model. Finally, sum the three results to obtain the final predicted value of the backflow volume.
[0089] like Figure 7 The diagram shown is a schematic representation of a time-series-based shale oil fracturing fluid flowback prediction device according to an embodiment of this specification. The basic structure of the device is illustrated in the diagram. The functional units and modules can be implemented using software, or using general-purpose or specific chips to predict shale oil fracturing fluid flowback based on time-series prediction. The device specifically includes:
[0090] The input unit 701 is used to input monitoring data into the backflow volume prediction model; wherein, the backflow volume prediction model includes: a first prediction model and a second prediction model, the first prediction model is obtained by training a time-series feature algorithm with historical backflow volume and historical backflow volume, and the second prediction model is obtained by training a non-time-series feature algorithm with historical non-time-series feature sample data and historical backflow volume;
[0091] The final backflow volume prediction unit 702 is used to assign weights to the first backflow volume prediction value and the second backflow volume prediction value output by the first prediction model and the second prediction model in the backflow volume prediction model, respectively, and sum them to obtain the final backflow volume prediction value.
[0092] like Figure 8 The diagram illustrates a computer device provided in an embodiment of this specification. The shale oil fracturing fluid flowback prediction method based on time-series prediction described in this application can be applied to the computer device. The computer device 802 may include one or more processors 804, such as one or more central processing units (CPUs), each of which can implement one or more hardware threads. The computer device 802 may also include any memory 806 for storing any kind of information such as code, settings, data, etc. Non-limitingly, for example, the memory 806 may include any type of RAM, any type of ROM, flash memory, hard disk, optical disk, etc. More generally, any memory can use any technology to store information. Further, any memory can provide volatile or non-volatile retention of information. Further, any memory can represent a fixed or removable component of the computer device 802. In one case, when the processor 804 executes associated instructions stored in any memory or combination of memories, the computer device 802 can perform any operation of the associated instructions. The computer device 802 also includes one or more drive mechanisms 808 for interacting with any memory, such as a hard disk drive mechanism, an optical disk drive mechanism, etc.
[0093] Computer device 802 may also include an input / output module 810 (I / O) for receiving various inputs (via input device 812) and providing various outputs (via output device 814). A specific output mechanism may include a presentation device 816 and an associated graphical user interface (GUI) 818. In other embodiments, the input / output module 810 (I / O), input device 812, and output device 814 may be omitted, and the device may function solely as a computer device within a network. Computer device 802 may also include one or more network interfaces 820 for exchanging data with other devices via one or more communication links 822. One or more communication buses 824 couple the components described above together.
[0094] Communication link 822 can be implemented in any way, such as via a local area network, a wide area network (e.g., the Internet), a point-to-point connection, or any combination thereof. Communication link 822 may include any combination of hardwired links, wireless links, routers, gateway functions, name servers, etc., governed by any protocol or combination of protocols.
[0095] Corresponding to Figures 1 to 6 In addition to the methods described above, embodiments of this specification also provide a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the methods described above.
[0096] This specification also provides computer-readable instructions, wherein when a processor executes the instructions, the program therein causes the processor to perform the following... Figures 1 to 6 The method shown.
[0097] It should be understood that in the various embodiments of this specification, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this specification.
[0098] It should also be understood that, in the embodiments of this specification, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this specification generally indicates that the preceding and following related objects have an "or" relationship.
[0099] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this specification can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this specification.
[0100] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0101] In the several embodiments provided in this specification, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the couplings or direct couplings or communication connections shown or discussed may be indirect couplings or communication connections through some interfaces, devices, or units, or they may be electrical, mechanical, or other forms of connection.
[0102] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments described in this specification, depending on actual needs.
[0103] Furthermore, the functional units in the various embodiments of this specification can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0104] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this specification, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this specification. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0105] This specification uses specific embodiments to illustrate the principles and implementation methods of this specification. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this specification. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this specification. Therefore, the content of this specification should not be construed as a limitation of this specification.
Claims
1. A method for predicting shale oil fracturing fluid flowback based on time-series prediction, characterized in that, The method includes: The monitoring data is input into the backflow volume prediction model; wherein, the backflow volume prediction model includes: a first prediction model and a second prediction model, the first prediction model is obtained by training a time-series feature algorithm with historical time-series feature data and historical backflow volume, and the second prediction model is obtained by training a non-time-series feature algorithm with historical non-time-series feature data and historical backflow volume; Based on the first predicted value of the backflow volume output by the first prediction model and the second predicted value of the backflow volume output by the second prediction model, weights are assigned to the first predicted value of the backflow volume and the second predicted value of the backflow volume, and the sums are obtained to obtain the final predicted value of the backflow volume.
2. The method according to claim 1, characterized in that, The backflow volume prediction model was trained in the following manner: Determine the training sample dataset, which includes: historical time-series feature sample data, historical non-time-series feature sample data, and the corresponding historical backflow volume; The historical time-series feature sample data is input into the first prediction model to obtain the initial prediction result output by the first prediction model; The historical non-time-series feature sample data is input into the second prediction model to obtain the initial prediction result output by the second prediction model; Based on the loss function of the first prediction model, calculate the loss value between the initial prediction result output by the first prediction model and the historical backflow volume, and iteratively train the first prediction model based on the loss value. Based on the loss function of the second prediction model, the loss value between the initial prediction result output by the second prediction model and the historical backflow volume is calculated, and the second prediction model is iteratively trained based on the loss value to obtain the backflow volume prediction model.
3. The method according to claim 2, characterized in that, The method further includes: The historical time series feature sample data is input into the LSTM initial model and the GRU initial model to obtain the first prediction result output by the LSTM initial model and the second prediction result output by the GRU initial model. The historical non-time-series feature sample data is input into the XGBoost initial model to obtain the third prediction result output by the XGBoost initial model.
4. The method according to claim 3, characterized in that, Based on the loss function of the first prediction model, the loss value between the initial prediction result output by the first prediction model and the historical backflow volume is calculated, and the first prediction model is iteratively trained based on the loss value, including: The first loss function is used to calculate the loss value of the first prediction result output by the LSTM initial model and the historical backflow volume, and the loss value of the second prediction result output by the GRU initial model and the historical backflow volume, respectively. If the loss value is greater than or equal to the preset threshold, iterate through the number of hidden layers, nodes, and iterations of the initial LSTM and GRU models until the loss value converges to the preset threshold.
5. The method according to claim 4, characterized in that, Based on the loss function of the second prediction model, the loss value between the initial prediction result output by the second prediction model and the historical backflow volume is calculated, and the second prediction model is iteratively trained based on the loss value to obtain the backflow volume prediction model, which includes: The loss value of the third prediction result output by the XGBoost initial model and the historical backflow volume is calculated based on the second loss function; If the loss value is greater than or equal to a preset threshold, the number of hidden layers, nodes, and iterations of the initial XGBoost model are iterated and optimized until the loss value converges to the preset threshold.
6. The method according to claim 5, characterized in that, The historical time-series feature sample data is determined in the following way: Based on the collected pressure, temperature, and flow data for historical periods, determine the changes in pressure, temperature, and flow rate during those periods. The pressure, temperature, and flow rate changes over historical periods are used to form a historical time-series feature sequence, which is the historical time-series feature sample data.
7. The method according to claim 6, characterized in that, The first and second predicted backflow volumes are weighted and summed to obtain the final predicted backflow volume, which includes: The model trained from the initial LSTM model is assigned a first weight, the model trained from the initial GRU model is assigned a second weight, and the model trained from the initial XGBoost model is assigned a third weight. The final predicted value of the backflow volume is determined based on the first weight, the second weight, the third weight, and the predicted values output by each initial model.
8. A time-series prediction device for predicting shale oil fracturing fluid flowback, characterized in that, An input unit is used to input monitoring data into a backflow volume prediction model; wherein, the backflow volume prediction model includes: a first prediction model and a second prediction model, the first prediction model is obtained by training a time-series feature algorithm with historical backflow volume and historical backflow volume, and the second prediction model is obtained by training a non-time-series feature algorithm with historical non-time-series feature sample data and historical backflow volume; The final backflow volume prediction unit is used to assign weights to the first backflow volume prediction value and the second backflow volume prediction value output by the first prediction model and the second prediction model in the backflow volume prediction model, respectively, and sum them to obtain the final backflow volume prediction value.
9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method according to any one of claims 1 to 7.