Oil and gas well dynamic development data prediction method and application thereof

Abstract

The application discloses an oil and gas well dynamic development data prediction method and application thereof, and relates to the technical field of oil and gas field exploration and development. The oil and gas well dynamic development data prediction method disclosed by the application comprises the following steps: S1, collecting and preprocessing oil and gas well dynamic development data; S2, constructing a neural network model based on a BP algorithm; S3, training the neural network model through the BP algorithm; and S4, inputting the oil and gas well dynamic development data to be predicted into the neural network model trained in step S3 to obtain a prediction result. The method improves the precision and accuracy of oil and gas well dynamic development data prediction and reduces prediction errors by preprocessing original data and training a BP neural network model.

Description

Technical Field

[0001] This invention relates to the field of oil and gas field exploration and development technology, and in particular to a method for predicting dynamic development data of oil and gas wells and its application. Background Technology

[0002] Artificial Neural Networks (ANNs) are information processing technologies similar to the human nervous system. They can be viewed as powerful and widely applicable machine learning algorithms, exhibiting strong intelligence in resolving fuzzy relationships and making nonlinear predictions. By training and learning from the input and output samples of a system using a well-designed neural network, ANNs can approximate arbitrarily complex nonlinear systems with appropriate accuracy. This superior performance makes ANNs a universal mathematical model for multidimensional nonlinear functions. In recent years, the development of ANN theory, particularly the emergence of the backpropagation (BP) algorithm, has made ANNs an effective method for analyzing and predicting the performance of complex nonlinear systems. Currently, ANNs are widely used in research fields such as pattern recognition, signal processing, optimization computation, parameter matching and prediction, and parameter spatial distribution prediction. In recent years, ANN technology has been introduced into the petroleum industry, effectively solving nonlinear problems in production processes. Its wide range of applications has yielded good results in reservoir parameter prediction, drilling downhole complexities and accident diagnosis and prediction, reservoir fracture identification, gas well management, and dynamic prediction, with continuous theoretical development and improvement.

[0003] Chinese patent CN113236228A discloses a method for predicting production using an LSTM model, specifically including: 1) acquiring the static and dynamic parameters of each oil well and constructing corresponding LSTM models; 2) performing sensitivity analysis on the network calculation parameters of each LSTM model; 3) establishing the correlation between the daily oil production fluctuation and the optimal network calculation parameters, forming a graph showing the relationship between the daily oil production fluctuation and the optimal network calculation parameters; 4) determining the optimal LSTM model; 5) acquiring the static and dynamic parameters of the oil well to be predicted and inputting them into the optimal LSTM model to obtain the predicted production value of the oil well. This method relies on complete historical data of oil and gas field development and requires both static data and dynamic parameters to achieve oil well production prediction. It has limitations in predicting dynamic production of new wells with only dynamic data.

[0004] Chinese patent CN113722997A discloses a method for dynamic production prediction of new wells based on static oil and gas field data. The method includes: summarizing and preprocessing oil and gas well data; training a BP neural network cumulative production prediction model and optimizing its hyperparameters; training a BP-LSTM monthly production prediction model and optimizing its hyperparameters; and obtaining the predicted monthly production data through inverse normalization. This model does not consider issues such as missing data and outliers.

[0005] In view of this, in order to overcome the shortcomings of the existing technology, the present invention provides a method for predicting dynamic development data of oil and gas wells and its application. Summary of the Invention

[0006] The purpose of this invention is to provide a method for predicting dynamic development data of oil and gas wells and its application, which can accurately predict dynamic development data of oil and gas wells and better capture the complex relationships and changes in the dynamic development process of oil and gas wells.

[0007] To achieve the above-mentioned objectives, the technical solution of this invention is as follows:

[0008] On the one hand, the present invention provides a method for predicting dynamic development data of oil and gas wells, comprising the following steps:

[0009] S1. Collect and preprocess dynamic development data of oil and gas wells;

[0010] S2. Construct a neural network model based on the BP algorithm:

[0011] S21. Design a neural network, which includes an input layer, a hidden layer, and an output layer;

[0012] S22. Initialize the weights and thresholds of the neural network;

[0013] S3. Train the neural network model using the BP algorithm:

[0014] The neural network model is trained using the preprocessed oil and gas well dynamic development data from step S1, and the weights and thresholds of the neural network are adjusted using the BP algorithm.

[0015] S4. Input the dynamic development data of the oil and gas wells to be predicted into the neural network model trained in step S3 to obtain the prediction results.

[0016] Preferably, the method for predicting dynamic development data of oil and gas wells further includes the following steps:

[0017] S5. Make optimization decisions on the development and management of oil and gas fields based on the prediction results obtained in step S4.

[0018] S6. Regularly update and optimize the neural network model.

[0019] Preferably, in step S1, the dynamic development data of the oil and gas well includes oil well production, pressure, temperature, gas-oil ratio, water cut, daily production, cumulative production, pump diameter, nozzle, stroke, stroke frequency, discharge rate, oil pressure, casing pressure, back pressure, total fluid volume, oil volume, water volume, gas volume, oil-gas ratio, pressure measurement, fluid level, and liquid mixing.

[0020] Preferably, in step S1, the data preprocessing includes data cleaning, data transformation, data normalization, data feature extraction, data partitioning, data standardization, and data randomization.

[0021] More preferably, in step S1, the data cleaning includes missing value processing, outlier processing, and data format standardization.

[0022] Preferably, in step S1, the data conversion includes converting the raw data into a form suitable for neural network processing and converting the time series data into a form suitable for time series analysis.

[0023] Preferably, in step S1, the data normalization refers to normalizing data of different dimensions and scales using the sigmoid activation function to make them into a uniform form, so as to better train and predict the model. The linear normalization formula is as follows:

[0024]

[0025] In the formula, x is the input of the neuron, e is the base of the natural logarithm (approximately 2.71828), and δ(x) is the normalized data.

[0026] Preferably, in step S1, the data feature extraction refers to extracting meaningful features from the original data. These features can reflect the changing patterns and trends of dynamic development of oil and gas wells. Values ​​that are not closely related to the data features are deleted, and the deleted data is normalized.

[0027] Preferably, in step S1, the data partitioning refers to dividing the data into a training set, a validation set, and a test set.

[0028] Preferably, in step S1, data standardization refers to standardizing the data so that the mean becomes 0 and the variance becomes 1.

[0029] Preferably, in step S1, data randomization refers to randomizing the data.

[0030] Preferably, in step S21, the number of input layer nodes is the same as the number of features in the dynamic development data of oil and gas wells, and the number of output layer nodes is the predicted target value. Those skilled in the art can adjust and optimize the number of hidden layer nodes and the number of layers according to the actual situation.

[0031] Preferably, in step S3, the BP algorithm includes batch training and stochastic gradient descent.

[0032] On the other hand, the present invention provides the application of the above-mentioned method for predicting dynamic development data of oil and gas wells in the process of oil and gas field exploration or development.

[0033] The beneficial effects of this invention are as follows:

[0034] This invention discloses a novel method for predicting dynamic development data of oil and gas wells. This method improves the accuracy and precision of predicting dynamic development data of oil and gas wells and reduces prediction errors by preprocessing raw data and training a BP neural network model. Attached Figure Description

[0035] Figure 1 This is a flowchart of the method for predicting dynamic development data of oil and gas wells in an embodiment of the present invention.

[0036] Figure 2 This is a schematic diagram of the distribution of the sigmoid activation function in the dynamic development prediction of oil and gas wells in an embodiment of the present invention.

[0037] Figure 3 This is a neural network structure diagram for dynamic development prediction of oil and gas wells in an embodiment of the present invention.

[0038] Figure 4 This is a graph showing the change in training error of the BP neural network model in an embodiment of the present invention. Detailed Implementation

[0039] The following non-limiting embodiments are intended to enable those skilled in the art to gain a more comprehensive understanding of the present invention, but do not limit the invention in any way. The following content is merely an exemplary description of the scope of protection claimed by the present invention, and those skilled in the art can make various changes and modifications to the invention based on the disclosed content, which should also fall within the scope of protection claimed in this application.

[0040] Example

[0041] A method for predicting dynamic development data of oil and gas wells includes the following steps:

[0042] Step S1: Preprocessing of dynamic development data for oil and gas wells, including data cleaning, data transformation, data normalization, data feature extraction, data partitioning, data standardization, and data randomization;

[0043] Step S11 Data Collection: Collect dynamic development data of oil and gas wells, including oil well production, pressure, temperature, gas-oil ratio, water cut, daily production, cumulative production, pump diameter, nozzle, stroke, stroke frequency, discharge rate, oil pressure, casing pressure, back pressure, total fluid volume, oil volume, water volume, gas volume, oil-gas ratio, pressure measurement, fluid level, and fluid mixing.

[0044] Table 1 shows the collected dynamic development data of oil and gas wells (taking partial data from well SHB5-12H in the Shunbei block as an example).

[0045]

[0046]

[0047]

[0048]

[0049]

[0050]

[0051]

[0052]

[0053]

[0054]

[0055]

[0056]

[0057]

[0058]

[0059]

[0060]

[0061] Step S12 Data Cleaning: The data cleaning refers to cleaning the collected dynamic development data of oil and gas wells. The cleaning includes handling missing values, handling outliers, and unifying data formats.

[0062] Table 2 Dynamic development data of oil and gas wells after cleaning (taking partial data from well SHB5-12H in the Shunbei block as an example)

[0063]

[0064]

[0065]

[0066]

[0067] Step S13 Data Transformation: The data transformation includes converting the raw data into a form suitable for neural network processing and converting the time series data into a form suitable for time series analysis;

[0068] Step S14 Data Normalization: Data normalization refers to normalizing data of different dimensions and scales using the sigmoid activation function to make them into a uniform form, so as to better facilitate model training and prediction (e.g., ...). Figure 2 The linear normalization formula is as follows:

[0069]

[0070] In the formula, x is the input of the neuron, e is the base of the natural logarithm (approximately 2.71828), and δ(x) is the normalized data.

[0071] The sigmoid activation function can compress the output of x∈R to the interval x∈(0,1). The sigmoid activation function tends to a saturation value in most of its domain. When x takes a very large positive value, the sigmoid activation function will saturate to a high value (infinitely close to 1); when x takes a very large negative value, the sigmoid activation function will saturate to a low value (infinitely close to 0).

[0072] Step S14 Data Feature Extraction: The data feature extraction refers to extracting meaningful features from the raw data. These features can reflect the changing patterns and trends of dynamic development of oil and gas wells. Values ​​that are not closely related to the data features are deleted, and the deleted data is normalized.

[0073] Step S15 Data partitioning: The data partitioning refers to dividing the data into training set, validation set and test set in order to better evaluate the model's predictive performance and generalization ability;

[0074] Step S16 Data Standardization: Data standardization refers to standardizing the data so that the mean becomes 0 and the variance becomes 1, which makes the training of the neural network more stable and reliable.

[0075] Step S17 Data Randomization: Data randomization refers to randomizing the data to avoid the impact of data order and other factors on model training. The final training data (training dataset is shown in Table 3) is as follows.

[0076] Table 3 Training Dataset

[0077]

[0078]

[0079] Step S2: Construct the neural network model:

[0080] Based on the characteristics and problem requirements of dynamic development data of oil and gas wells, a multi-layer neural network model is constructed (e.g., Figure 3 ).

[0081] The neural network model includes an input layer, a hidden layer, and an output layer.

[0082] The number of input layer nodes is the same as the number of features in the dynamic development data of oil and gas wells, and the number of output layer nodes is the predicted target value. Those skilled in the art can adjust and optimize the number of hidden layer nodes and layers according to the actual situation.

[0083] The input layer receives preprocessed dynamic development data of oil and gas wells, the hidden layer transforms the input into meaningful feature representations through a series of complex calculations, and the output layer transforms the results of the hidden layer into specific predicted values.

[0084] Step S21: Initialize the weights and thresholds of the neural network: Using random initialization or pre-training methods, assign initial values ​​to the weights and thresholds of the neural network. The weights and thresholds are continuously adjusted according to the loss function and optimization algorithm to minimize the prediction error of the neural network (e.g., ...). Figure 4 ).

[0085] Step S3: Train the neural network model using the BP algorithm:

[0086] The neural network model constructed in step S2 is trained using the dynamic development data of oil and gas wells obtained in step S1. The weights and thresholds of the neural network are adjusted by the BP algorithm so that the neural network model can better fit the actual data.

[0087] The BP algorithm includes batch training and stochastic gradient descent.

[0088] Step S31: Evaluate the performance of the neural network model: Use metrics such as cross-validation and mean squared error to evaluate the trained neural network model to verify its accuracy and generalization ability. The superiority of the method can also be evaluated by comparing it with other prediction methods.

[0089] Step S4: Predict dynamic development data of oil and gas wells:

[0090] The dynamic development data of the oil and gas wells to be predicted are input into the BP neural network model obtained in step S3 to obtain the prediction results (as shown in Table 4).

[0091] Table 4 Prediction Results of Examples

[0092] Serial Number Actual value (m) Predicted value (m) The difference between the actual value and the predicted value (m) 0 0.041336617 -0.063160369 0.104496985 1 0.834220714 0.669984051 0.164236662 2 0.359827183 0.25222233 0.107604853 3 0.219665458 0.21338895 0.006276508 4 0.42451721 0.407618524 0.016898686 5 0.281660067 0.222560888 0.059099179

[0093] Step S5: Application of the prediction results:

[0094] Based on the prediction results obtained in step S4, optimization decisions are made for the development and management of oil and gas fields, including formulating reasonable development plans, adjusting production parameters, and optimizing resource allocation, in order to improve the output and efficiency of oil and gas fields.

[0095] Preferred:

[0096] Regularly update and optimize neural network models: Based on actual conditions and new data, regularly update and optimize neural network models to adapt to new needs and challenges in oil and gas field development and management. Methods such as using new training data, adjusting network structure, and optimizing hyperparameters can be used to improve model performance.

[0097] Comparative Example

[0098] Step 1: Collect dynamic development data of oil and gas wells. The collected dynamic development data of oil and gas wells is the same as in the example.

[0099] Step 2: Transform the raw data into a format suitable for neural network processing;

[0100] Step 3: Construct the neural network model, the construction process is the same as in the example;

[0101] Step 4: Train the neural network model using the BP algorithm. The training process is the same as in the example.

[0102] Step 5: Predict dynamic development data of oil and gas wells:

[0103] The dynamic development data of the oil and gas wells to be predicted are input into the trained neural network model to obtain the prediction results (as shown in Table 5).

[0104] Table 5 shows the prediction results for the comparative examples.

[0105] Serial Number Actual value (m) Predicted value (m) The difference between the actual value and the predicted value (m) 0 0.041336617 -0.263160369 0.304496986 1 0.834220714 0.469984051 0.364236663 2 0.359827183 -0.15222233 0.512049513 3 0.219665458 0.01338895 0.206276508 4 0.42451721 0.107618524 0.316898686 5 0.281660067 -0.122560888 0.404220955

[0106] Compared to the examples, the comparative model omits the data preprocessing step, resulting in decreased prediction accuracy, weakened model generalization ability, and unstable prediction results.

[0107] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting dynamic development data of oil and gas wells, characterized in that, Includes the following steps: S1. Collect and preprocess dynamic development data of oil and gas wells; S2. Construct a neural network model based on the BP algorithm: S21. Design a neural network, which includes an input layer, a hidden layer, and an output layer; S22. Initialize the weights and thresholds of the neural network; S3. Train the neural network model using the BP algorithm: The neural network model is trained using the preprocessed oil and gas well dynamic development data from step S1, and the weights and thresholds of the neural network are adjusted using the BP algorithm. S4. Input the dynamic development data of the oil and gas wells to be predicted into the neural network model trained in step S3 to obtain the prediction results.

2. The method for predicting dynamic development data of oil and gas wells according to claim 1, characterized in that, It also includes the following steps: S5. Make optimization decisions on the development and management of oil and gas fields based on the prediction results obtained in step S4. S6. Regularly update and optimize the neural network model.

3. The method for predicting dynamic development data of oil and gas wells according to claim 1, characterized in that, In step S1, the dynamic development data of the oil and gas well includes oil well production, pressure, temperature, gas-oil ratio, water cut, daily production, cumulative production, pump diameter, nozzle, stroke, stroke frequency, discharge rate, oil pressure, casing pressure, back pressure, total fluid volume, oil volume, water volume, gas volume, oil-gas ratio, pressure measurement, fluid level, and liquid mixing.

4. The method for predicting dynamic development data of oil and gas wells according to claim 1, characterized in that, In step S1, the data preprocessing includes data cleaning, data transformation, data normalization, data feature extraction, data partitioning, data standardization, and data randomization.

5. The method for predicting dynamic development data of oil and gas wells according to claim 4, characterized in that, In step S1, the data cleaning includes missing value processing, outlier processing, and data format standardization.

6. The method for predicting dynamic development data of oil and gas wells according to claim 4, characterized in that, Data normalization refers to the process of normalizing data of different dimensions and scales. The sigmoid activation function is used to transform data into a uniform form, which facilitates better model training and prediction. The linear normalization formula is as follows: In the formula, x is the input of the neuron, e is the base of the natural logarithm, and δ(x) is the normalized data.

7. The method for predicting dynamic development data of oil and gas wells according to claim 4, characterized in that, In step S1, the data partitioning refers to dividing the data into a training set, a validation set, and a test set.

8. The method for predicting dynamic development data of oil and gas wells according to claim 4, characterized in that, In step S1, data standardization means changing the mean of the data to 0 and the variance to 1.

9. The method for predicting dynamic development data of oil and gas wells according to claim 1, characterized in that, In step S21, the number of input layer nodes is the same as the number of features in the dynamic development data of oil and gas wells, and the number of output layer nodes is the predicted target value.

10. The method for predicting dynamic development data of oil and gas wells according to claim 1, characterized in that, In step S3, the BP algorithm includes batch training and stochastic gradient descent.

11. The application of the oil and gas well dynamic development data prediction method according to any one of claims 1-10 in the process of oil and gas field exploration or oil and gas field development.

Images