Oil field geological parameter intelligent prediction system based on machine learning

Abstract

The invention relates to the technical field of oil field geological prediction, in particular to an oil field geological parameter intelligent prediction system based on machine learning, which acquires static geological data and environmental data in real time through a data acquisition module, and performs noise reduction, normalization and cleaning on the acquired data through a data preprocessing module, so as to obtain an oil field geological parameter prediction result. The feature extraction module adopts an XGBoost model and a long-short-term memory network model to process static features and time sequence data respectively. And the model optimization and self-adaption module dynamically adjusts model parameters according to real-time feedback data by using a reinforcement learning algorithm so as to ensure the accuracy and adaptability of a prediction result. The visualization module provides visual display of a prediction trend, a geological data prediction value and a risk indicator, and helps an operator to make a decision based on data driving. According to the method, the prediction precision of the oil field geological data can be effectively improved, the adaptability to environmental changes is enhanced, and a scientific basis is provided for oil field management and production optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of oilfield geological prediction, and in particular to an intelligent prediction system for oilfield geological parameters based on machine learning. Background Art

[0002] As oilfield development deepens, accurate prediction of reservoir geological data becomes increasingly important, which not only helps to improve oil recovery efficiency, but also effectively prolongs the life cycle of oilfields. The dynamic changes of reservoir geological data are affected by many factors, including geological conditions, production activities and environmental changes. Therefore, real-time monitoring and prediction of geological data are essential for optimizing oilfield production management.

[0003] However, existing technologies still face several key problems in application. First, it is difficult for a single model to process both static feature data and time series data at the same time, resulting in deviations in the prediction results and a lack of comprehensiveness. Second, traditional prediction methods lack the ability to dynamically optimize model parameters and cannot adapt to real-time environmental changes in oil fields in a timely manner, thus reducing the accuracy and applicability of predictions. Summary of the invention

[0004] Based on the above objectives, the present invention provides an intelligent prediction system for oilfield geological parameters based on machine learning.

[0005] The intelligent prediction system of oilfield geological parameters based on machine learning includes data acquisition module, data preprocessing module, feature extraction module, prediction modeling module, model optimization and adaptation module and visualization module, among which;

[0006] The data acquisition module collects geological data and environmental data through sensors arranged in the oil field, the geological data including rock porosity, permeability, density and oil saturation, and the environmental data including ambient temperature, air pressure, humidity, precipitation, wind speed and wind direction;

[0007] The data preprocessing module preprocesses the collected geological data and environmental data using noise reduction, data normalization and data cleaning algorithms;

[0008] The feature extraction module uses a machine learning algorithm to extract key features that affect geological parameters, including temperature, pressure, rock porosity, permeability, and oil saturation, from preprocessed geological data and environmental data;

[0009] The prediction modeling module establishes a geological data prediction model in parallel through the XGBoost model and the long short-term memory network model. The XGBoost model is used to process static features, and the long short-term memory network model is used to process time series features. The prediction results of the XGBoost model and the long short-term memory network model are fused by weighted average to generate the final geological data prediction value, so as to help oilfield managers make scientific decisions in oilfield production activities, optimize resource utilization and improve production efficiency;

[0010] The model optimization and adaptation module combines the real-time feedback data of field operations and uses the reinforcement learning algorithm to dynamically adjust the parameters of the XGBoost model and the long short-term memory network model;

[0011] The visualization module provides a visualization interface for displaying prediction trends, geological data prediction values ​​and risk indicators, helping operators make scientific decisions based on a data-driven approach.

[0012] Optionally, the data acquisition module specifically includes:

[0013] Sensor network deployment: A sensor network is deployed in the oil field to collect geological and environmental data, including rock porosity, permeability, density and oil saturation sensors, as well as ambient temperature, air pressure, humidity, precipitation, wind speed and wind direction sensors;

[0014] Geological data collection: rock porosity sensors are used to measure the porosity data of rock formations to reflect the oil storage capacity of oil reservoirs. Permeability sensors collect rock formation permeability data to evaluate the fluidity of fluids in rocks. Density sensors measure rock density to identify the characteristics of different rock formations. Oil saturation sensors are used to detect the oil saturation in the reservoir to provide a basis for oil production prediction.

[0015] Environmental data collection: Ambient temperature sensors are used to measure the ambient temperature of the oil field area and the temperature changes downhole in real time to analyze the impact of temperature on the oil reservoir. Air pressure sensors collect air pressure data and provide pressure information of the surrounding environment to monitor changes in external conditions. Humidity sensors are used to detect humidity data of the oil field environment to help assess the corrosion risk and environmental impact of equipment. Precipitation sensors record precipitation data in the oil field area to facilitate analysis of the impact of moisture on the formation structure. Wind speed and direction sensors are used in oil fields to measure wind speed and direction data in real time.

[0016] Data collection cycle setting: set the frequency of data collection according to the actual needs of the oil field;

[0017] Data transmission and storage: The collected geological data and environmental data are transmitted to the preset data storage unit through a wireless communication network (such as LoRa or 5G) for data storage.

[0018] Optionally, the data preprocessing module includes:

[0019] Noise reduction: noise reduction processing is performed on the collected geological data and environmental data;

[0020] Data normalization: normalize the geological data and environmental data after noise reduction;

[0021] Data cleaning: Clean the normalized geological data and environmental data;

[0022] Data verification: Verify the data after noise reduction, normalization and cleaning to ensure that the data quality meets the requirements of subsequent analysis;

[0023] Data storage: Store the preprocessed and verified geological data and environmental data in the preset data storage unit.

[0024] Optionally, the feature extraction module includes:

[0025] Feature importance evaluation: XGBoost algorithm is used to evaluate the feature importance of preprocessed geological data and environmental data;

[0026] Feature selection: Identify key features that have an impact on geological data prediction based on feature importance assessment;

[0027] Recursive feature elimination: Use recursive feature elimination algorithm to screen the selected key features;

[0028] Feature set generation: The filtered features are used to form the final feature set, and the feature set is stored in a structured manner;

[0029] Feature set verification: Verify the generated feature set.

[0030] Optionally, the prediction modeling module includes:

[0031] Model building: Build an XGBoost model for static feature prediction to capture nonlinear relationships in geological data. Static features include rock porosity, permeability, and oil saturation.

[0032] Model training: Extract static features (such as rock porosity, permeability, and oil saturation) from geological data and input them into the XGBoost model to train the XGBoost model;

[0033] Static feature prediction: Apply the trained XGBoost model to new static features to generate preliminary prediction results for the static features.

[0034] Optionally, the prediction modeling module further includes:

[0035] Model building: Build a long short-term memory network model for time series features to capture the temporal dependency of geological data;

[0036] Model training: extract time series features from environmental data (such as historical series of ambient temperature, humidity, and air pressure) and input them into the constructed long short-term memory network model. Based on the input time series features, train and optimize the long short-term memory network model.

[0037] Time series feature prediction: Apply the trained LSTM network model to new time series features to generate preliminary prediction results for time series features, which are used to predict the dynamic change trend of geological data;

[0038] Fusion of prediction results: The time series feature prediction results of the long short-term memory network model are weighted averaged with the static feature prediction results of the XGBoost model to generate the final geological data prediction value.

[0039] Optionally, the model optimization and adaptation module includes:

[0040] Real-time feedback data collection: Collect real-time feedback data from field operations, including permeability changes, oil saturation, ambient temperature, humidity and moisture;

[0041] Introduction of reinforcement learning algorithm: The reinforcement learning algorithm is used to adaptively optimize the parameters of the XGBoost model;

[0042] Reward function design: Define the reward function based on the difference between the prediction results of the XGBoost model and the real-time feedback data to guide parameter optimization;

[0043] Parameter update and optimization: The reinforcement learning algorithm dynamically adjusts the XGBoost model parameters based on the feedback results of the reward function;

[0044] Result storage and application: The optimized XGBoost model parameters are stored and applied to subsequent geological data prediction tasks.

[0045] Optionally, the model optimization and adaptation module further includes:

[0046] Real-time feedback data collection: collect real-time feedback data from field operations, including geological data and environmental data change information in time series data;

[0047] Introduction of reinforcement learning algorithm: The reinforcement learning algorithm is used to adaptively optimize the parameters of the long short-term memory network model;

[0048] Reward function design: Design a reward function based on the difference between the prediction results of the LSTM network model and the real-time feedback data to optimize the LSTM network parameters;

[0049] Parameter update and optimization: The reinforcement learning algorithm dynamically adjusts the parameters of the long short-term memory network model based on reward feedback;

[0050] Result storage and application: The optimized long short-term memory network model parameters are stored and applied to subsequent time series feature predictions.

[0051] Optionally, the visualization module includes:

[0052] Data display interface construction: Design a visual interface, including multiple data display windows, to display geological data prediction values, prediction trend charts and risk indicators in real time, so that operators can fully understand the reservoir status;

[0053] Prediction trend graph drawing: display the prediction trend of geological data through time series curve;

[0054] Geological data prediction value display: Display the current and future geological data prediction values ​​in the form of tables or icons;

[0055] Risk indicator design: set up risk indicator areas on the visual interface, and provide real-time risk warnings based on abnormal conditions predicted by geological data;

[0056] Historical data comparison function: Provides a comparison chart between historical data and current forecast values, allowing operators to refer to historical data trends and verify the rationality of current forecasts;

[0057] Customizable interface: Allows operators to customize the visualization interface based on oilfield monitoring needs, including selecting the type of geological data displayed and adjusting risk indicator thresholds, enhancing the flexibility and applicability of the interface.

[0058] Beneficial effects of the present invention:

[0059] The present invention realizes the comprehensive collection of static geological data (such as rock porosity, permeability, density and oil saturation) and dynamic environmental data (such as temperature, air pressure, humidity, precipitation, wind speed and wind direction) through the data acquisition module, ensuring the integrity of geological and environmental information. The data preprocessing module further uses noise reduction, data normalization and data cleaning algorithms to efficiently process the collected data, eliminate outliers and optimize data consistency, providing high-quality, noise-free data input for subsequent modeling and prediction, laying the foundation for accurate prediction.

[0060] The present invention greatly improves the accuracy of geological data prediction and the adaptability to environmental changes by combining the prediction modeling module with the model optimization and adaptation module. The prediction modeling module combines the XGBoost model and the long short-term memory network model to fully utilize the advantages of static feature data and time series data to generate comprehensive geological data prediction results. The model optimization and adaptation module uses reinforcement learning algorithms combined with real-time feedback data to dynamically adjust model parameters, ensuring that the prediction model can be continuously optimized as environmental conditions change, providing reliable support for oil field dynamic monitoring and long-term trend prediction.

[0061] The visualization module of the present invention significantly improves the readability of data and the decision-making efficiency of operators by providing real-time prediction trend charts, geological data prediction value displays and risk indicators. The risk indicator prompts potential problems in real time based on a three-level early warning mechanism (low risk, medium risk, and high risk), helping operators to quickly respond and adjust production strategies. At the same time, the system supports historical data comparison and interface customization functions, giving users flexible analysis capabilities, enabling operators to intuitively grasp key data changes in complex oilfield dynamics and scientifically optimize production plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0063] Figure 1 A schematic diagram of a system flow of an embodiment of the present invention;

[0064] Figure 2 Schematic diagram of the feature extraction module flow in an embodiment of the present invention. DETAILED DESCRIPTION

[0065] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0066] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0067] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0068] like Figure 1-Figure 2 As shown, the intelligent prediction system of oilfield geological parameters based on machine learning includes a data acquisition module, a data preprocessing module, a feature extraction module, a prediction modeling module, a model optimization and adaptation module and a visualization module, wherein;

[0069] The data acquisition module collects geological data and environmental data through sensors arranged in the oil field. The geological data includes rock porosity, permeability, density and oil saturation, and the environmental data includes ambient temperature, air pressure, humidity, precipitation, wind speed and wind direction.

[0070] The data preprocessing module preprocesses the collected geological and environmental data using noise reduction, data normalization, and data cleaning algorithms;

[0071] The feature extraction module uses machine learning algorithms to extract key features that affect geological parameters, including temperature, pressure, rock porosity, permeability, and oil saturation, from preprocessed geological and environmental data.

[0072] The prediction modeling module establishes a geological data prediction model in parallel through the XGBoost model and the long short-term memory network model. The XGBoost model is used to process static features, and the long short-term memory network model is used to process time series features. The prediction results of the XGBoost model and the long short-term memory network model are fused through weighted average to generate the final geological data prediction value, which helps oilfield managers make scientific decisions in oilfield production activities, optimize resource utilization and improve production efficiency.

[0073] The model optimization and adaptation module combines real-time feedback data from field operations and uses reinforcement learning algorithms to dynamically adjust the parameters of the XGBoost model and the long short-term memory network model to ensure the accuracy of predictions and enhance adaptability to environmental changes;

[0074] The visualization module provides a visualization interface to display prediction trends, geological data prediction values ​​and risk indicators, helping operators make scientific decisions based on data-driven methods.

[0075] The data acquisition module specifically includes:

[0076] Sensor network deployment: A sensor network is deployed in the oil field to collect geological and environmental data, including rock porosity, permeability, density and oil saturation sensors, as well as ambient temperature, air pressure, humidity, precipitation, wind speed and wind direction sensors;

[0077] Geological data collection: rock porosity sensors are used to measure the porosity data of rock formations to reflect the oil storage capacity of oil reservoirs. Permeability sensors collect rock formation permeability data to evaluate the fluidity of fluids in rocks. Density sensors measure rock density to identify the characteristics of different rock formations. Oil saturation sensors are used to detect the oil saturation in the reservoir to provide a basis for oil production prediction.

[0078] Environmental data collection: Ambient temperature sensors are used to measure the ambient temperature of the oil field area and the temperature changes downhole in real time to analyze the impact of temperature on the oil reservoir. Air pressure sensors collect air pressure data and provide pressure information of the surrounding environment to monitor changes in external conditions. Humidity sensors are used to detect humidity data of the oil field environment to help assess the corrosion risk and environmental impact of equipment. Precipitation sensors record precipitation data in the oil field area to facilitate analysis of the impact of moisture on the formation structure. Wind speed and direction sensors are used in oil fields to measure wind speed and direction data in real time to ensure the safety of oil field platforms and equipment.

[0079] Data collection cycle setting: Set the frequency of data collection according to the actual needs of the oil field to ensure real-time monitoring of oil field environmental changes and meet the timeliness requirements of subsequent analysis modules for data;

[0080] Data transmission and storage: The collected geological data and environmental data are transmitted to the preset data storage unit through a wireless communication network (such as LoRa or 5G) for data storage to facilitate subsequent data preprocessing and analysis;

[0081] Through the reasonable layout of sensor networks and efficient data transmission systems, comprehensive collection of oilfield geological data and environmental data is achieved. Through real-time collection of key data (such as rock porosity, permeability, temperature and humidity, etc.), accurate and timely data input is provided for subsequent data preprocessing, feature extraction and predictive modeling.

[0082] The data preprocessing module includes:

[0083] Noise reduction: Perform noise reduction processing on the collected geological data and environmental data to remove noise data caused by sensor errors or environmental interference. The specific steps include:

[0084] Use wavelet denoising algorithm to filter out high-frequency noise from seismic activity data and retain key frequency components to improve data quality;

[0085] Use moving average filtering on time series data such as temperature and pressure to smooth data fluctuations and reduce the impact of random noise;

[0086] Data normalization: The geological data and environmental data after noise reduction are normalized to ensure the scale consistency between different data types. The specific steps include:

[0087] (1) Use the Min-Max normalization algorithm to normalize temperature, pressure, humidity and other data to the [0,1] interval. The definition formula is:

[0088]

[0089] Among them, x is the original data value, x m in and x m ax is the minimum and maximum value of the data type respectively;

[0090] The normalized data is saved in the preprocessing data set to facilitate subsequent feature extraction and model training;

[0091] Data cleaning: Perform data cleaning on the normalized geological data and environmental data to ensure the integrity and consistency of the data. The specific steps include:

[0092] (1) Detect and eliminate outliers, mark and delete abnormal data that does not meet the preset range;

[0093] (2) Use the mean filling method to fill in the missing data and replace the missing values ​​with the mean of the corresponding data column;

[0094] (3) For consecutive missing data segments, a time interpolation algorithm is used to estimate the missing values ​​to ensure the continuity of the time series data;

[0095] Data verification: Verify the data after noise reduction, normalization and cleaning to ensure that the data quality meets the requirements of subsequent analysis, including:

[0096] (1) Check whether each data type is within the normalization range and confirm that the data meets the preprocessing standards;

[0097] (2) Perform data consistency checks to ensure the temporal and spatial matching of geological data and environmental data;

[0098] Data storage: The preprocessed and verified geological data and environmental data are stored in the preset data storage unit to provide high-quality input data for the subsequent feature extraction module;

[0099] Through steps such as noise reduction, data normalization and data cleaning, the clarity, consistency and integrity of geological and environmental data are ensured, providing high-quality input data for feature extraction and predictive modeling, and improving the overall data reliability and analysis accuracy of the system.

[0100] The feature extraction module includes:

[0101] Feature importance evaluation: The XGBoost algorithm is used to evaluate the feature importance of the preprocessed geological data and environmental data, including:

[0102] The characteristic variables such as temperature, pressure, rock porosity, permeability and oil saturation are input into the XGBoost model for training;

[0103] Based on the characteristics of the gradient boosting tree, the importance score of each feature to the target variable (prediction value of geological data) is calculated to obtain the feature importance ranking;

[0104] Feature selection: Based on feature importance assessment, identify key features that have an impact on geological data prediction. The specific steps include:

[0105] Features with high importance scores were selected, including temperature, pressure, rock porosity, permeability, and oil saturation;

[0106] Combined with the actual situation of the reservoir, the features with less impact in the current environment are eliminated, and the most representative features are retained to simplify the subsequent modeling process;

[0107] Recursive feature elimination: The recursive feature elimination algorithm is used to screen the selected key features, including:

[0108] In each iteration, a low-importance feature is recursively removed based on the feature’s importance score;

[0109] Repeat the iteration until a preset number of features are retained to ensure the most concise and strongest relevance of the input data;

[0110] Feature set generation: The filtered features are used to form the final feature set, and the feature set is stored in a structured manner, including:

[0111] Storing selected temperature, pressure, rock porosity, permeability and oil saturation characteristic data into a preset characteristic library;

[0112] Add timestamps or geolocation tags to each feature for efficient recall during predictive modeling;

[0113] Feature set verification: Verify the generated feature set to ensure its integrity and consistency, including:

[0114] Check whether the feature set is consistent with the preset feature selection criteria and verify the validity of the feature set;

[0115] Ensure that features in geological and environmental data match in time and space to ensure accuracy and stability of feature sets;

[0116] The XGBoost algorithm is used to evaluate feature importance and combined with recursive feature elimination to optimize the feature set, ultimately generating concise and efficient feature data, providing high-quality input for predictive modeling, thereby improving the accuracy and computational efficiency of the predictive model.

[0117] The predictive modeling module includes:

[0118] Model construction: Build an XGBoost model for static feature prediction to capture nonlinear relationships in geological data. Static features include rock porosity, permeability, and oil saturation. The steps to build the XGBoost model include:

[0119] (1) Use the gradient boosting tree method to build the XGBoost model structure;

[0120] (2) Define the loss function of the XGBoost model as mean square error, and the calculation formula is expressed as:

[0121]

[0122] Among them, y i is the actual value, is the predicted value, λ is the regularization parameter, and m is the number of model parameters;

[0123] Model training: Extract static features (such as rock porosity, permeability, and oil saturation) from geological data and input them into the XGBoost model to train the XGBoost model. The steps include:

[0124] (1) Through the static features extracted from geological data (such as rock porosity, permeability, and oil saturation), the XGBoost model can learn the complex relationship between these static features and the target geological parameters, and then identify the characteristics and patterns of geological data;

[0125] (2) During the training process, the XGBoost model will gradually adjust the parameters so that the model's predicted value is closer to the true value, thereby minimizing the loss function (such as mean square error) and improving the model's prediction accuracy;

[0126] (3) By inputting different static features into the training model, XGBoost can generate highly adaptable prediction results, ensuring accurate geological data prediction when new data is input;

[0127] Static feature prediction: Apply the trained XGBoost model to new static features to generate preliminary prediction results for the static features.

[0128] The predictive modeling module also includes:

[0129] Model construction: Build a long short-term memory network model for time series features to capture the temporal dependency of geological data. The specific steps include:

[0130] Design a multi-layer LSTM network structure, where each LSTM network unit contains a forget gate, an input gate, and an output gate to dynamically adjust the flow of information;

[0131] The loss function is defined as mean square error to minimize the error between the predicted value and the actual value and improve the accuracy of time series prediction;

[0132] Model training: Extract time series features from environmental data (such as historical series of ambient temperature, humidity, and air pressure) and input them into the constructed long short-term memory network model. Based on the input time series features, train and optimize the long short-term memory network model. The specific steps include:

[0133] Use the back-propagation algorithm to optimize the weights and biases of the LSTM network to gradually improve the prediction accuracy of the model on time series features;

[0134] The core calculation of the LSTM model includes the following formula:

[0135] f t =σ(W f ·[h t-1 ,x t ]+b f );

[0136] i t =σ(W i·[h t-1 ,x t ]+b i );

[0137] C t =f t *C t-1 +i t *tanh(W C ·[h t-1 ,x t ]+b C );

[0138] o t =σ(W o ·[h t-1 ,x t ]+b o );

[0139] h t =o t *tanh(C t );

[0140] Among them, f t 、i t , C t and t are the calculation results of the forget gate, input gate, cell state and output gate respectively, W is the weight matrix, and b is the bias term;

[0141] Time series feature prediction: Apply the trained long short-term memory network model to new time series features to generate preliminary prediction results for time series features, which are used to predict the dynamic change trend of geological data;

[0142] Fusion of prediction results: The time series feature prediction results of the long short-term memory network model are weighted averaged with the static feature prediction results of the XGBoost model to generate the final geological data prediction value. The specific calculation formula is:

[0143] Final prediction value = w1×XGBoost prediction result + w2×LSTM prediction result;

[0144] Among them, w1 and w2 are the weights of XGBoost and LSTM network models, respectively, which are dynamically adjusted based on the prediction accuracy of the models to ensure the optimal performance of the fusion results;

[0145] By analyzing the dynamic trend of time series data and integrating it with the static feature prediction results of the XGBoost model, the final geological data prediction value is generated. The advantages of static features and time series features are combined to provide support for the comprehensive and accurate prediction of geological data.

[0146] The model optimization and adaptation module includes:

[0147] Real-time feedback data collection: Collect real-time feedback data from field operations, including permeability changes, oil saturation, ambient temperature, humidity and moisture, to provide real-time information for model optimization;

[0148] Introduction of reinforcement learning algorithm: The reinforcement learning algorithm is used to adaptively optimize the parameters of the XGBoost model. The specific steps include:

[0149] (1) Define state variables, including the current parameter settings and real-time feedback data of the XGBoost model, reflecting the current state of the model;

[0150] (2) Define the action space, including adjusting the key parameters of the XGBoost model (such as learning rate, tree depth, etc.) to optimize model performance;

[0151] Reward function design: Define the reward function based on the difference between the prediction results of the XGBoost model and the real-time feedback data to guide parameter optimization, including:

[0152] When the prediction result error is small, a positive reward is given to encourage the current parameter setting;

[0153] When the prediction error is large, a negative reward is given to encourage the algorithm to adjust its parameters;

[0154] Parameter update and optimization: The reinforcement learning algorithm dynamically adjusts the XGBoost model parameters based on the feedback results of the reward function. The specific steps include:

[0155] Adopt Q-learning algorithm to iteratively update the model parameters based on reward feedback;

[0156] Through multiple iterations of optimization, the optimal parameter combination is finally determined to improve the prediction accuracy of the XGBoost model;

[0157] Result storage and application: Store the optimized XGBoost model parameters and apply them to subsequent geological data prediction tasks;

[0158] Dynamic optimization of XGBoost model parameters is achieved through real-time feedback and reinforcement learning algorithms, which improves the accuracy of static feature data prediction and enhances the model's adaptability to environmental changes.

[0159] The model optimization and adaptation module also includes:

[0160] Real-time feedback data collection: Collect real-time feedback data from field operations, including geological data and environmental data change information in time series data, to provide a basis for the optimization of long-term and short-term memory network models;

[0161] (1) The geological data change information in time series data includes:

[0162] Reservoir pressure: The internal pressure of the reservoir changes with the production process. Real-time monitoring of pressure changes helps optimize production strategies;

[0163] Temperature: Changes in reservoir and downhole temperature can reflect fluid status and heat conduction, which is very important for reservoir stability analysis;

[0164] Oil saturation: The relative saturation of oil in a reservoir changes over time, affecting oil recovery efficiency and production prediction;

[0165] Fluid flow rate: monitor the flow rate of fluid in the reservoir to understand the flow law of oil and gas;

[0166] Permeability changes: As production increases, the permeability of different areas of the reservoir will change over time, affecting fluid transport efficiency;

[0167] (2) Environmental data change information in time series data includes:

[0168] Ambient temperature: changes due to factors such as seasons and weather, affecting the heat conduction in the reservoir and underground as well as the operating status of equipment;

[0169] Air pressure: Changes in external atmospheric pressure affect the downhole pressure balance, thereby indirectly affecting the behavior of reservoir fluids;

[0170] Humidity: Especially in onshore oil fields, changes in ambient humidity may affect the corrosion of equipment and have a certain impact on the production process;

[0171] Precipitation: Changes in precipitation affect surface and groundwater levels, and may affect downhole pressure, especially in shallow oil fields;

[0172] Introduction of reinforcement learning algorithm: The reinforcement learning algorithm is used to adaptively optimize the parameters of the long short-term memory network model. The steps include:

[0173] Define state variables, including the current parameter configuration and real-time feedback data of the LSTM network model to reflect the current state of the model;

[0174] Define the action space, including adjusting the learning rate, time step, and number of hidden units in the LSTM network model to improve model performance;

[0175] Reward function design: Design a reward function based on the difference between the prediction results of the LSTM network model and the real-time feedback data to optimize the LSTM network parameters, including:

[0176] When the prediction error is small, give positive rewards and keep the current parameter settings;

[0177] When the prediction error is large, a negative reward is given to guide the algorithm to adjust parameters;

[0178] Parameter update and optimization: The reinforcement learning algorithm dynamically adjusts the parameters of the long short-term memory network model based on reward feedback. The specific steps include:

[0179] Use a deep Q-network or other reinforcement learning method to update the parameter settings of the LSTM model through multiple rounds of iterations;

[0180] According to the changes in reward feedback, it gradually converges to the optimal parameter combination to achieve the optimal prediction performance of the LSTM model on time series data;

[0181] Result storage and application: The optimized LSTM model parameters are stored and applied to subsequent time series feature prediction to ensure the real-time adaptability of the model to environmental changes;

[0182] By combining real-time feedback and reinforcement learning algorithms, dynamic optimization of the long short-term memory network model is achieved, the accuracy of time series data prediction is enhanced, and the model's response and adaptability to environmental changes is improved.

[0183] The visualization modules include:

[0184] Data display interface construction: Design a visual interface, including multiple data display windows, to display geological data prediction values, prediction trend charts and risk indicators in real time, so that operators can fully understand the reservoir status;

[0185] Prediction trend graph drawing: Display the prediction trend of geological data through time series curves, including:

[0186] Use line graphs or bar graphs to show the changing trends of geological data (such as pressure, temperature and oil saturation);

[0187] Support users to select a specific time period to zoom in and observe changes in a specific period;

[0188] Geological data prediction value display: Display the current and future geological data prediction values ​​in the form of tables or icons, including:

[0189] The predicted values ​​of each geological data are numerically labeled for easy visual viewing;

[0190] Combined with the difference between the actual value and the predicted value, color coding is provided to intuitively display the size of the prediction error;

[0191] Risk indicator design: Set up risk indicator areas on the visual interface, and provide real-time risk warnings based on abnormal conditions predicted by geological data, including:

[0192] Based on the set safety threshold, the risk level is indicated by color or icon (such as a red warning light);

[0193] Implement a three-level early warning mechanism (low risk, medium risk, high risk) to facilitate operators to respond to potential risks in a timely manner;

[0194] Historical data comparison function: Provides a comparison chart of historical data and current forecast values, allowing operators to refer to historical data trends and verify the rationality of current forecasts, including:

[0195] Displays an overlay of historical geological data and current forecast trends to analyze the accuracy of forecast models;

[0196] Support selection of different historical time periods for multi-angle comparative analysis;

[0197] Customizable interface: Allows operators to customize the visualization interface based on oilfield monitoring needs, including selecting the type of geological data displayed and adjusting risk indicator thresholds, enhancing the flexibility and applicability of the interface.

[0198] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0199] The above are only preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. Intelligent prediction system for oilfield geological parameters based on machine learning, characterized by: It includes data acquisition module, data preprocessing module, feature extraction module, prediction modeling module, model optimization and adaptation module and visualization module, among which; The data acquisition module collects geological data and environmental data through sensors arranged in the oil field, the geological data including rock porosity, permeability, density and oil saturation, and the environmental data including ambient temperature, air pressure, humidity, precipitation, wind speed and wind direction; The data preprocessing module preprocesses the collected geological data and environmental data using noise reduction, data normalization and data cleaning algorithms; The feature extraction module uses a machine learning algorithm to extract key features that affect geological parameters, including temperature, pressure, rock porosity, permeability, and oil saturation, from preprocessed geological data and environmental data; The prediction modeling module establishes a geological data prediction model in parallel through the XGBoost model and the long short-term memory network model, the XGBoost model is used to process static features, and the long short-term memory network model is used to process time series features, and the prediction results of the XGBoost model and the long short-term memory network model are fused by weighted average to generate the final geological data prediction value; The model optimization and adaptation module combines the real-time feedback data of field operations and uses the reinforcement learning algorithm to dynamically adjust the parameters of the XGBoost model and the long short-term memory network model; The visualization module provides a visualization interface for displaying prediction trends, geological data prediction values ​​and risk indicators, helping operators make scientific decisions based on a data-driven approach.

2. The intelligent prediction system for oilfield geological parameters based on machine learning according to claim 1 is characterized in that: The data acquisition module specifically includes: Sensor network deployment: A sensor network is deployed in the oil field to collect geological and environmental data, including rock porosity, permeability, density and oil saturation sensors, as well as ambient temperature, air pressure, humidity, precipitation, wind speed and wind direction sensors; Geological data collection: rock porosity sensors are used to measure the porosity data of rock formations, permeability sensors collect rock permeability data, density sensors measure rock density, and oil saturation sensors are used to detect the oil saturation in the reservoir; Environmental data collection: The ambient temperature sensor is used to measure the ambient temperature of the oil field area and the temperature changes downhole in real time. The air pressure sensor collects air pressure data. The humidity sensor is used to detect the humidity data of the oil field environment. The precipitation sensor records the precipitation data of the oil field area. The wind speed and wind direction sensor is used in the oil field to measure the wind speed and wind direction data in real time. Data collection cycle setting: set the frequency of data collection according to the actual needs of the oil field; Data transmission and storage: The collected geological data and environmental data are transmitted to the preset data storage unit through the wireless communication network for data storage.

3. The intelligent prediction system for oilfield geological parameters based on machine learning according to claim 2 is characterized in that: The data preprocessing module comprises: Noise reduction: noise reduction processing is performed on the collected geological data and environmental data; Data normalization: normalize the geological data and environmental data after noise reduction; Data cleaning: Clean the normalized geological data and environmental data; Data verification: Verify the data after noise reduction, normalization and cleaning to ensure that the data quality meets the requirements of subsequent analysis; Data storage: Store the preprocessed and verified geological data and environmental data in the preset data storage unit.

4. The intelligent prediction system for oilfield geological parameters based on machine learning according to claim 3 is characterized in that: The feature extraction module comprises: Feature importance evaluation: XGBoost algorithm is used to evaluate the feature importance of preprocessed geological data and environmental data; Feature selection: Identify key features that have an impact on geological data prediction based on feature importance assessment; Recursive feature elimination: Use recursive feature elimination algorithm to screen the selected key features; Feature set generation: The filtered features are used to form the final feature set, and the feature set is stored in a structured manner; Feature set verification: Verify the generated feature set.

5. The intelligent prediction system for oilfield geological parameters based on machine learning according to claim 4 is characterized in that: The prediction modeling module includes: Model building: Build an XGBoost model for static feature prediction to capture nonlinear relationships in geological data. Static features include rock porosity, permeability, and oil saturation. Model training: Extract static features from geological data and input them into the XGBoost model to train the XGBoost model; Static feature prediction: Apply the trained XGBoost model to new static features to generate preliminary prediction results for the static features.

6. The intelligent prediction system for oilfield geological parameters based on machine learning according to claim 5 is characterized in that: The predictive modeling module also includes: Model building: Build a long short-term memory network model for time series features to capture the temporal dependency of geological data; Model training: extract time series features from environmental data and input them into the constructed long short-term memory network model. Based on the input time series features, train and optimize the long short-term memory network model. Time series feature prediction: Apply the trained long short-term memory network model to new time series features to generate preliminary prediction results for time series features, which are used to predict the dynamic change trend of geological data; Fusion of prediction results: The time series feature prediction results of the long short-term memory network model are weighted averaged with the static feature prediction results of the XGBoost model to generate the final geological data prediction value.

7. The intelligent prediction system for oilfield geological parameters based on machine learning according to claim 6 is characterized in that: The model optimization and adaptation module includes: Real-time feedback data collection: Collect real-time feedback data from field operations, including permeability changes, oil saturation, ambient temperature, humidity and moisture; Introduction of reinforcement learning algorithm: The reinforcement learning algorithm is used to adaptively optimize the parameters of the XGBoost model; Reward function design: Define the reward function based on the difference between the prediction results of the XGBoost model and the real-time feedback data to guide parameter optimization; Parameter update and optimization: The reinforcement learning algorithm dynamically adjusts the XGBoost model parameters based on the feedback results of the reward function; Result storage and application: The optimized XGBoost model parameters are stored and applied to subsequent geological data prediction tasks.

8. The intelligent prediction system for oilfield geological parameters based on machine learning according to claim 7 is characterized in that: The model optimization and adaptation module also includes: Real-time feedback data collection: collect real-time feedback data from field operations, including geological data and environmental data change information in time series data; Introduction of reinforcement learning algorithm: The reinforcement learning algorithm is used to adaptively optimize the parameters of the long short-term memory network model; Reward function design: Design a reward function based on the difference between the prediction results of the LSTM network model and the real-time feedback data to optimize the LSTM network parameters; Parameter update and optimization: The reinforcement learning algorithm dynamically adjusts the parameters of the long short-term memory network model based on reward feedback; Result storage and application: The optimized long short-term memory network model parameters are stored and applied to subsequent time series feature predictions.

9. The intelligent prediction system for oilfield geological parameters based on machine learning according to claim 8, characterized in that: The visualization module comprises: Data display interface construction: Design a visual interface, including multiple data display windows, to display geological data prediction values, prediction trend charts and risk indicators in real time, so that operators can fully understand the reservoir status; Prediction trend graph drawing: display the prediction trend of geological data through time series curve; Geological data prediction value display: Display the current and future geological data prediction values ​​in the form of tables or icons; Risk indicator design: set up risk indicator areas on the visual interface, and provide real-time risk warnings based on abnormal conditions predicted by geological data; Historical data comparison function: Provides a comparison chart between historical data and current forecast values, allowing operators to refer to historical data trends and verify the rationality of current forecasts; Customizable interface: Allows operators to customize the visualization interface based on their field monitoring needs, including selecting the type of geological data displayed and adjusting risk indicator thresholds.

Images