Proceedings of the International Field Exploration and Development Conference 2023

Green Sustainable Development of Oil and Gas Fields considers oil and gas exploration as well as environmental conservation. Currently, national ecological protection requires that oil and gas exploration wells be shut down and closed when they withdraw, and that oil and gas facilities be demolished to restore the surrounding geomorphology and ecology. As a result, the condition of ecological restoration of oil and gas sites withdrawal is an essential component to evaluate the ecological protection of oil and gas fields. In this article, multi-temporal high-resolution satellite remote sensing big data is employed to achieve intelligent monitoring and assessment of green recovery at oil and gas sites. The technical process of remote sensing intelligent monitoring of oil and gas well withdrawal and ecological restoration includes three steps: 1. Determine the different types of well sites. Identify well sites where oil and gas facilities depart using high-resolution remote sensing oil and gas well site interpretation markings; 2. Detect well site modification. 3. Evaluate well site recovery by using the GRNDVI vegetation growth index into the well site vegetation change over time. To evaluate well site recovery, set a threshold value based on change detection data. Using Dabusu in Jilin Oilfield and Liaohekou in Liaohe Oilfield as experimental areas, remote sensing monitoring results show that 18 well sites in Dabusu experimental region were withdrawn in 2019 and all achieved vegetation restoration; 15 well sites in Liaohekou experimental region were withdrawn from April to November 2018 to achieve natural restoration; and PetroChina has achieved results in ecological restoration of oil and gas well site withdrawal; remote sensing intelligent monitoring technology of oil and gas well site withdrawal and ecological restoration can realize large-scale and large-quantity well site withdrawal as well as efficient and accurate vegetation restoration monitoring. This technology ought to be used and popularized.

A large number of major oil fields in China have entered the late stages of development, and the decreasing production is increasingly unable to meet the continuously growing demand for energy. Therefore, it is crucial for oilfield production to accurately and rapidly predict the effects of production-increasing measures based on existing data. This paper comprehensively considers three types of data: geological static parameters, production dynamic parameters, and process parameters of measures. Advanced machine learning algorithms such as random forest (RF), support vector regression (SVR), and extreme gradient boosting (XGBoost) are separately used, together with data augmentation techniques and Bayesian optimization algorithms to construct the different enhancing production through measures prediction model. The best prediction model is optimized by comparing the scores of each model. The results of a comprehensive comparison of various models based on the mean absolute error (MAE), mean squared error (MSE), and coefficient of determination (R2) indicate that the model based on the extreme gradient boosting algorithm performs the best. The application of data augmentation and optimization algorithms significantly improves the model performance. The accuracy of predicting the oil production enhancement effect for a given measure can reach over 90%. Compared with traditional methods for predicting the effects of measures, this paper addresses the issues of long computational time in numerical simulations and difficulty in exploring the mechanism of oil production enhancement measures in depth, and achieves a rapid and accurate prediction of the multidimensional effect of measures for increasing oil production. This paper employs machine learning algorithms to fully explore the relationship between three types of data and oil production enhancement effects, accurately predicting the effect of measures for increasing oil production. It provides a technical foundation for selecting reasonable measures to increase oil production in oilfields and has certain guiding significance for actual production.

Casing damage is a common problem faced in the development of old oilfields. Some areas with weak mechanical properties of rock layers and other unfavorable factors are prone to severe casing damage, such as time of occurrence and plane distribution of casing damage wells. The stable state of damaged rock formations is a prerequisite for the implementation of workover and remaining oil tapping. Traditional stability evaluation of damaged areas often relies on qualitative analysis through well condition investigation and engineering logging, with incomplete considerations and a high rate of judgment errors, often leading to secondary concentrated casing damage after treatment. Therefore, based on the research and understanding of centralized casing damage mechanism, taking into account the changes in casing damage situation and the control of inducing factors, 9 indicators are selected to establish an evaluation index system for stability of casing damage areas. Gradient lifting algorithm is applied to achieve quantitative grading evaluation of the stability, with a verification compliance rate of 83.3%; meanwhile, classified adjustment measures are implemented to overcome unstable aspects, shorten the adjustment cycle, determine the timing of overall governance as soon as possible, and ensure the oil recovery of casing damage areas. This method was applied in the X67 casing damage area, guiding the implementation of various stability control workloads for 49 wells, and effectively improving the stability of the casing damage area.

The objective of this study is to analyze dominant controlling factors of the EUR of shale gas wells and then to forecast the EUR precisely by employing knowledge graph and automated machine learning techniques. First, an ontology knowledge representation model and a set of classification system for shale gas production are constructed, which include 13 shale gas objects such as basin, shale gas play, shale gas field, shale gas reservoir, and shale gas well, and their 112 geological, engineering and production parameters, such as mineral brittleness, fracturing section length, sanding intensity, and first-year production, and so on. Subsequently, structured data from existing databases are transformed, and loaded into the knowledge base. Large amount of unstructured data from papers, presentations, professional books are extracted and loaded by using various natural language processing (NLP) tools. The final shale gas knowledge base contains 56 shale gas plays and more than 1,000 shale gas wells worldwide. Based on the shale gas knowledge base, the graph embedding algorithm is used to convert the graph into a vector in order to train the machine learning models. Various automated machine learning frameworks such as TPOT, H2O, Auto-Sklearn, and AutoGluon are implemented and the performances are compared. According to the model with best performance, the main controlling factors of the EUR of shale gas wells are high-quality bed thickness, fracturing section length, and fracturing fluid volume, etc., which are consistent with shale gas production practices. The MSE and MAE of the best model on the testing dataset are 0.06 and 0.19, respectively. The approach of knowledge base construction and application developed in this paper can be extended to the entire life cycle of E&P process, which can make full use of various documents, data and knowledge accumulated in the oil and gas industry to conduct decision support.

Based on the engineering technology data in the ultra-deep oil & gas fields, this paper utilizes data insight tool to identify and extract information from various types of data stored in documents with text or tables, which meets the needs of data governance project. If the document information is about text content, the natural language processing (NLP) method is directly selected for recognition; If the document information is a table, it is necessary to convert the table into a heterogeneous data table with Date-Frame format first by Python language, and then recognize and extract it. These two processing methods can successfully convert unstructured data to structured data, solving the problem of low accuracy and low timeliness of extracting information from different documents. The NumPy & Pandas learning with Python language and other algorithms/functions play an important role in building metadata models, labeling fields, and training backend algorithms of data insight tool structure. The target trained extraction model is very crucial to the identification and extraction of various information. Relying on this and later, the qualified data generated after steps of extraction of target documents, selection of matching data for review and multi-level audit evaluation will be marked with “EDG”, which is the main data source of various professional databases of Tari Oilfield and the guarantee of the capacity and quality of the data lake. Examples show that the data insight tool has strong adaptability, obvious optimization effects, and superior performance compared to other extraction tools. The development and application of data insight tool have significantly improved the identification and extraction ability for engineering data of ultra-deep oil & gas fields, improved the identification accuracy and extraction speed, and met the needs of data governance.

The quality of cement bond is related to the safety of oil and gas well production and the service life of casing. At present, acoustic variable density logging (VDL) is the most widely used method for evaluating cementing quality in oil fields. The data interpretation of VDL still needs to rely on manpower, and the accuracy of interpretation results is restricted by human factors, and the workload is heavy. Oilfields have accumulated a large number of practically verified VDL interpretation results. It is of great research value and application potential to sort out these historical data and mine them with the help of deep learning technology, and establish an intelligent analysis method instead of humans to explain the cementing quality. In this study, the VDL cementing quality evaluation reports of several oil wells were collected. Through data preprocessing, the acoustic variable density images were standardized and segmented along the borehole direction. The cementation conditions of the first interface and the second interface corresponding to each segment of the acoustic variable density image were marked, and a sample set for cement bond quality evaluation was established. The cementing quality evaluation problem is transformed into an image classification problem, and the convolutional neural network method is introduced. On the basis of LeNet5, AlexNet and other classic image recognition architectures, considering the characteristics of acoustic variable density images, a personalized convolutional neural network (CBQNet) for cementing quality evaluation is designed, including 28 layers and more than 32 million learnable parameters. Using historical cementing quality evaluation samples to train and analyze the performance of convolutional neural network, the results show that: CBQNet has a training accuracy rate of 95.9% and a verification accuracy rate of 95.4% in the first interface cementing quality evaluation. In the cementing quality evaluation of the second interface, the training accuracy rate reached 90.8%, and the verification accuracy rate reached 88.1%. It shows that the convolutional neural network realizes efficient and accurate interpretation of cementing quality by mining and learning the interpretation results of historical VDL data, and provides a new method for cementing quality evaluation.

China is rich in tight oil resources, with a wide distribution range and a large amount of resources, making it one of the key areas for strategic replacement of future oil reserves and production. In response to issues such as strong heterogeneity of terrestrial tight oil reservoirs, difficulty in drilling high-quality oil layers, large production differences, and unclear main control factors for production capacity, a detailed analysis of dynamic and static data of production wells was conducted to analyze production performance and decline patterns. Production wells were classified according to production characteristics, and development indicators at different stages were statistically analyzed based on actual production days. Using a combination of principal component analysis and Pearson correlation coefficient, based on multiple dynamic and static data such as geological factors, fracturing factors, and development factors, and analyzing the correlation between different single and combined factors and cumulative oil production at different stages, the main control factors for different production stages of tight oil were obtained. A production capacity prediction model for tight oil fracturing horizontal wells was established based on machine learning intelligent algorithms, A production capacity evaluation and prediction technology for tight oil fracturing horizontal wells has been developed. By comparing with actual production data, the accuracy of the predicted results can meet production needs, providing a strong technical foundation for precise prediction and guidance of tight oil production in China.

Uncertainties in the characterization of new-found, ultra-deep, thin and low porosity Permian gas reservoir reduce feasibility for development index (DI) prediction through reservoir simulation. DI prediction with big data analysis approach are studied. Geology and production data from 30 mature gas fields are reviewed and 13 parameters are selected to represent geological features, deliverability and DI of individual reservoir. Based on the BP neural network algorithm, proxy models are established to correlate DI with geology and deliverability data, and the bagging method is used to effectively improve the experimental accuracy and stability while avoiding over-fitting phenomenon in the case of limited sample data. The coefficient of determination coefficient (R2) are selected to evaluate the prediction effect of DI. The mean value of the prediction results of the model with higher R2 value in 2000 numerical experiments was selected as the final prediction result. With the established proxy model, DI for QX reservoir in Permian formation are predicted and the influence of heterogeneity are also evaluated.

Mud logging serves as the “eyes” of exploration and development, acting as a counselor for drilling safety, the center of information transmission, and holding the first-hand data on oil and gas exploration and development. With the rapid development of informatization, digitization, intelligence, and remote support systems, the demand for high-quality mud logging data has continuously risen, where sensor calibration and calibration technology serve as the foundation for ensuring accuracy and reliability. This paper proposes an artificial intelligence-based comprehensive mud logging instrument sensor calibration and calibration technology, targeting the issues of prolonged service life, low precision, and low inspection rate of traditional mud logging instruments. The technology primarily involves collecting and pre-processing sensor output data such as filtering, sampling to eliminate noise, and improve the dataset's quality. Mathematical models of sensors were constructed using machine learning or deep learning algorithms to analyze the relationship between sensor outputs and actual values, which could also compute sensor errors and uncertainties. Algorithm optimization methods such as wavelet transform and adaptive filtering were used to process and analyze sensor data for different types of sensors and environmental conditions. The adaptive control algorithm was then utilized based on the predicted model results and actual measurement results to calibrate the sensor, ultimately helping to avoid errors and uncertainty in the traditional manual calibration process. Experimental results show that this technology has higher accuracy and reliability than traditional calibration techniques while maintaining simple operation, fast speed, and cost-effectiveness. This technology improves the level of detection and evaluation technology of comprehensive mud logging instruments, Standardizes mud logging equipment management, and plays an essential role in timely discovering, evaluating oil and gas layers, and optimizing drilling construction safety.

Major productivity breakthroughs have been achieved in key production layers of Jiyang shale, such as lower Es3 and upper Es4 producing layer of Shahejie Formation, and remarkable development have been gained. However, it is also limited by short production time, large production difference of single well, and unclear of production influencing factors. Comprehensive analysis of the main controlling factors of production for horizontal shale oil wells has become the research focuses. Field production data was taken to clarify the influence of various factors on the production of horizontal wells. Grep correlation analysis and principal component analysis were used to quantitatively analyze the sensitivity of 90-day average oil production, 180-day average oil production, and 270-day average oil production to the influencing factors, such as the amount of fluid used and the sand added. Research indicates that the amount of fluid used, the amount of sand added and the number of fracture events are the main engineering parameters affecting the production, while the content of gray matter, TOC and shale porosity are the main geological parameters affecting the production. The influence of geological factors on production gradually increase, while that of engineering factors on the production is gradually weakened in the late flowing production stag. The main controlling factors and variation rules of the production are preliminarily identified, which could provide guidance for the deployment of shale oil wells and fracturing design.

In the development of oil and gas fields, the accurate measurement of oil, gas and water production rate is the basis for calculating the key parameters such as water cut (WCT) and gas-oil ratio (GOR), as well as the important basis for formulating the stimulation measures and field development plans. The traditional surface flowing test is mainly carried out with the help of three-phase Separator or multiple phase flowmeters (MPFM), which is not only complicated operation process and long operating duration time, but also cannot be implemented for environmental protection reasons in some special location. In recent years, with the acceleration of digital oilfield transformation, many wellhead digital multi-phase flowmeters have appeared in the market. However, due to the interference of flow rate, high water cut, high gas content and other factors, the measurement accuracy of most three-phase flowmeters in the market is not good enough, which cannot meet the actual requirement of the customers. In view of the above problems, the design and development of a new on-line three-phase flowmeter is carried out, and the hardware and software system of the flowmeter is upgraded with an iterative and innovative method. Through large-scale field pilot tests to evaluate the performance of equipment, find out the problems during the testing process, and continuously improve the hardware design, software function and core model of the product, so that the instrument can detect the gas and liquid production rate online in real time, and realize real-time data collection and communication through the deep integration with the Internet of Things technology. Present the data to the user in a visual manner in the same time, thus, the whole field data sharing can be realized. The product was put on line in a domestic oilfield after pilot test. The application results show that the product has high stability and the ability to work under complex conditions, and can meet the needs of flow measurement under high GOR and high water cut in the oilfield, and has the feasibility to widely expand the application in domestic and foreign oil fields.

“Massive” logging data assets, due to their insufficient storage methods and normalization, cannot be quickly and accurately called up, become a “data island”, so that their value has not been fully explored. The current application scope of artificial intelligence is focused on single method research, with few system applications. However, intelligent interpretation requires the use of a large amount of logging data and related standard data. Based on a large number of documents related to large logging database and logging artificial intelligence, starting with supervised, unsupervised and semi-supervised intelligent algorithms, this paper expounds the application status quo and applicability of intelligent logging interpretation technology through machine learning for conventional logging lithology identification, automatic layering, sedimentary microfacies identification and reservoir identification. This paper briefly introduces the application status quo of logging data governance and mining technology. This paper summarizes the process of intelligent interpretation method, as well as the intelligent logging interpretation method and system based on a physical model under a standard big data platform. This paper discusses the existing problems in intelligent logging interpretation and evaluation and the feasible development direction of future research.

In oilfield production, the liquid production is an important indicator for measuring the production capacity of sucker rod wells and determining reasonable production parameters. Therefore, accurate metering of liquid production in sucker rod wells holds significant importance for oilfield automation production management. This paper proposed a physical-data hybrid-driven liquid production prediction method based on the attention mechanism to improve the accuracy of sucker rod well production metering. First, a physical-driven model for measuring liquid production based on the sucker rod well dynamometer cards is established, which ensures the rationality and interpretability of predicting liquid production. Then, a ResNet-based data-driven model is established to uncover the hidden features in downhole pump dynamometer cards and oil well production data. Finally, an attention mechanism is employed to couple the physical-driven and data-driven models, facilitating the identification of crucial features for liquid production prediction. The proposed method was tested on actual production data, and the average accuracy rate reached 95.67%, which was at least 2.43% higher than other best benchmark models for production prediction, and demonstrating good prediction accuracy and stability in special operating conditions. This approach successfully fuses the physical analytical model and data mining model of sucker rod wells, ultimately enhancing the interpretability and reliability of the model, thereby promoting efficient production management in oilfields.

Numerical reservoir simulation is an important technology in reservoir production development, but the computational consumption of numerical simulation is a key factor affecting reservoir history matching, production prediction, and optimization. By constructing a computationally fast machine learning model to learn the mapping relationship between reservoir model parameters and production data, a maximum alternative to the numerical simulation process can be achieved to improve the efficiency of reservoir management and decision making. The current surrogate models of reservoir numerical simulation for large spatial variables, including permeability and porosity fields, often extract spatial features by convolutional neural networks and later use recurrent neural networks to learn the time-series relationships of production data. In this work, we study the method using convolutional neural networks to extract spatial parameters of reservoir models and propose a new module to convert the temporal and spatial features of surrogate models. By converting the spatial features extracted by convolution and adapting the input features and dimensions of the recurrent neural network, maximum extraction of spatial feature parameters is achieved. The proposed method was verified on a 3D reservoir model, and the results indicate that the method can enhance the accuracy of the surrogate model.

This study focuses on the application of modern artificial intelligence (AI) techniques to improve the accuracy of drilling status recognition in the oil drilling industry, with the aim of enhancing safety and efficiency. To address the limitations of existing research in evaluation methods and practical applications, we constructed a unified drilling status dataset, introduced a more scientific evaluation criterion, the F1 score, and conducted a comprehensive evaluation and improvement of existing oil drilling status recognition methods. This paper provides an in-depth analysis of various drilling status characteristics and explores the applicability and limitations of different AI algorithms in drilling status recognition. Based on these findings, we propose optimized general drilling status recognition algorithms and validate the performances in real drilling status. Our research offers valuable insights and guidance for future oil drilling status recognition studies and is expected to promote safer and more efficient development in the oil drilling industry.

The theoretical regulation of production parameters in oil production engineering plays a significant role in the management of beam pumps. However, it falls short in identifying the inherent relationships among historical production data, thus failing to address the problem at its core. Valuable information can be extracted from historical well experiences through data mining techniques, offering new insights for adjusting production measures. To achieve this objective, an analysis is conducted to explore the factors and patterns influencing the exemption period of oil wells. Various methods, including expert experience and correlation analysis, are employed to process and selectively identify relevant features. Drawing upon the principles of oil production engineering and leveraging advanced big data processing techniques, these features are encoded to construct a comprehensive sample set that represents long-life wells. Subsequently, association rule mining is applied to uncover frequent patterns exhibited by these long-life wells. By setting a minimum support threshold of 0.01, the mining process encompasses a substantial dataset comprising over 1700 wells, leading to the discovery of more than 100 meaningful association rules. These rules are further prioritized and visualized based on their lift values, providing valuable insights into the experiential knowledge base related to effective measures for long-life well patterns. Consequently, this knowledge base becomes an invaluable asset, offering support for informed decision-making in terms of production parameter control and aiding in the development of scientifically guided production strategies.

The existing pumping unit downhole working condition diagnosis system has a high false alarm rate and a low accuracy rate of diagnosis for complex working conditions and abnormal working conditions. To address this problem, a diagnostic system of pumping unit workings is developed. First of all, the displacement-load data of the workover diagrams were converted into images, and then, through preliminary screening, manual review and data balancing from hundreds of millions of workover diagrams accumulated over the years, a sample database of 28 types of workover conditions, such as normal production, insufficient fluid supply, gas influence, rod breakage and tubing leakage, was established, with a total of about 760,000 samples, to compile a data set that is leading in quality and quantity in China. The project adopts “graphic + data” composite diagnosis, “graphic” corresponds to the power diagram, “data” refers to the electrical parameters, set pressure and other production parameters, and transforms the fault diagnosis problem of the power diagram into a deep learning-based image classification problem. Deep learning based image classification problem. A fault diagnosis method based on migration learning and category imbalance loss is designed. Better diagnostic results are obtained, with the single diagnostic accuracy of no less than 98% for common working conditions, 99.5% for normal production, 98.4% for insufficient fluid supply, and 97.2% for gas influence.

Drilling sensor failure leads to unavailability of kick monitoring parameters and the inability to apply intelligent kick detection methods. To solve this problem, a confidence evaluation indicator based on softmax is designed to measure the difficulty of kick identification, and the appropriate monitoring parameters are adaptively selected based on this indicator. Finally, an intelligent kick detection method using a cascaded GRU network with adaptive monitoring parameters is proposed in this paper. Kick identification experiments were conducted using simulated and measured data. The experimental results show that, when one monitoring parameter is unavailable, the recognition accuracy of the cascaded network proposed is improved by 10.61% on average and the computational load is reduced by 38.5% compared with the traditional gate recurrent unit network. The applicability of intelligent kick detection methods is significantly improved.

CO2 huff and puff is an important replacement method for the subsequent improvement of oil recovery after the elastic development of multi-stage fracturing horizontal wells in tight reservoirs. The rates optimization of huff and puff injection and production has the advantages of low cost, easy implementation, and obvious effects. At present, the rates optimization method of huff and puff injection and production is insufficient, and the interference between different huff and puff cycles and different stages were not fully considered. This paper established a CO2 huff and puff injection and production rates optimization method for multi-stage fractured horizontal wells based on the proximal policy optimization algorithm. We took the net present value as the optimization objective and huff and puff injection and production rates parameters as the optimization variables. The new method realized dynamic injection and production rates optimization with different huff and puff cycles and variable injection and production speed and variable injection and production duration, and considered the interference between various stages. The rates optimization results of CO2 huff and puff in multi-stage fracturing horizontal wells indicate that the optimal project extends the backflow time by reducing the backflow rate, fully improving the utilization degree of the backflow stage. At the same time, improving the efficiency of CO2 injection and reducing the number of cycles significantly reduce gas injection costs, achieving optimal economic benefits and providing guidance for on-site actual CO2 huff and puff injection and production rates optimization.

The development of strike slip faults in the central part of the Sichuan Basin is influenced by the structure, and the high yield wells reveal that strike slip faults have a close relationship with reservoir control. This article uses 3D seismic data from the HC125 work area in the Hechuan Tongnan area of the Sichuan Basin to carry out identification of strike slip faults based on artificial intelligence. Firstly, preprocess seismic data to improve the imaging characteristics of strike slip faults in seismic profiles. Secondly, developing method for edge coherence enhancement to highlight the faults boundaries. Finally, U-Net convolutional neural network machine learning method is used to identify main faults, and disorder detection technology is used to identify associated fractures and small-scale faults. According to this, a comprehensive detection technology suitable for strike-slip faults in the central Sichuan region will be formed. Compared with conventional fracture detection technology, artificial intelligence technology for fracture detection has a relatively high fault resolution, and the continuity and interpretability of deep fracture have been greatly improved; Results of multi-scale Fault Detection guarantees the research on the Mechanism of Controlling hydrocarbon accumulation through Strike-slip Faults in the Central Sichuan Basin. This technology effectively improves the interpretation accuracy and classification accuracy of strike slip faults in the central Sichuan region.

The massive data and information in the petroliferous basins formed by exploration and development are extremely valuable. Thus, they need to be deeply mined and utilized by new technologies to provide data support and decision basis for exploration and development. The knowledge graph can well integrate the knowledge contained in these data and documents. However, its concept and relationship rely on manual construction, which results in its limited coverage of knowledge areas. The traditional question-and-answer (Q&A) method can get relevant answers from documents according to questions, which has the characteristics of wide knowledge coverage. Nevertheless, it is difficult to understand the contents of professional fields, which leads to its low accuracy in petroliferous basins. In order to address the above problems, this paper proposes a hybrid Q&A method for merging the knowledge graph and documents in petroliferous basins. The method takes the knowledge graph of petroliferous basins as the knowledge base of professional background. Additionally, it obtains professional-related knowledge contents from documents. In particular, to answer the question on petroliferous basins, the method firstly extracts entities from the question according to the knowledge graph. Then, with these entities, the method converts the question into a query in the knowledge graph, obtaining partial candidate answers to the question. For obtaining candidate answers from documents, the method constructs a deep semantic matching model which incorporates knowledge graph embedding. The model can match the question and answers in documents base on the information from the knowledge graph. Finally, the method leverages a sort algorithm to reorder the above two types of candidate answers from the knowledge graph and documents respectively. Compared with traditional Q&A methods, the hybrid Q&A method supports professional Q&A scenarios for the knowledge graph and documents of petroliferous basins, improving the efficiency of users’ knowledge query and increasing the recall rate while ensuring the retrieval accuracy. The hybrid Q&A method has the characteristics of convenient operation, strong interaction and high accuracy, etc., which provides a platform on knowledge deep sharing and application for the study of petroliferous basins.

With the application of deep learning algorithms in the industry, artificial intelligent technology has been developed in the field of seismic data interpretation in petroleum geophysical prospecting. This paper first starts from the analysis and research of Fully Convolutional Networks (FCN), U-Net model, the calculation of its lower accuracy results were analyzed, and the shortcomings of the model were found and pointed out; then it was proposed to introduce the High-Resolution Network (HR-Net) model into the field of intelligent interpretation of seismic data, and improve its network algorithm to make it more suitable for 3D space seismic data analysis and processing. Considering that the interpretation results of the FCN, U-Net, HR-Net algorithm cannot fully reflect the periodic phenomena and laws in the depth of the formation, the author improves HR-Net model and the high-resolution semantic fusion of the HR-Net model is also improved. The research result is the improved HR-Net algorithm model, which has certain application and promotion value in interpreting reservoirs and predicting faults from seismic image data.

In view of the problems that the load cell used to test the indicator diagram in the pumping unit is easy to drift in the long term and needs manual regular maintenance, a new method is proposed to demonstrate the indicator diagram of the electrical parameters in the pumping unit well. By combining neural network and big data analysis technology, BP neural network model is established to carry out learning、training and simulation analysis on the historical data of pumping unit, to find the corresponding relationship between electrical parameters and indicator diagram, and to realize the direct conversion of indicator diagram using electrical parameters. After145 field tests, the accuracy of electrical parameter inversion diagram based on neural network reaches 93.2%. This method has the advantages of low model complexity, fast operation speed and high accuracy, which provides a new way to obtain the indicator diagram of pumping unit well, and has great significance for the digital construction of pumping unit well.

The correct identification of reservoir fractures is of great practical significance for accurately evaluating the oil and gas reserves of reservoirs and the effectiveness of hydraulic fracturing, especially for macroscopic fractures between wells with an aperture greater than 1cm, which are often the culprit of hydraulic fracturing failure. However, traditional reservoir fracture identification methods face difficulties in feature extraction and illposedness in inversion problems, making it difficult to ensure the accuracy of results. Regression prediction models based on convolutional neural networks have powerful nonlinear data mapping capabilities and can replace traditional geophysical inversion calculations. To address the issue that traditional convolutional neural networks can only handle scalar data while the actual electromagnetic field data is vector valued, this paper proposes a macroscopic fracture identification method for inter-well reservoirs based on complex-valued convolutional neural networks. By using the real and imaginary parts or amplitude and phase data of the observed field as the input to the complex-valued convolutional neural network, the information input to the network is increased, enabling the network to extract more target features and improve the identification ability of reservoir fractures. Through comparative experi-ments based on amplitude scalar data and complex-valued convolutional neural networks, the results demonstrate that the electromagnetic detection data inversion for fracture identification based on complex-valued convolutional neural networks has higher resolution and provides a new approach for the accurate identification of macroscopic fractures in inter-well reservoirs.

As the revitalization and development of Daqing oilfield has entered a new stage, new requirements have been put forward for the large-scale processing of massive seismic data, and the seismic data processing has entered the era of basin-level massive pre-stack data. The quality of massive pre-stack seismic data is closely related to the quality of migration imaging, which greatly affects the optimization of trap target, well position deployment and adjustment of engineering scheme. Therefore, the comprehensive treatment of massive prestack seismic data is of great significance to the high-quality development of oilfield exploration and development. Due to issues such as missing data header information, non-standard archive content, difficulty in data back-tracking, unsynchronized data structure and document storage, the quality of pre-stack seismic data is affected to varying degrees, unable to meet the needs of high-quality, efficient, and full lifecycle exploration and development. Professional big-data comprehensive management means are needed. Through continuous exploration and practice, a “Five Refined Quality Control” has been established for pre-stack big-data management methods, which includes institutionalized modes, quality control processes, evaluation forms, software standardization, and management informatization, achieving standardized management and standardized governance of pre-stack seismic data. The method is applied to massive seismic pre-stack data, and established a pre-stack middle results database for the entire exploration area of Daqing. In the continuous processing task of tract 24 in the Gulong area of the Songliao Basin, high-quality middle results data were used to shorten the 26 month continuous processing cycle to 9 months, greatly improving the seismic processing efficiency and achieving significant promotion and application results.

In the process of digital transformation of logging, how to integrate traditional business applications with digital and intelligent technologies that have been accumulated for many years has become an important topic for research and exploration in the logging industry. The migration of traditional logging applications to the cloud has gradually become an effective means of digital transformation.China Petroleum Organization has launched the construction of the Exploration and Development Dream Cloud Platform. This article analyzes the characteristics of traditional logging heterogeneous systems and refers to the Dream Cloud technology framework. It classifies and studies the cloud migration modes, proposes three cloud migration modes, namely basic infrastructure migration, thin client migration, and microservice migration, based on the complexity of the architecture of traditional logging applications. It forms a complete set of cloud migration solutions for traditional logging applications.In response to the characteristics of the logging industry software, breakthroughs have been made in the communication integration of heterogeneous systems, data transparency exchange, and cloud-based data processing, effectively improving the overall service capabilities of specialized logging software.

As the important equipment in the oil drilling rig system, the operation status of drilling pump and winch directly affects the safety and efficiency of oilfield drilling production. The existing inspection and maintenance of drilling equipment mainly rely on manual patrol inspection and post-maintenance. The intelligent monitoring and health diagnosis technology of drilling key equipment can help to realize the life cycle management of drilling key equipment and significantly improve the level of drilling equipment evaluation business. For the health monitoring and fault diagnosis analysis of drilling pump and winch, this paper focuses on the research and design of the online monitoring and fault diagnosis system of drilling pump and winch using the Internet of Things technology. This system can realize remote real-time monitoring and fault diagnosis of the drilling process, reduce the workload of on-site personnel, improve the management efficiency of equipment, and more safely ensure the exploration and development of oil and gas resources.

Since the utilization of data analytics in oil field industry, data mining has become increasingly important. Various decision-making algorithms derived from data are closely related to the quality of data, which makes data quality assessment an indispensable part of the intelligent construction of oilfield. General data quality assessment models are not suitable for centralized oilfield scenarios because the quality of datasets depends on their usage rather than a simple stacking of individual data units. For example, datasets containing data units with good quality yet serious homogeneity cannot meet the data requirements in deep learning. This paper is based on the theoretical model of process measurement and adopts the second-level fuzzy comprehensive evaluation model. We calculate the member-ship degree of each factor set based on the business demand by the AHP. The oriented oilfield structured data quality assessment model is then established. This model provides theoretical basis and technical support for oilfield data preprocessing, decision-making and staged evaluation of data governance.

Leakage of premium thread connections tubing is the main reason for annulus pressure and affecting well integrity level. At present, helium gas seal detection and manual monitoring of make-up torque curve are mainly used to ensure the integrity of gas seal of well string. However, the helium seal detection environment is static detection, which fails to describe the air tightness of the tubing under complex downhole load; The manual monitoring of the make-up torque curve depends on the field engineer with certain experience to check the standard curve one by one. The results are greatly affected by subjective factors, so it is difficult to unify the measurement standard. Therefore, a machine learning method based on convolutional neural network (CNN) is proposed to automatically identify and classify the makeup torque curve of special threaded tubing. In order to achieve this goal, firstly, according to the failure of gas seal detection, manufacturer's manual and field experience, the categories of makeup torque curve are divided, including typical curve, acceptable curve and unacceptable curve. Secondly, in order to improve the model training accuracy and further improve the prediction results, the data expansion technology is used to expand the training database. Finally, the multilayer convolutional neural network model is built and trained and verified based on the data. In the model verification stage, four comprehensive evaluation methods are used: typical rate (TR), acceptable rate (AR), unacceptable rate (UR) and accuracy (P). The proposed CNNs model is evaluated accurately and compared with state-of-art machine learning algorithms such as SVM and logistic regression.

In order to diagnose and predict the drilling complexities before drilling operations, and provide the drilling operators on well site with some hints, so that they are mentally aware of what kind of drilling complexity will occur in the future, and they could take the corresponding preventive measures to prevent the occurrence of some drilling complexities with the lowest possible economic cost. The handling methods of the drilling complexity cases that have occurred are treatment solutions made by the experts on well site based on their professional knowledge and years of rich drilling experience, which has a very important reference value for the later drilling operations. On the basis of case-based reasoning method, according to the adjacent well data, the computer technology, the artificial intelligence, and the data mining technology, this paper will construct a drilling complexity intelligent prediction model to utilize an open-source software and regression analysis method of causal relationship model. Use the ROC curve and confusion matrix to evaluate the performance of the drilling complexity intelligent prediction model, and the accuracy of the model is between 70–80%. It is recommended to use more drilling complexity cases to train the model in the later stage to improve the accuracy of the prediction model.

Production optimization is a crucial technology for efficient development of water-driven reservoirs. By adjusting the injection and production rate of oil and water wells in a reservoir block, the optimal production solution can be provided for the field to maximize the economic benefits while minimizing the costs. In this paper, a soft actor-critic (SAC) based reinforcement learning offline production optimization method are proposed, which models the production optimization problem as a Markov sequence decision process. Specifically, the deep reinforcement learning (DRL) agents aimed at maximizing the economic efficiency. The agent updated the policy model incrementally using the data obtained by interactive sampling with the environment to accelerate the convergence of the optimization process. In addition, to achieve offline optimization, a state transfer model is constructed that captures the dynamics of the reservoir under time-varying well control conditions using historical regulation experience. In the offline deployment stage of the cloud platform, the trained policy network and state transition network are utilized. In this way, the well control scheme for multiple future time steps can be calculated using only the current state of the reservoir. Reservoir instances show that this method is highly efficient and can provide optimized solutions within seconds, and the optimization performance is also remarkable. With the good effect of water control and oil increment, the target model can achieve higher net present value (NPV). The proposed offline method, which embedding control strategies into the model and utilizing a state transition model to capture the dynamics of the system, offers a novel approach to intelligent production optimization. By enabling offline optimization deployment on a cloud platform, this approach provides a practical solution to meet the demand for intelligent oilfield construction.

Digital transformation is a key factor to improve the efficiency of upstream production companies and accelerate key business decisions in the oil and gas industry. The goal of enterprise digital transformation is to transform traditional business into digital business. The essence of digital business is to process data as a new production factor and build products with data as the main form of existence. With the accelerated development of digital transformation, efficient management and use of massive data has become one of the core challenges in the process of digital transformation of enterprises. Oilfield data governance is an important means to improve oilfield production efficiency and economic benefits, and is also a necessary condition for data driven decision-making. Breaking data islands through effective data governance can promote the application of big data technology. This paper takes the data governance methods and key technologies in the application of oil and gas field exploration and development data as the research object, expounds the data governance system of exploration and development, and gives the solution of data governance and unified data management including data resource inventory, data standards, data quality management, data services and big data operation and maintenance, covering the main contents of the entire life cycle of data governance to solve the data governance problems encountered in the application of big data in the process of exploration, development and production, as well as in the construction of intelligent oil fields.

Paleozoic carbonate rock is the key field of oil-gas exploration in Tarim Basin. Fracture-cavity reservoirs are often developed in carbonate reservoirs in Tarim Oilfield, a large amount of oil and gas resources are dis-tributed in Paleozoic reservoirs with different burial depths and scales, accurately and quickly identifying these fracture-cavity reservoirs is Significant to the oil and gas exploration, development, and production of the Tarim Oilfield. With the continuous development of artificial intelligence technology, machine learning methods have been widely applied in various scenarios of oil and gas exploration and development, bringing new opportunities for the development of carbonate reservoir prediction technology. Based on drilling, logging and seismic data, this study comprehensively analyzes the structure, rock and physical properties of carbonate reservoirs in the study area, exploring the main controlling factors of carbonate reservoirs. On this basis, a sample set corresponding to the fracture-cavity reservoirs in the study area was constructed, by using machine learning methods, a prediction model for carbonate rock fracture-cavity reservoirs has been established, which can intelligently predict carbonate rock fracture and cave reservoirs in the research area. The trained model of carbonate rock fracture and cave reservoir prediction can quickly and accurately identify fracture and cave reservoirs on post stack seismic data. This study demonstrates that methods are based on machine learning can quickly and efficiently predict carbonate rock fracture-cavity reservoirs.

As an efficient and intelligent means of knowledge organization, knowledge graph has become the core force driving the development of artificial intelligence. Hydraulic fracturing is an important measure for increasing production and injection in oil and gas fields, with complex design processes and numerous influencing factors. In order to achieve rapid and accurate optimization of fracturing plan, this paper proposes a method for optimizing fracturing plan based on knowledge graph. By combing the system of fracturing domain knowledge, fracturing knowledge graph is constructed. Extracting characteristic parameters describing the geological engineering double sweet spot in multiple dimensions and multiple scales, and showing the characteristic parameter-related entities, relationships, and attributes as vectors via graph embedding technique. Integrate expert knowledge with artificial intelligence to build a fracturing effect prediction model and optimize the fracturing plan. In this study, more than 500 fracturing oil wells in a tight sandstone block are taken as objects to build a knowledge graph. Based on well test and production test data and historical production, this study predicts the fracturing stimulation effect and optimizes the fracturing engineering parameters. The calculation results indicate that factors such as reservoir thickness, oil saturation, number of fracture clusters, and half length of fractures have a significant impact on the fracturing effect. The coincidence rate between the predicted capacity of production and the actual capacity of production is over 91%, and the efficiency of fracturing plan design is increased by more than 20 times. The research results can provide scientific basis for predicting fracturing effects and optimizing fracturing engineering parameters, greatly improving the efficiency and quality of fracturing design, and improving the success rate of fracturing construction.

Based on the dream cloud platform, and on the Jidong oilfield basis of basin-level regional lake construction, some basic requirements for oilfield production analysis have been realized. However, with the development of informatization and the deepening of research work, it is necessary to develop business characteristic scenarios such as modern oil reservoir visualization and collaborative management. Using WebGL and GPU acceleration technology, an integrated collaborative online work platform is formed that integrates the “five fields” of reservoirs 4D virtual visualization and dynamic analysis results. The results show that: ①Through the standardization of model volume analysis and storage format, online unified management and visual display of 3D reservoir numerical simulation models with a large number of grids based on WebGL and GPU acceleration technology have been realized. ②Based on the simulation results of the “five fields” of four-dimensional geological field, pressure field, saturation field, seepage field and chemical field, by mining the internal data relationship of the reservoir dynamic model, the dynamic analysis and 2D-3D correlation graph could be visualized online, providing support for deep mining of reservoir five-field model data. ③Reservoir research that IFEDC-202315248 2 integrates 2D reservoir dynamic analysis, 3D model visualization, 4D model visualization and five-field quantitative analysis is an internal fluid migration based on macro development situation analysis during the entire life cycle of the reservoir In-depth analysis of the rules. The platform realizes scientific research collaboration among development plate research data, software, and achievements, lies a solid foundation for high-efficiency and high-quality oil and gas exploration business work, and provides strong support for the high-level application of Dream Cloud.

Under certain conditions, when crude oil is moved by external forces, the property of internal friction generated between crude oil molecules is called crude oil viscosity. The viscosity of crude oil reflects its complex seepage state in porous media. Underground crude oil with high viscosity, always has great flow resistance in porous media, thus the flowing becomes more difficult. Oil viscosity is an indispensable key parameter in the process of dynamic analysis, reservoir engineering calculation and reservoir numerical simulation, which has critical influence on the field of well production or crude oil storage and transportation. Due to different oil viscosity, recovery approach of oil reservoirs, technical measures for storage and transportation, and the quality of oil products will be affected. The composition of crude oil is complicated, but it is mainly composed of carbon and hydrogen elements. The composition has a crucial effect on oil viscosity. Therefore, according to composition data of the actual oil sample, the determination dataset of oil viscosity is constructed together with other key parameters that affect the viscosity of crude oil within the reservoirs. Based on various machine learning algorithms, like extremely randomized trees and XGBoost, determination methods of oil viscosity based on component data and machine learning algorithms are established. In the construction process of computational model of oil viscosity, whole dataset is parted to the training dataset and the testing dataset in the ratio of 8:2. The training dataset is mainly used to determine the best hyper-parameter combination of machine learning algorithm, while the testing dataset is used to determine the accuracy and adaptability of the corresponding method. Compared with methods such as experimental method and empirical formula method, the determination method of oil viscosity based on component data and machine learning algorithm does not require extra experimental costs and has a considerable degree of accuracy. Once the relevant input parameters are determined, the viscosity determination of multiple groups of oil samples could be completed quickly and accurately.

Drilling and completion is the core of energy production and plays an irreplaceable role in improving oil and gas exploration and development and increasing crude oil recovery. Similarly, drilling data is of great significance in drilling optimization, perforation, hydraulic fracturing, and oil and gas production. However, in actual production, data quality and application often fall short, making it difficult to further enhance and optimize oilfield engineering technology. Therefore, effective data governance to address data quality and application problems has become an urgent and critical issue. Additionally, the scale, quality, complexity, and security of data are essential issues that must be considered in data governance processes. To address these issues, this paper proposes an engineering technology data governance approach covering data quality control, data standardization, data modeling, and data mining. These methods and tools have been applied in production practices, and have achieved good results. Additionally, this paper explores the application of intelligent data governance, which aims to quickly and efficiently manage data. This paper compares the advantages and disadvantages of existing data governance algorithms in data governance, providing reference application scenarios for more efficient, reliable, and accurate drilling data governance. In summary, the proposed data governance approach and techniques provide the foundation for further improvements in oilfield engineering technology and management, as well as enhancing data utilization and decision-making for the petroleum service industry. Effective methods and tools must be used to address the challenges of data governance, including data scale, quality, complexity, and security. The proposed data governance approach for engineering technology, along with its applications and exploration in intelligent data governance, provides significant support for the exploration and implementation of data governance in the petroleum industry.

With the gradual application of knowledge atlas technology in oil and gas exploration and development fields such as intelligent evaluation of oil and gas reservoir and intelligent identification of oil and gas reservoir through logging, the application of knowledge Atlas in oil and gas industry has attracted more and more attention. In view of the problem of knowledge expression and application in E&P sector field, a professional word bank of E&P was established based on web crawler technology, a feature corpus of E&P field was established by extracting E&P sector related achievement documents and literature data, and the principle of E&P knowledge graph ontology construction was formulated and the construction process was defined. The multistage classification system of oil and gas exploration and development knowledge atlas ontology is described and constructed from multiple dimensions such as sector, object and feature, and the knowledge atlas ontology model of oil and gas exploration and development sector domain is formed, which lays a foundation for the in-depth application of knowledge atlas technology in petroleum exploration and development.

Natural fractures are effective storage spaces and important seepage channels for oil and gas reservoirs. Accurately identifying natural fractures in reservoirs is crucial for the exploration and development of oil and gas resources. This article combines conventional and imaging logging data and uses machine learning to automatically identify natural fractures in reservoirs. The fracture labels of conventional logging come from imaging logging. Conventional logging data is decomposed through multi-scale wavelet to extract components that reflect fracture information, and further build the original data set. The AdaBoost model is trained based on a modified dataset of balanced samples for automatic fractures recognition in logging. The research results indicate that the approximate component and high-frequency component reflect the fluctuation of the formation and noise information respectively, and have little impact on the reservoir fractures identification; The medium frequency component can reflect the characteristic information of fractures and can be used for model training; After hyper-parameter optimization, the AdaBoost model has high accuracy and generalization ability, and can still accurately identify the types and distribution of natural fractures from the actual unbalanced logging data. This research has important guiding significance for the accurate characterization and construction of reservoir.

In order to ensure the safe production of oil drilling equipment, about 2000 sets of in-service oil drilling rigs and workover rigs need to be tested and evaluated every year, covering Daqing, Changqing, southwest, Tarim, Dagang and other major oil and gas fields as the key equipment of the drilling rig system, the bearing capacity of the Derrick is directly related to the safety of production. In view of the complex data transmission line, low efficiency and high labor intensity of the wired strain tester, a wireless strain testing system with the function of testing and analyzing the bearing capacity of Derrick is developed the integration of wireless transmission of Derrick bearing capacity test data, on-line analysis and load capacity evaluation is realized. Through the quasi-calibration of Beijing Institute of Metrology and Field Test of Oil Derrick, it is proved that the research results have high accuracy in strain detection (JJG623–2005), strong stability, high efficiency, low power consumption, friendly interpersonal interface and broad application prospects.

To solve the problems of complex in-situ stress of tight sandstone reservoir, few sample points of experimental data, difficulty in in-situ stress prediction, etc., a method for one-dimensional, two-dimensional and three-dimensional in-situ stress prediction based on geomechanics and BP neural network was innovatively proposed by comprehensively using various data such as core data, mechanical experimental data, logging data, etc. In this method, the rock mechanics parameters of single well in the study area were predicted by neural network method using the logging data as the learning sample and measured rock physical parameters as the monitoring data first; then the in-situ stress of single well was accordingly calculated by empirical formula, and predicted and analyzed by neural network algorithm using the calculated in-situ stress of single well selected by error analysis and the indoor measured in-situ stress as the monitoring data and the conventional logging data as the learning samples. The application in the actual areas shows that the predicted results of in-situ stress not only conform to the measured data, but also follow the logging curves, and thus provide an important basis for the design of integrated geological engineering scheme.

Based on the production data management system of oil, gas and water wells (A2), Changqing Oilfield has developed comprehensive query, big data analysis and other deepening application functions, which play an important supporting role in oil and gas exploration deployment, production dynamic analysis and other work. At present, Changqing A2 Regional Center has many problems in daily production data synchronization, such as long synchronization time, low timeliness, weak verification logic, and data timeliness and accuracy are facing challenges. In view of the above difficulties, this paper designed a new data synchronization scheme and completed the feasibility test of key technical points by analyzing the publish and unlock process of A2 data and the current synchronization logic, and built a full process closed-loop synchronization system of “publish awareness, decentralized synchronization, unlock identification and local update”. Through actual operation, the average data synchronization time was shortened from 83 min to 2 min, and the unlock recognition rate was improved to 100%, which realized the efficient application of A2 production data and boosted the high-quality secondary accelerated development of Changqing Oilfield. At the same time, the scheme has good guidance and reference significance for other oilfield companies that need to build A2 regional data center.

Reservoir numerical simulation is a time-consuming and expensive procedure, especially for production optimization and history matching, which requires multiple calls to the reservoir numerical simulator. The surrogate model based on deep learning can provide a proxy solution to the process of calculating the production curve by the numerical reservoir simulator with approximate accuracy and higher computational efficiency, while an attention mechanism can better capture the local time characteristics of production time series. The attention mechanism is introduced based on deep CNN and LSTM to build an attention-based temporal and spatial convolution recursive neural network for surrogate modeling which is capable of extracting spatial features of reservoir static parameters and handling temporal data, and establishing an image-sequence mapping relationship from reservoir static parameters to reservoir production curve which is used to predict the reservoir production curve. The constructed surrogate model can fast and accurately predict production curves and improves the computational efficiency of production optimization and history matching.

Production optimization is an important way to improve technical and economic benefits in the process of reservoir development. Generally, most production optimization problems of chemical flooding are solved separately using mathematical algorithms, which limits optimization efficiency. This paper introduces the prior scheme obtained from reservoir engineering method into the optimization mathematical model to improve the efficiency of production optimization problems of chemical flooding. Firstly, the reservoir numerical simulation model and optimization mathematical model for chemical flooding are established. Secondly, the injection and production allocations are carried out through statistical analysis of the present development performance of reservoir, and a prior scheme based on reservoir engineering method is obtained. Finally, the prior scheme is used as the initial scheme for optimization. The optimization mathematical model takes net present value as the objective function, and the injection-production volume and chemical agent concentration as the optimization variables. The solving algorithm adopts particle swarm optimization. It can be seen from the results that the net present value of the uniform scheme is 0.761 × 108 RMB while 0.963 × 108 RMB for the prior scheme, which has an increase of 26.54%. Moreover, the conventional method converges to 1.317 × 108 RMB after 22 iterations, while the proposed method converges to 1.328 × 108 RMB after 11 iterations. The proposed method reduces calculation amount by 50% with satisfactory accuracy. Therefore, the proposed method using the prior scheme obtained from reservoir engineering method as the initial scheme achieves better optimization performance than conventional method. This method achieves the combination of mathematical theory and engineering experience, and providing an effective way to reduce calculation costs and increase efficiency for solving reservoir optimization production problems.

In recent years, the country has successively passed a series of laws and regulations on the reform of the mining rights system, especially related to exploration and construction, land use, and ecological environment restrictions, which have a profound impact on the exploration decision-making and development of oil fields. Analyzing the available dimensions of mining rights, comprehensively evaluating the ground environment, determining whether the ground conditions can be constructed, and the difficulty of construction, has become one of the important tasks in oilfield exploration deployment and mining rights evaluation. GlobeLand30 is the world's first high-resolution land cover dataset, with the advantages of comprehensive information coverage and high display accuracy. This article takes GlobeLand30 as the starting point, and through key steps such as data collection and download, coordinate projection conversion, and vectorization into maps, it achieves the accurate display of all surface elements in oilfield exploration areas for the first time, accurately implementing the distribution and morphology of various surface elements. Meanwhile, combined with the analysis of oilfield exploration and construction technology and policy restrictions, the surface types of difficult operation areas in the exploration area have been clarified, and precise evaluation of the surface environment has been achieved through hierarchical classification and vectorization mapping. The results indicate that the main types of surface cover in the Daqing Oilfield exploration area of the Songliao Basin are cultivated land, grassland, water bodies, wetland, and artificial surfaces; The most difficult construction operations are water bodies and wetland, followed by artificial surfaces, and the construction difficulty of cultivated land, grasslands, and forest also needs to consider policy restrictions. The ground environmental assessment method based on GlobeLand30 has important scientific guidance for oilfield exploration deployment decision-making and mining protection plan formulation.

In conventional reservoir modeling, the well log curve shape information is usually lost when it is up-scaled. In this study, we implemented machine learning method to capture the well log curve shape information of each well and clustering those wells in target spatial domain. One state-of-art machine learning algorithm used in most of time series classification area is implemented for feature extraction. All the wells are grouped into different groups according to the similarity of extracted features from multiple log curves. The final spatial reservoir segmentation is obtained through spatial interpolation from the clustered well groups. The results of the 2D spatial map provides valuable insights to the depositional background analysis through providing the lateral geological heterogeneity features. One small synthetic data set is used in case study to illustrate this method as an effective way to characterize the spatial heterogeneity. As illustrated in the case study, the proposed spatial segmentation technique can be used as a fast geological modelling method to integrate geological heterogeneity features embedded in multiple well logs.

Continued low oil prices and the demand for green and low-carbon development have put forward urgent requirements for the traditional oil and gas industry to upgrade its production model. In recent years, with the deep integration of emerging technologies such as 5G, Internet of Things, artificial intelligence, cloud computing, big data, edge computing and the energy sector, new opportunities and challenges have been created for the traditional oil and gas industry to seek industrial upgrading, and digital transformation and intelligent development have become a way to achieve oil and gas fields Digital transformation and intelligent development have become the inevitable trend to achieve high quality development of oil and gas fields. As a guarantee unit for seasonal peaking and emergency gas supply in the Beijing-Tianjin-Hebei region, Jidong Gas Storage is responsible for the stable delivery of natural gas energy supply in the region. The non-linear relationship between gas supply capacity and formation pressure changes in natural gas storage reservoirs makes accurate regulation extremely difficult; the frequent changes of gas injection and extraction wellbore conditions make the control of wellbore integrity a IFEDC-202315047 2 great challenge; the non-linear changes of formation supply capacity in upstream storage reservoirs and the rapid changes of downstream customer demand make the ground system load fluctuate greatly, making it difficult to ensure the smooth operation of the pipeline network. In the face of these challenges, exploring the construction of an intelligent gas storage reservoir with integrated and coordinated regulation of the ground, wellbore and formation is the best solution to meet this demand.

The original reservoir dynamic monitoring data management system of Changqing Oilfield was built in 2008. With the continuous development of oilfield, dynamic monitoring technology is also developing, business management is becoming more rigorous and standardized, and the operation problems of the original database are becoming increasingly prominent. Unable to meet and adapt to the actual needs of enterprise development, it also faces the bottleneck of iterative upgrade due to the difficulty of function expansion, which greatly hinders the subsequent digital transformation. System problems are mainly reflected in the following aspects: outdated system architecture, incomplete application functions, and inconsistent data standards and specifications. Therefore, it is urgent to reconstruct the original dynamic monitoring database system based on cloud native architecture and business application. Based on the development planning of digital transformation and upgrading of petroleum enterprises, this paper comprehensively expounds the construction and application of reservoir dynamic monitoring data management system based on H5 technology and oriented to micro service. Through the cloud upgrade of reservoir dynamic monitoring database, to solve the problem of timely, complete and accurate business data; Form closed-loop management of business chain including plan delivery, plan distribution, data archiving and plan tracking; By using the functions of report statistics, data analysis, graph drawing and comparison, we can better assist the management and technical personnel in application analysis, and guide the efficient development of reservoir by giving full play to the value of data assets.