Digital intelligent risk assessment method, platform and equipment for oil and gas field environment
By employing a digital risk assessment method that combines environmental monitoring data and environmental regulations knowledge graphs, the problem of insufficient intelligence in data collection and processing in oil and gas field environmental management has been solved, achieving efficient and accurate environmental risk assessment and promoting the green development and sustainability of oil and gas fields.
Patent Information
- Application Number
- CN202511031917.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-25
- Publication Date
- 2025-11-07
AI Technical Summary
In the current environmental management of oil and gas fields, the low frequency of data collection, reliance on manual experience for data processing, and lack of standardized preprocessing result in a low level of intelligence in environmental risk assessment, making it impossible to identify potential risks in a timely manner, failing to meet environmental protection regulations, and lacking systematicity, leading to environmental damage and resource waste.
The digital risk assessment method is adopted. After obtaining environmental monitoring data from oil and gas fields and preprocessing it, it is matched with the environmental regulations knowledge graph. A deep neural network model is used to assess the environmental risk level and output a risk assessment report. The regulations knowledge graph is dynamically adjusted to adapt to environmental changes.
It enables intelligent assessment of environmental risks, improves data quality and assessment accuracy, supports preventive intervention, reduces pollution incidents, and enhances environmental compliance and management efficiency in oil and gas fields.
Smart Images

Figure CN120911755A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of oil and gas field environment monitoring, and particularly relates to a digital risk assessment method, platform and equipment for oil and gas field environment. BACKGROUND
[0002] With the continuous growth of global energy demand and the promotion of clean energy transformation, oil and gas fields as important energy production bases inevitably produce various environmental pollution problems in the process of exploration, development and production. These problems mainly include waste solids (such as drilling waste and sludge), waste gas (such as volatile organic compounds and combustion emissions), waste liquid (such as oily wastewater and acidic waste liquid) and noise pollution, etc. These pollutants not only cause direct damage to the surrounding ecological environment, but also may cause long-term effects such as soil degradation, water pollution and air quality decline. At the same time, affected by factors such as wind speed, wind direction and rainfall, the diffusion path of pollutants is complex and variable, further increasing the difficulty of environmental risk assessment. According to relevant statistical data, in recent years, oil and gas field environmental accidents have occurred frequently, some of which are due to the delay in identifying pollution sources and the lag in management response, resulting in high economic losses and ecological restoration costs.
[0003] Traditional oil and gas field environmental management mainly relies on manual inspection, laboratory sampling analysis and basic sensor monitoring. For example, field staff regularly collect pollutant concentration data and combine meteorological observations to manually record and preliminarily judge. Although this method can obtain basic environmental information, it has many limitations. First, the data collection frequency is low, usually once a day or once a week, which cannot capture real-time changes, especially in bad weather or emergencies, the monitoring data is easy to be missing or biased. Second, data processing relies on human experience, lacks standardized preprocessing, resulting in noise interference and heterogeneous data integration difficulties, affecting analysis accuracy. In addition, existing environmental assessment is often limited to a single pollutant type or a local area, without fully considering the comprehensive influence of multi-source data and the corresponding relationship with environmental protection regulations. The environmental protection regulation system is complex, including national pollutant emission standards, enterprise internal specifications, etc., but traditional methods are difficult to achieve efficient retrieval and application, resulting in low compliance inspection efficiency, and unable to identify potential risks in time.
[0004] In the prior art, some digital tools such as environmental monitoring software or simple data management systems have been preliminarily applied to oil and gas fields, but most of them are limited to data storage and basic statistics. For example, some systems can record pollutant concentrations and meteorological parameters, but lack deep analysis capabilities and cannot quantify risk levels or generate systematic reports. This makes it difficult for environmental management to achieve preventive intervention, often responding passively after pollution incidents occur, causing irreversible environmental damage. In particular, in large oil and gas field projects, the processing bottleneck of massive data is further highlighted, and traditional computing resources are insufficient to meet real-time needs, leading to decision delays and resource waste. Overall, the existing environmental risk assessment methods have low intelligence level and lack of systematization, which are difficult to adapt to the complex and variable operating environment of oil and gas fields and the increasingly stringent environmental protection requirements. Therefore, it is urgent to develop an efficient and intelligent environmental risk assessment technology to improve the sustainability of green development in oil and gas fields, reduce the probability of pollution incidents, and meet the needs of government regulation and enterprise compliance. SUMMARY
[0005] The purpose of the present application is to solve the problems existing in the prior art, and a digital risk assessment method, platform and equipment for the environment of oil and gas fields are proposed. In order to achieve the above purpose, the embodiments of the present application adopt the following technical solutions:
[0006] In a first aspect, the embodiments of the present application propose a digital risk assessment method for the environment of oil and gas fields, the method comprising:
[0007] Obtaining environmental monitoring data of the oil and gas field, the environmental monitoring data comprising pollutant concentration data and meteorological data;
[0008] Preprocessing the environmental monitoring data;
[0009] Constructing an environmental protection regulation knowledge graph;
[0010] Matching the preprocessed environmental monitoring data with the environmental protection regulation knowledge graph to generate a matching result;
[0011] Evaluating the environmental risk level of the oil and gas field according to the matching result;
[0012] Outputting an environmental risk assessment result according to the environmental risk level.
[0013] Preferably, the step of preprocessing the environmental monitoring data comprises removing noise from the environmental monitoring data and normalizing the environmental monitoring data after removing the noise.
[0014] Preferably, the step of constructing the environmental protection regulation knowledge graph comprises:
[0015] extracting legal provisions, pollutant emission standards and index thresholds related to oil and gas field environmental protection from an environmental protection law database or an official regulation platform, and constructing an environmental protection regulation knowledge graph according to the extracted legal provisions, pollutant emission standards and index thresholds.
[0016] Preferably, the step of matching the pre-processed environmental monitoring data with the environmental protection regulation knowledge graph comprises:
[0017] The corresponding environmental indicators in the pre-processed environmental monitoring data are obtained, and the environmental indicators are associated with the corresponding nodes in the environmental protection regulation knowledge graph using a knowledge graph reasoning rule engine or a matching algorithm based on semantic similarity.
[0018] Preferably, the step of evaluating environmental risk according to the matching result comprises:
[0019] The pre-trained deep neural network model is used to input the associated environmental indicators, compare the environmental indicators with the index thresholds of the corresponding regulations in the environmental protection regulation knowledge graph, and determine the environmental risk level according to the output result.
[0020] Preferably, the step of outputting the environmental risk evaluation result according to the environmental risk level comprises:
[0021] An environmental risk evaluation report is generated according to the environmental risk level; wherein the environmental risk evaluation report includes the environmental risk level of the oil and gas field and the risk warning suggestions corresponding to the environmental risk level.
[0022] Preferably, the method further comprises updating the environmental protection regulation knowledge graph when new relevant environmental protection regulation information is detected.
[0023] Preferably, based on historical environmental monitoring data and real-time environmental monitoring data, the index thresholds defined in the environmental protection regulation knowledge graph are dynamically adjusted through a machine learning algorithm to adapt to changes in environmental conditions and improve the accuracy of environmental risk assessment.
[0024] In a second aspect, an embodiment of the present application proposes a digital risk assessment platform for oil and gas field environment, comprising: an environmental monitoring data acquisition module, a data preprocessing module, an environmental protection regulation knowledge graph construction module, a matching module, a risk assessment module and an output module.
[0025] The environmental monitoring data acquisition module is used to acquire oil and gas field environmental monitoring data, and the environmental monitoring data includes pollutant concentration data and meteorological data of the oil and gas field.
[0026] The data preprocessing module is used to preprocess the environmental monitoring data.
[0027] The environmental protection regulation knowledge graph construction module is used to construct an environmental protection regulation knowledge graph.
[0028] a matching module configured to automatically match the preprocessed environmental monitoring data with the environmental regulation knowledge graph;
[0029] a risk assessment module configured to assess the environmental risk level of the oil and gas field according to the matching result;
[0030] an output module configured to output the environmental risk assessment result according to the environmental risk level.
[0031] In a third aspect, an embodiment of the present application provides an electronic device including a memory and a processor, the memory storing a program for implementing smart risk assessment, and the processor executing the program in sequence to: acquire environmental monitoring data of an oil and gas field, the environmental monitoring data including pollutant concentration data and meteorological data; preprocess the environmental monitoring data; construct an environmental regulation knowledge graph; match the preprocessed environmental monitoring data with the environmental regulation knowledge graph to generate a matching result; assess the environmental risk level of the oil and gas field according to the matching result; and output an environmental risk assessment result according to the environmental risk level.
[0032] Advantages:
[0033] Firstly, by acquiring environmental monitoring data including pollutant concentration and meteorological data and preprocessing, data noise and inconsistency can be effectively eliminated, data quality can be improved, a reliable foundation can be provided for subsequent analysis, and delay and errors caused by manual collection and processing in traditional methods can be avoided. Secondly, by constructing an environmental regulation knowledge graph and matching preprocessed data with it, automated compliance checking of environmental indicators is realized, the intelligent level of assessment is significantly improved, the fragmentation of regulation application and the inefficiency of manual retrieval in the prior art are solved, and risk identification is more comprehensive and timely. Thirdly, according to the matching result, the environmental risk level is assessed and the assessment result is output, which helps to quickly quantify potential risks, supports preventive intervention, reduces the probability of pollution incidents, and promotes green development and sustainable development of oil and gas fields. In addition, this method integrates multi-source data and knowledge graph, promotes the systematicness and closed-loop effect of environmental management, facilitates relevant parties to make scientific decisions and optimize resources, reduces the overall environmental management cost, and improves the environmental compliance and efficiency of oil and gas field operation. BRIEF DESCRIPTION OF DRAWINGS
[0034] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments with reference made to the accompanying drawings. The drawings are for purposes of illustration only and are not intended to be limiting in any respect. Moreover, the use of the same reference symbols in different drawings indicates similar or identical items.
[0035] Figure 1A flowchart of a method for digital risk assessment in an oil and gas field environment according to an embodiment of the present application is shown in FIG. 1.
[0036] Figure 2 A structural diagram of a platform for digital risk assessment in an oil and gas field environment according to an embodiment of the present application is shown in FIG. 2. DETAILED DESCRIPTION
[0037] Embodiments of the present application are described in detail below, examples of which are shown in the accompanying drawings, in which the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application and should not be understood as limiting the present application. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.
[0038] In the first aspect, please refer to Figure 1 A method for digital risk assessment in an oil and gas field environment according to an embodiment of the present application is applied to a platform for digital risk assessment in an oil and gas field environment. The platform for digital risk assessment in an oil and gas field environment can be executed by a computer device with certain computing resources, such as a personal computer (PC), a smart phone, a personal digital assistant (PAD), or a platform server, but is not limited thereto. The embodiment provides a method for digital risk assessment in an oil and gas field environment, which is suitable for environmental management of oil and gas exploration and development sites and risk assessment of various pollution sources such as solid waste, waste gas, waste liquid, and noise. By integrating big data, cloud computing, and artificial intelligence technology, intelligent supervision from the source to the terminal is realized. Specifically, the method comprises the following steps:
[0039] In step S1, environmental monitoring data of the oil and gas field is obtained, including pollutant concentration data and meteorological data.
[0040] In implementation, the environmental monitoring data of the oil and gas field is sourced from Internet of Things (IoT) devices, sensors and monitoring stations deployed on site. These data acquisition modules collect multi-source heterogeneous data in real time, such as pollutant concentration data covering continuous parameters such as H2S concentration in exhaust gas, VOCs emission, combustion efficiency, etc.; discrete parameters such as COD value, oil content, pH value in waste liquid; time series data such as decibel value, frequency spectrum characteristics, duration of noise, etc. At the same time, meteorological data includes wind speed, wind direction, temperature, humidity, rainfall, etc., which are obtained by accessing real-time data from the meteorological bureau or on-site weather stations. For example, in a certain oil and gas field application scenario, the platform integrates multi-source data through DCS system process parameters (collected once per second), environmental monitoring station data (updated every 5 minutes) and equipment management system maintenance records. Data acquisition uses Beidou satellite positioning system coordinate conversion to establish a unified time scale system (Beijing time ± 30 seconds) to ensure spatio-temporal alignment processing. This step provides comprehensive basic data for subsequent analysis, avoiding the problems of data missing or non-real-time in traditional methods.
[0041] Step S2, preprocessing the environmental monitoring data.
[0042] This step aims to clean and standardize the raw data to improve data quality. The specific implementation includes data cleaning specification: using cubic spline interpolation method to repair sensor fault data, implementing truncation processing on data exceeding the range and marking abnormalities, and establishing a data confidence evaluation model (0-1 score). For example, when processing exhaust gas concentration (ppm), it is converted to log value, waste liquid pH value is linearly mapped to 0-14, and noise decibel is standardized to Z-score, the formula is Z=(X-μ) / σ, where Z is the standardized value, μ is the mean of the data set, σ is the standard deviation of the data set, and X is a specific observation value or data point in the data set. For example, a pollutant concentration (such as H2S concentration) in the oil and gas field environmental monitoring data, waste liquid pH value or noise decibel value, etc. By subtracting the mean μ of the data set from X and dividing by the standard deviation σ, it can be converted to a standardized Z value (Z-score), thereby eliminating the dimensional difference, facilitating subsequent data analysis and comparison. At the same time, multi-dimensional correlation analysis is carried out to construct a pollution parameter matrix, including parameter groups of exhaust gas, waste liquid and noise, and the Pearson correlation coefficient matrix is used to calculate the correlation. In addition, the IQR (Interquartile Range) algorithm is used for outlier detection: Q1=25th percentile, Q3=75th percentile, IQR=Q3-Q1, and the outlier range is [Q1-1.5IQR, Q3+1.5IQR]. Through these preprocessing, the platform can process massive heterogeneous data to ensure the accuracy of subsequent steps. In actual cases, after the application of this preprocessing step in a large oil and gas field enterprise, the proportion of human resources has been significantly reduced, and management efficiency has been improved.
[0043] Step S3, build an environmental protection regulation knowledge graph.
[0044] The knowledge graph is based on laws and regulations, standards and specifications, and enterprise rules and regulations, forming a structured knowledge network. The specific process includes extracting relevant content from environmental protection regulation databases or official regulation platforms, such as the official websites of the Ministry of Ecology and Environment and the National Standardization Management Committee, using web crawlers to continuously capture environmental protection laws and regulations, including the "Oil and Natural Gas Exploration Pollutant Discharge Standard" (GB 31570). The digital processing of clauses uses natural language processing techniques for word segmentation, entity recognition, and relationship extraction to establish a three-level knowledge network: the first node is the name of the law and regulation (such as GB31570), the second node is the clause content (such as Article 12 of the Waste Gas Emission Limit), and the third node is the applicable scenario (drilling platform, refinery workshop, etc.). A dynamic updating mechanism sets up a semantic monitoring module that automatically triggers a version comparison program when it detects keywords such as "revision" or "abolition" in policy documents, identifies the changes through a difference analysis algorithm, and updates the timeliness status of the corresponding knowledge nodes. This graph supports benchmarking and scoring, finding vulnerabilities and defects, such as parameter mapping models that build a corresponding relationship between sensor data and regulation indicators, including the direct comparison of H2S detector ppm values with GB31570 waste gas emission limits, and the equivalent sound level calculation formula for noise meters:
[0045] where L eq is the equivalent continuous sound level, representing the average energy level of the noise over a certain time period T, measured in decibels (dB); T is the length of the measurement period, usually measured in seconds, minutes, or hours, representing the sampling time range of the noise data; L i is the sound pressure level (Sound Pressure Level) of the i-th noise sampling point within the time period T, measured in decibels (dB); is the exponential operation and summation of the sound pressure levels of all sampling points, reflecting the cumulative noise energy; is the average noise energy over the time period T, resulting in the average energy per unit time; 10lg(·) is the logarithm of the average energy multiplied by 10, converting it to the equivalent sound level in decibels. In the oil and gas field scenario, this knowledge graph covers a three-dimensional weight distribution model of pollution contribution rate (weight ratio 30%), management controllability (weight ratio 40%), and regulation sensitivity (weight ratio 30%), ensuring that all relevant legal provisions, pollutant emission standards, and index thresholds related to oil and gas field environmental protection are fully integrated.
[0046] Step S4, match the pre-processed environmental monitoring data with the environmental protection regulation knowledge graph to generate a matching result.
[0047] This step uses an intelligent matching engine. First, the scene recognition module automatically matches applicable regulatory provisions based on device geographic coordinates, process flow parameters, and pollutant types, such as automatically associating Article 18 of the Regulations on the Management of Nature Reserves when the GPS coordinates are within 5 kilometers of the boundary of an ecological protection zone and the production stage is during oil testing operations. Second, the parameter mapping model processes the correspondence between sensor data and regulatory indicators, using a fuzzy matching algorithm to calculate the similarity of unstructured data, such as automatically associating the relevant provisions of the Emission Standard for Odorous Pollutants with the "obvious odor in the oil tank area" in the field inspection record, and using a cosine similarity algorithm to match the text feature vector. At the same time, the corresponding environmental indicators in the pre-processed environmental monitoring data are obtained, and the knowledge graph reasoning rule engine or the matching algorithm based on semantic similarity is used to associate the environmental indicators with the corresponding nodes in the environmental protection regulation knowledge graph. For example, feature importance is sorted by random forest algorithm, 30% of the parameters with the highest scores are retained, Spearman rank correlation coefficient analysis is applied to remove redundant parameters, and finally expert verification is organized for ± 15% manual correction. The results of this matching include the corresponding relationship between the associated environmental indicators and the regulatory requirements, providing a basis for risk assessment. In practical applications, this matching in the identification of abnormal waste gas emission in an oil and gas field timely determines the valve sealing problem, reducing the amount of volatilization.
[0048] Step S5: According to the matching results, the environmental risk level of the oil and gas field is evaluated.
[0049] The evaluation uses a pre-trained deep neural network model, taking the associated environmental indicators as input, compares the environmental indicators with the threshold values of the corresponding regulations in the environmental protection regulation knowledge graph, and determines the environmental risk level according to the output result. Specifically, the model uses a CNN-RNN hybrid network, CNN extracts spatial features, and RNN captures time series features. The input layer receives pollution data such as exhaust gas volume, waste liquid concentration, and noise intensity, the hidden layer uses a convolutional neural network to extract spatial features (such as pollution distribution heat map, three-dimensional concentration field generated by Kriging interpolation), and a recurrent neural network to capture time series features (such as trend prediction and periodicity analysis, using LSTM to capture long-term dependencies). The output layer generates intelligent diagnosis results such as pollution source and treatment path suggestion. The evaluation steps include a factor screening process: initial screening by random forest algorithm, fine screening by Spearman rank correlation coefficient, and expert verification correction. The cloud map generation technology uses an improved Kriging interpolation algorithm, sets the semi-variogram parameter γ(h) = 0.3 + 0.7[1-exp(-h / 250)], h is the spatial distance, and 250 meters is the optimal range. Multidimensional data fusion establishes a five-layer superposition model: base satellite map (transparency 100%), pollution heat cloud map (60%), equipment vector icon (80%), pipe network line layer (50%), and meteorological dynamic flow field (40%). The dynamic rendering mechanism data refresh cycle is 5 seconds, the color gradient algorithm is HSL color space interpolation, and the abnormal flicker frequency is 2Hz. Pollution source analysis applies back propagation neural network (BPNN) to establish a pollution diffusion model D is the diffusion coefficient, v is the wind speed vector, and Q is the source intensity. The responsibility matrix determines a four-dimensional responsibility tracing model, including the owner party (rationality of equipment selection, weight 30%), the construction party (installation qualification certificate, 40%), the treatment party (integrity of operation records, 20%), and the supervision party (inspection frequency compliance rate, 10%). The risk level is divided into first level (strategic level, such as 80% of the pollutant discharge exceeding the limit value of GB31570), second level (tactical level, 15%-30% of pollution load), and third level (execution level, <10% of pollution load). Through this model, the environmental protection violation risk in a certain refinery application is significantly reduced.
[0050] Step S6, outputting the environmental risk assessment result according to the environmental risk level.
[0051] The step generates an environmental risk assessment report, which includes displaying the environmental risk level of the oil and gas field and the risk warning suggestion corresponding to the environmental risk level. For example, a first-level risk is immediately disposed of by the general manager, a second-level risk is disposed of by the department head within 2 hours, and a third-level risk is disposed of by the team leader within 24 hours. The report adopts visual presentation: a three-dimensional situation map (spatial dimension: factory area plan map marked with exceeding standard points, time dimension: trend curve showing historical score changes, and index dimension: radar chart showing compliance of each dimension). The intelligent report generation application automatically outputs using natural language generation technology, for example, “The environmental protection comprehensive score this quarter is X points, the main progress is that the oil-containing sludge disposal compliance rate is improved, and the key attention is that the torch combustion efficiency is decreased.” The output supports cloud technology, realizes a four-dimensional integrated management mode of the owner, the construction party, the processing party, and the supervision party, and ensures closed-loop management. In actual cases, the output helps enterprises to find and fill gaps in time, realize problem-to-point, responsibility-to-post, and quantitative evaluation.
[0052] Preferably, the step of preprocessing the environmental monitoring data comprises removing noise of the environmental monitoring data and normalizing the environmental monitoring data after removing the noise. Specifically, noise removal uses an outlier detection algorithm, such as the IQR method combined with principal component analysis (PCA) to reconstruct the feature space of the cleaned data, and the formula is PCA dimension reduction processing. At the same time, the normalization processing includes linear normalization (such as mapping the pH value to the 0-14 interval), Z-score standardization (noise decibel value), and logarithmic transformation (high skewness of pollutant concentration data). The spatio-temporal alignment processing uses a dynamic time warping (DTW) algorithm: D(i,j) = dist(x i ,y j )+min(D(i-1,j),D(i,j-1),D(i-1,j-1)), where D(i,j) is the value of the cumulative distance matrix at position (i,j), representing the minimum total cost of aligning the first i elements of sequence X with the first j elements of sequence Y; dist(x i ,y j ) is a local distance function that calculates the similarity difference between the i-th element x i of sequence X and the j-th element y j of sequence Y; x i is the i-th element of sequence X, representing the observation value at time point i; y jFor the jth element of sequence Y, represents the observation value at time point j; min(D(i-1,j),D(i,j-1),D(i-1,j-1)) is the minimum value selected from three adjacent position cumulative distances, corresponding to deletion, insertion and matching operations respectively; i and j are matrix indexes, representing the current positions of sequences X and Y. Sequences X and Y are two time series data sequences (time series) that need to be aligned, which represent observation values collected at different time points, and are used for spatiotemporal alignment of time series data to ensure the consistency of multi-source environmental monitoring data (such as pollutant concentration, meteorological data, etc.) in the time dimension. In the oil and gas field special feature network, deep feature mining uses an adaptive crawling strategy, priority score = 0.4 real-time abnormality + 0.3 historical failure rate + 0.2 environmental sensitivity + 0.1 equipment criticality, and an incremental learning mechanism uses the elastic weight consolidation (EWC) algorithm to prevent catastrophic forgetting. This preferred method ensures high-quality data and is suitable for massive pollution data processing.
[0053] Preferably, the step of constructing the environmental protection regulation knowledge graph comprises: extracting legal provisions, pollutant emission standards and index thresholds related to oil and gas field environmental protection from an environmental protection regulation database or an official regulation platform, and constructing the environmental protection regulation knowledge graph according to the extracted legal provisions, pollutant emission standards and index thresholds. In specific implementation, a network crawler is deployed to collect PDF and Word format documents through multi-source data acquisition, and a smart parsing algorithm is used to identify structured data. The knowledge graph is constructed in three levels, and is dynamically updated through semantic monitoring and difference analysis. The grading theory is based on the principles of petroleum engineering environment and system dynamics theory, and a three-dimensional weight distribution model is established. The screening conditions include first-level influence factors (80% of the pollutant emission amount exceeds the limit value of GB31570, involving HW08 hazardous waste), second-level (pollution load 15%-30%, correlation coefficient > 0.6), and third-level (pollution load < 10%). This preferred construction method makes the graph more targeted and supports specific scenarios in oil and gas fields.
[0054] Preferably, the step of matching the preprocessed environmental monitoring data with the environmental protection regulation knowledge graph comprises: obtaining the corresponding environmental indicators in the preprocessed environmental monitoring data, and using a knowledge graph reasoning rule engine or a matching algorithm based on semantic similarity to associate the environmental indicators with the corresponding nodes in the environmental protection regulation knowledge graph. Specifically, the matching engine includes scene recognition (such as GPS and regulation association), parameter mapping (sensor and index correspondence table), and fuzzy matching (cosine similarity). The factor screening process is: preliminary screening with random forest, fine screening with Spearman, and expert verification. The cloud map generation uses Kriging interpolation and a five-layer model for multi-dimensional fusion. This preferred matching improves the automation level and reduces manual intervention.
[0055] Preferably, the step of evaluating the environmental risk according to the matching result comprises: using a pre-trained deep neural network model, inputting the associated environmental indicators, comparing the environmental indicators with the indicator thresholds of the corresponding regulations in the environmental protection regulation knowledge graph, and determining the environmental risk level according to the output result. The specific model is a CNN-RNN hybrid, and the space-time feature extraction is as follows: CNN constructs a pollution heat map, LSTM constructs a 24-hour time window sample, and an attention mechanism is used for cross-modal fusion. The model is trained and optimized by using a composite loss function, progressive training (pre-training historical data for 3 years, fine-tuning recent 6 months, online learning real-time data), and Bayesian optimization of hyperparameters. The pollution tracing BPNN diffusion model and the four-dimensional judgment of the responsibility matrix are used. The preferred evaluation improves the accuracy and supports the prediction modeling.
[0056] Preferably, the step of outputting an environmental risk evaluation result according to the environmental risk level comprises: generating an environmental risk evaluation report according to the environmental risk level; and the environmental risk evaluation report includes the environmental risk level of the oil and gas field and the risk warning suggestion corresponding to the environmental risk level. The specific visualization three-dimensional situation map and intelligent report generation, and the disposal scheme such as immediate disposal of the first level promote the decision optimization.
[0057] Preferably, the method further comprises: updating the environmental protection regulation knowledge graph when new relevant environmental protection regulation information is detected. The specific version comparison algorithm processes changes to ensure the timeliness of the graph. The preferred dynamic maintenance adapts to changes in regulations.
[0058] Preferably, based on historical environmental monitoring data and real-time environmental monitoring data, the indicator thresholds defined in the environmental protection regulation knowledge graph are dynamically adjusted by a machine learning algorithm to adapt to changes in environmental conditions and improve the accuracy of environmental risk evaluation. The specific ARIMA model is used to predict trends, Bayesian networks are used to calculate responsibility probabilities, and evidence theory is used to fuse multi-modal information. The dynamic weight adjustment improves the indicator weight based on historical violation data. The preferred adjustment makes the evaluation more robust and suitable for complex environments.
[0059] By the method of the embodiment, the oil and gas field environmental risk assessment is intelligent and efficient, such as dimensional AI diagnosis and four-dimensional management mode, ensuring cloud supervision and management from the source of storage, transportation and disposal. In practical application, the method performs well in reducing pollution and improving management efficiency, such as adjusting the process to improve the compliance rate in waste liquid treatment and reducing volatile emissions in waste gas emission. The technical architecture of the embodiment adopts a "cloud + edge" distributed framework, and the functional modules include big data computing, intelligent diagnosis and cloud technology, supporting neural network algorithm optimization such as ST-Conv three-dimensional convolution structure. The software design process forms a closed loop from data acquisition to result output. The neural network structure diagram includes an input layer (pollution data), a hidden layer (CNN spatial features, RNN time sequence features) and an output layer (diagnosis results). The spatio-temporal feature fusion improves the traceability accuracy through attention mechanism, solving the limitation of single feature. The technical innovations of the method include multi-dimensional AI diagnosis, big data cloud computing fusion, four-dimensional management mode and neural network optimization, which are suitable for governments, enterprises and operators, and promote clean energy green development.
[0060] In the second aspect, please refer to Figure 2 A digital risk assessment platform for oil and gas field environment is provided for the embodiment, which is suitable for environmental management of oil and gas field exploration and development operation sites, and conducts intelligent risk assessment on pollution sources such as waste solid, waste gas, waste liquid and noise. The platform adopts a "cloud + edge" distributed architecture, integrates big data analysis, cloud computing, artificial intelligence (AI) and Internet of Things (IoT) technologies, and realizes intelligent management of oil and gas field pollution in all directions. The platform includes an environmental monitoring data acquisition module, a data preprocessing module, an environmental regulation knowledge graph construction module, a matching module, a risk assessment module and an output module.
[0061] The environmental monitoring data acquisition module is used to acquire oil and gas field environmental monitoring data, including pollutant concentration data and meteorological data of the oil and gas field. The module acquires multi-source data in real time through IoT devices, sensors and monitoring stations deployed on site. For example, the pollutant concentration data includes H2S concentration in waste gas, VOCs emission, combustion efficiency, etc.; COD value, oil content, pH value, etc. in waste liquid; decibel value, frequency spectrum characteristics, etc. of noise. At the same time, meteorological data includes wind speed, wind direction, temperature, humidity, etc., which are obtained by accessing meteorological bureaus or on-site stations. The module supports integration of DCS system process parameters (1 per second), environmental monitoring station data (updated every 5 minutes) and equipment maintenance records, uses Beidou satellite positioning and unified time scale system (Beijing time ± 30 seconds) to ensure spatio-temporal alignment. In practical application, the module efficiently collects massive data in a certain oil and gas field scene, avoiding the delay of traditional manual inspection.
[0062] A data preprocessing module is configured to preprocess the environmental monitoring data. The module cleans and standardizes the data, including noise removal and normalization processing. Specifically, cubic spline interpolation is used to repair sensor failure data, data exceeding the range is truncated and marked as abnormal, and a 0-1 confidence model is established. The exhaust gas concentration is converted to a log value, the waste liquid pH value is linearly mapped to 0-14, and the noise is Z-score standardized Z=(X-μ) / σ, where Z is the standardized value, μ is the mean of the data set, σ is the standard deviation of the data set, and X is a specific observation value or data point in the data set. Multidimensional correlation analysis constructs a pollution parameter matrix, and the Pearson coefficient is calculated. The IQR algorithm is used for outlier detection: IQR=Q3-Q1, and the abnormal range is [Q1-1.5IQR, Q3+1.5IQR]. Principal component analysis (PCA) reconstructs the feature space, and dynamic time warping (DTW) processes the time series data: D(i,j)=dist(x i ,y j )+min(D(i-1,j),D(i,j-1),D(i-1,j-1)), where D(i,j) is the value of the cumulative distance matrix at position (i,j), representing the minimum total cost of aligning the first i elements of sequence X with the first j elements of sequence Y; dist(x i ,y j ) is a local distance function that calculates the similarity difference between the ith element x i of sequence X and the jth element y j of sequence Y; x i is the ith element of sequence X, representing the observation value at time point i; y j is the jth element of sequence Y, representing the observation value at time point j; min(D(i-1,j),D(i,j-1),D(i-1,j-1)) is the minimum value selected from the three adjacent position cumulative distances, corresponding to deletion, insertion, and matching operations, respectively; i and j are matrix indices, representing the current positions of sequences X and Y. Sequences X and Y are two time series data sequences (time series) that need to be aligned, representing observation values collected at different time points, used to align time series data in space and time to ensure the consistency of multi-source environmental monitoring data (such as pollutant concentration, meteorological data, etc.) in the time dimension. This module improves data quality, and after application in a large oil and gas field, management efficiency is significantly improved.
[0063] An environmental regulation knowledge graph construction module is used to construct an environmental regulation knowledge graph. This module extracts relevant content from an environmental protection regulation database or official regulation platform, such as the "Petroleum and Natural Gas Exploitation Pollutant Discharge Standard" (GB31570), uses a web crawler to capture PDF / Word documents, and uses NLP technology for word segmentation and entity recognition. A three-level knowledge network is constructed: the first level is the legal regulation name, the second level is the clause content, and the third level is the applicable scenario. A dynamic update mechanism monitors keywords such as "revision" to trigger version comparison. The grading basis includes pollution contribution rate (30%), management controllability (40%), and regulation sensitivity (30%). Screening conditions: first-level factors exceeding 80% of GB31570 limits, involving HW08 hazardous waste; secondary pollution load 15%-30%; and tertiary <10%. The parameter mapping library corresponds to sensors and indicators, such as H2S ppm compared to waste gas limits. This module ensures that the graph is targeted at oil and gas field scenarios and supports benchmarking.
[0064] A matching module is used to automatically match pre-processed environmental monitoring data with the environmental regulation knowledge graph. This module uses an intelligent matching engine, and scenario recognition is based on GPS / process parameter matching regulations, such as the "Nature Reserve Management Regulations" within the ecological zone. Parameter mapping processes sensor-indicator correspondence, and fuzzy matching uses cosine similarity to associate non-structured data. The knowledge graph reasoning rule engine or semantic similarity algorithm is used to associate environmental indicators with graph nodes. Factor selection: random forest screening for importance in the top 30%, Spearman for redundancy removal >0.8, and expert correction ±15%. Cloud map generation uses Kriging interpolation γ(h) = 0.3 + 0.7[1-exp(-h / 250)], and a five-layer superposition model (base 100%, pollution 60%, etc.). This matching identifies the source in waste gas anomaly cases in a timely manner.
[0065] A risk assessment module is used to assess the environmental risk level of oil and gas fields based on matching results. This module uses a pre-trained deep neural network (CNN-RNN hybrid), inputs associated indicators, compares with graph thresholds, and outputs risk levels. CNN extracts spatial features (such as heat maps), and RNN captures timing (LSTM 24-hour window). Cross-modal fusion attention mechanism. Model training with compound loss, progressive (3-year history pre-training, latest 6-month fine-tuning, real-time online). Pollution traceability BPNN diffusion model The responsibility matrix is four-dimensional: owner 30%, construction 40%, treatment 20%, and supervision 10%. Risk classification: strategic level, tactical level, and execution level. Based on historical / real-time data, machine learning (ARIMA prediction, Bayesian network) dynamically adjusts thresholds. This module reduces the risk of violations in oil refineries.
[0066] An output module is configured to output an environmental risk assessment result according to the environmental risk level. The module generates a report including the risk level and warning suggestions, such as immediate overall manager-led for a first-level. A three-dimensional situation map (spatial over-standard points, time trend, and index radar chart) is visualized, and an NLG automatically outputs a report “comprehensive score X points, progress items: sludge compliance rate improved, and attention items: combustion efficiency decreased”. Four-dimensional management closed-loop is supported to ensure the cooperation of owners / construction / treatment / supervision. The output promotes preventive intervention and realizes green development.
[0067] The platform forms a software design process through module cooperation: collection-preprocessing-mapping-matching-evaluation-output. The neural network structure is: input pollution data, hidden CNN / RNN, and output diagnosis. Dimensional AI diagnosis and big data fusion have significant advantages in reducing pollution (such as reducing volatile waste gas by 30%) and improving efficiency (such as reducing manpower by 40%). It is suitable for governments / enterprises and promotes environmental protection in oil and gas field construction.
[0068] In the third aspect, the embodiment provides an electronic device suitable for an oil and gas field environmental risk assessment scenario, which can be deployed in a cloud server, an edge computing node, or a field terminal device, and supports intelligent management of pollution sources such as waste solids, waste gases, waste liquids, and noise. The electronic device includes a memory and a processor, wherein the memory stores a program for implementing intelligent risk assessment, and the processor executes the following steps in sequence according to the program:
[0069] Step S1, environmental monitoring data of the oil and gas field is acquired, and the environmental monitoring data includes pollutant concentration data (such as H2S concentration, COD value, and noise decibel) and meteorological data (such as wind speed and temperature). The processor collects multi-source data in real time through an integrated IoT interface, supports a DCS system (1 time per second) and an environmental monitoring station (updated every 5 minutes), and uses Beidou positioning to ensure spatio-temporal alignment.
[0070] Step S2, the environmental monitoring data is preprocessed. The processor executes cleaning algorithms, such as three times spline interpolation to repair faulty data, IQR abnormal value detection (IQR=Q3-Q1), and normalization (Z=(X-μ) / σ), including log conversion of waste gas concentration and linear mapping of pH value, to improve data quality.
[0071] Step S3, an environmental protection regulation knowledge graph is constructed. The processor extracts standards such as GB31570 from a database, uses NLP word segmentation to construct a three-level network (laws and regulations-clauses-scenarios), and uses a dynamic update mechanism to monitor policy changes, supporting a three-dimensional weight model (pollution contribution rate 30%, management controllability 40%, and regulation sensitivity 30%).
[0072] Step S4, the pre-processed environmental monitoring data is matched with the environmental protection regulation knowledge graph to generate a matching result. The processor uses an inference engine and a cosine similarity algorithm to associate indicators, such as matching the GPS scene with the “Nature Reserve Management Regulations” and mapping the H2S ppm parameter with the waste gas limit value.
[0073] Step S5, according to the matching result, the environmental risk level of the oil and gas field is evaluated. The processor runs the CNN-RNN model to extract the space-time features (CNN heat map, LSTM time sequence), compares the threshold value, and traces the BPNN diffusion model Classification level one (over 80%), level two (15%-30% load), and level three (<10%), with dynamic threshold adjustment using ARIMA.
[0074] Step S6, according to the environmental risk level, the environmental risk assessment result is output. The processor generates a report including level visualization (three-dimensional situation map, radar chart) and early warning suggestions (such as immediate disposal for level one), supporting four-dimensional management closed loop.
[0075] The device adopts a distributed architecture, the memory optimizes mass data processing, and the processor integrates AI acceleration to improve efficiency. In a certain oil and gas field application, pollution incidents are reduced, and green development is achieved.
[0076] In the description of the present specification, the description of the terms “one embodiment”, “some embodiments”, “example”, “specific example”, “one implementation”, “one preferred implementation” or “some examples” and the like means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.
[0077] Although the embodiments of the present application have been shown and described, those skilled in the art can understand that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and purposes of the present application, and the scope of the present application is defined by the claims and their equivalents.
Claims
1. A method for digital risk assessment for oil and gas field environment characterized in that, The method comprises: acquiring environmental monitoring data of the oil and gas field, the environmental monitoring data comprising pollutant concentration data and meteorological data; preprocessing the environmental monitoring data; constructing an environmental protection regulation knowledge graph; matching the preprocessed environmental monitoring data with the environmental protection regulation knowledge graph to generate a matching result; evaluating the environmental risk level of the oil and gas field according to the matching result; outputting an environmental risk evaluation result according to the environmental risk level.
2. The digital risk assessment method of claim 1, wherein, The preprocessing step of the environmental monitoring data comprises removing noise from the environmental monitoring data and normalizing the environmental monitoring data after removing the noise.
3. The digital risk assessment method of claim 2, wherein, The step of constructing an environmental protection regulation knowledge graph comprises: extracting legal provisions, pollutant emission standards and index thresholds related to oil and gas field environmental protection from an environmental protection regulation database or an official regulation platform, and constructing the environmental protection regulation knowledge graph according to the extracted legal provisions, pollutant emission standards and index thresholds.
4. The digital risk assessment method of claim 3, wherein, The step of matching the preprocessed environmental monitoring data with the environmental protection regulation knowledge graph comprises: obtaining corresponding environmental indicators in the preprocessed environmental monitoring data, and using a knowledge graph reasoning rule engine or a matching algorithm based on semantic similarity to associate the environmental indicators with corresponding nodes in the environmental protection regulation knowledge graph.
5. The digital risk assessment method of claim 4, wherein, The step of evaluating the environmental risk according to the matching result comprises: using a pre-trained deep neural network model, taking the associated environmental indicators as input, comparing the environmental indicators with the index thresholds of the corresponding regulations in the environmental protection regulation knowledge graph, and determining the environmental risk level according to the output result.
6. The digital risk assessment method of claim 5, wherein, The step of outputting an environmental risk evaluation result according to the environmental risk level comprises: generating an environmental risk evaluation report according to the environmental risk level; wherein the environmental risk evaluation report includes the environmental risk level of the oil and gas field and risk warning suggestions corresponding to the environmental risk level.
7. The digital risk assessment method of claim 5, wherein, The method further comprises updating the environmental protection regulation knowledge graph when new relevant environmental protection regulation information is detected.
8. The digital risk assessment method of claim 7, wherein, Based on historical environmental monitoring data and real-time environmental monitoring data, the index thresholds defined in the environmental protection regulation knowledge graph are dynamically adjusted through a machine learning algorithm to adapt to changes in environmental conditions and improve the accuracy of environmental risk assessment.
9. A digital risk assessment platform for oil and gas field environment characterized in that, It comprises: an environmental monitoring data acquisition module, a data preprocessing module, an environmental protection regulation knowledge graph construction module, a matching module, a risk assessment module and an output module; the environmental monitoring data acquisition module is used to acquire environmental monitoring data of the oil and gas field, the environmental monitoring data comprising pollutant concentration data and meteorological data of the oil and gas field; the data preprocessing module is used to preprocess the environmental monitoring data; the environmental protection regulation knowledge graph construction module is used to construct an environmental protection regulation knowledge graph; the matching module is used to automatically match the preprocessed environmental monitoring data with the environmental protection regulation knowledge graph; the risk assessment module is used to evaluate the environmental risk level of the oil and gas field according to the matching result; the output module is used to output an environmental risk evaluation result according to the environmental risk level.
10. An electronic device comprising a memory and a processor, characterized in that, The memory stores a program for implementing intelligent risk assessment, and the processor sequentially executes the following according to the program: acquiring environmental monitoring data of an oil and gas field, the environmental monitoring data including pollutant concentration data and meteorological data; preprocessing the environmental monitoring data; constructing an environmental protection regulation knowledge graph; matching the preprocessed environmental monitoring data with the environmental protection regulation knowledge graph to generate a matching result; evaluating the environmental risk level of the oil and gas field according to the matching result; outputting an environmental risk assessment result according to the environmental risk level.
Citation Information
Cited By
Laboratory hazard source identification method based on AI
CN121834391A
