Oilfield station unattended intelligent operation and maintenance centralized control method and system based on intelligent agent

Abstract

The invention discloses an intelligent agent-based oil field station unattended intelligent operation and maintenance centralized control method and system, and belongs to the technical field of oil field station intelligent operation and maintenance management. The system comprises a data acquisition module, a data summarization decision module, a station coordination execution module, an emergency intervention module and an output module; pressure sensors, temperature sensors, flow sensors, liquid level sensors and water content sensors are arranged at oil field inter-well stations, transfer stations and united stations, and crude oil conveying equipment data of the inter-well stations, the transfer stations and the united stations are collected in real time. According to the intelligent agent-based unattended intelligent operation and maintenance centralized control method and system for the oil field station, station correlation analysis, instant optimization adjustment strategy and target optimization adjustment strategy generation, station specific parameter execution, feedback and optimization adjustment strategy fine adjustment and decision process recording can be realized, and a large model and a small model are closely combined; and an up-and-down collaborative hierarchical control framework is formed, so that the practicability and functionality are enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent operation and maintenance management of oilfield stations, and specifically relates to an unattended intelligent operation and maintenance centralized control method and system for oilfield stations based on intelligent agents. Background Art

[0002] In the current oilfield industry, the requirements for production efficiency, energy utilization rate, and safety are continuously increasing. However, the traditional manual monitoring and point-to-point control methods are increasingly showing their limitations when facing the complex working conditions of multiple devices. There are various types of equipment in oilfield stations, including but not limited to heating furnaces, pumps, water blending equipment, and free water separators. These devices have dynamically changing working conditions during operation, and there is a high degree of coupling and complex correlation among them; Oilfield well-to-well stations, transfer stations, and gathering stations involve many crude oil transportation equipment, and their operating states are affected by various complex parameters such as pressure, temperature, flow rate, liquid level, and water cut. In the past, it was difficult to achieve efficient integration and accurate analysis based on manual detection and analysis of the massive data generated by these devices. It was impossible to make reasonable operation and maintenance decisions in a timely manner according to data changes, resulting in frequent equipment failures, affecting the overall production efficiency and economic benefits of the oilfield. It was impossible to immediately identify and adjust abnormal parameters, often resulting in abnormal data synchronization at the source site affecting the production activities of associated sites; This case proposes an unattended intelligent operation and maintenance centralized control method and system for oilfield stations based on intelligent agents to solve the above technical problems. Summary of the Invention

[0003] The present invention aims to solve at least one of the technical problems existing in the prior art. For this purpose, the present invention proposes an unattended intelligent operation and maintenance centralized control method and system for oilfield stations based on intelligent agents, and solves the above technical problems by improving the detection method and processing method.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions: An unattended intelligent operation and maintenance centralized control system for oilfield stations based on intelligent agents, including a data acquisition module, a data summary and decision-making module, a site coordination and execution module, an emergency intervention module, and an output module; The data acquisition module, by deploying pressure, temperature, flow rate, liquid level, and water cut sensors at oilfield well-to-well stations, transfer stations, and gathering stations, real-time collects the crude oil transportation equipment data of well-to-well stations, transfer stations, and gathering stations; The data aggregation and decision-making module receives the real-time crude oil transportation equipment data of the subordinate well-to-station, transfer station, and joint station obtained by the data acquisition module. Based on the data changes of the real-time sensors in the well-to-station and transfer station, it generates cross-station instant optimization and adjustment strategies for the transfer station and joint station through a large model. At the same time, based on the natural language processing technology in the large model, it semantically analyzes the data in the oilfield production logs and operation instructions, and generates target optimization and adjustment strategies for different stations through the analysis model combined with historical data, including combustion control, pump speed adjustment, and water injection ratio optimization, and transmits them to the site coordination and execution module; The site coordination and execution module, based on the instant optimization and adjustment strategies and target optimization and adjustment strategies generated by the data aggregation and decision-making module for each site, combines the data optimization small models deployed at each site for in-depth optimization and fine control of specific equipment and functions, executes the corresponding adjustment strategies for each site respectively, and at the same time, based on the data changes during the execution of the small models at different sites, further fine-tunes the large model optimization and adjustment strategies according to the real-time monitoring data.

[0005] Furthermore, when the data aggregation and decision-making module generates instant optimization and adjustment strategies and target optimization and adjustment strategies through the large model, it includes the following steps: Receive the real-time pressure, temperature, flow rate, liquid level, and water content data of the subordinate well-to-station and transfer station obtained by the data acquisition module. Based on the abnormal changes in the real-time data, identify the abnormal well-to-station and transfer station. According to the crude oil transportation relationship network of different abnormal well-to-station and transfer station, generate cross-station instant optimization and adjustment strategies for the relevant transfer station and joint station through the LMM large model; Based on the natural language processing technology in the large model, semantically analyze the data in the oilfield production logs and updated operation instructions of the well-to-station, transfer station, and joint station, and generate target optimization and adjustment strategies for different stations through the analysis model combined with historical data, including combustion control, pump speed adjustment, and water injection ratio optimization, and transmit them to the site coordination and execution module.

[0006] Furthermore, the specific steps for constructing the LMM large model for the subordinate well-to-station, transfer station, and joint station are as follows: Statistical information of the subordinate well-to-station, transfer station, and joint station of the system, including the site location and the upstream and downstream relationships of the sites, establish a relationship network of the well-to-station, transfer station, and joint station, and collect historical sensor data of different sites respectively, including pressure data, temperature data, flow rate data, liquid level data, and crude oil water content data, and calculate the mean value in the historical sensor data of the corresponding equipment at different sites And the standard value , through the 3sigma principle, set the normal range of the sensor data of different sites ; Based on the normal ranges of sensor data from different sites, record the abnormal parameter situations of the sensor data at each site, including those below and above the normal ranges, which are respectively recorded as low anomalies and high anomalies. Collect the adjustment strategies of associated sites in the relationship network under the corresponding abnormal parameter situations, extract the data of the upstream and downstream sites of the abnormal site within the same time window, form an associated adjustment strategy event set, and combine the abnormal parameter situations with the adjustment strategies of the associated sites into training samples. Each sample contains an input and an output. Select an open-source large language model as the framework and make specific modifications according to the characteristics of the oilfield site data. Add an input layer to process structured abnormal parameter information and set an output layer to generate the adjustment strategy text corresponding to the input layer. Input the constructed training samples into the model in batches, perform forward propagation to calculate the model output, calculate the loss value according to the loss function, and update the model parameters through the Adam optimization algorithm. Repeat multiple rounds until the model converges or reaches the preset number of training rounds.

[0007] Furthermore, based on the real-time data abnormal changes, identify abnormal well-to-station and transfer stations. According to the crude oil transportation relationship networks of different abnormal well-to-station and transfer stations, generate cross-site instant optimization adjustment strategies for relevant transfer stations and joint stations through the LMM large model, including the following steps: Receive the real-time pressure, temperature, flow rate, liquid level, and water content data of the subordinate well-to-station and transfer stations obtained by the data acquisition module, and calculate the mean value according to the time window. Based on the normal ranges of the sensor data from different sites, determine the mean value data that is not within the range. When the sensor mean value data is less than the normal range of the current sensor data of the corresponding site, determine that the current sensor data of the current site is a low anomaly. When the sensor mean value data is higher than the normal range of the current sensor data of the corresponding site, determine that the current sensor data of the current site is a high anomaly; One-hot encode the abnormal parameter situations of the abnormal well-to-station and transfer stations and the location information of the sites in the relationship network. Generate cross-site instant optimization adjustment strategies for relevant transfer stations and joint stations through the LMM large model, obtain the probability of each adjustment strategy through the softmax function, and transmit the instant optimization adjustment strategy with the maximum probability to the site coordination execution module.

[0008] Furthermore, the generation of target optimization adjustment strategies for different sites based on the natural language processing technology in the large model includes the following steps: Collect production log and operation instruction data, including operation scenarios and equipment information. Use the labeled production log and operation instruction data to train the BiLSTM-CRF model. Analyze the operation instructions using a dependency parsing tool to construct a dependency syntax tree to clarify the dependency relationships between words. Based on the dependency syntax tree and the identified entities, extract the cause, action, and target parameters in the causal relationship to form a structured triple, where triple = (cause, action, target parameter). Match the appropriate target optimization and adjustment strategies suitable for the subordinate sites according to the generated triples and transmit them to the site coordination execution module.

[0009] Furthermore, for the immediate optimization and adjustment strategies and target optimization and adjustment strategies generated by each site by the data aggregation and decision-making module, use a small model to execute the corresponding adjustment strategies for each site respectively, including the following steps: For different subordinate sites, based on the historical data of the equipment in each site, use a neural network to establish a data adjustment prediction model for the corresponding equipment respectively. According to the immediate optimization and adjustment strategies and target optimization and adjustment strategies transmitted by the data aggregation and decision-making module, determine the optimized adjustment sites and equipment involved. Based on the parameter fluctuation value of the impact source, calculate the data adjustment value through the corresponding equipment data adjustment prediction model of the involved sites. The specific steps are as follows: For different equipment in different wellhead stations, transfer stations, and joint stations, use a long short-term memory network with the mean square error as the loss function combined with the Adam optimizer to construct an equipment parameter adjustment prediction model for the corresponding site respectively. Based on the immediate optimization and adjustment strategies and target optimization and adjustment strategies of the data aggregation and decision-making module, Perform strategy-equipment matching, where are the strategy feature vector and the equipment historical data feature vector respectively; Determine the key parameter fluctuation source through sensitivity analysis and calculate the parameter for the target gradient, where: ; where, represents the output value of the th hidden layer in the neural network, is the optimization target variable, is the equipment parameter, represents the total number of hidden layers in the neural network; Input the strategy target and the current parameter into the model to solve for the parameter adjustment amount , combined with the equipment safety threshold, where the input model is: ; where, is the parameter adjustment amount, that is, the value of the change in the device parameters to be calculated. is the neural network prediction model, that is, the input parameters and the predicted target value after input. represents the regularization coefficient, and then constructs an optimization problem with constraints: ; Among them, respectively represent the physical security thresholds of the device parameters. The Lagrange multiplier method is used to finally obtain the optimal adjustment amount, and the parameters are adjusted at the corresponding sites through the optimal adjustment amount.

[0010] Furthermore, based on the data changes during the execution of the small models at different sites, the optimization adjustment strategy of the large model is further fine-tuned according to the real-time monitoring data, including the following steps: During the parameter adjustment process of the small models, each site's small model uploads the real-time data after execution to the central data pool, and calculates the deviation between the actual value and the target value , respectively represent the actual value and the target value, and the deviation is decomposed into the contributions of each associated site through the chain rule: ; Among them, represents the key parameter of the th site, is the actual adjustment amount. Based on the real-time feedback data, the policy value function of the large model is updated , the influence matrix between sites is defined and the constrained optimization problem is reconstructed. At the same time, the adjusted policy parameters are verified to ensure that the policy parameters are between the safety thresholds, where the policy parameters . If the policy parameters are out of bounds, the projection method is used for correction; At the same time, record the policy parameters and the corresponding performance indicators after each adjustment to build a historical version library, and re-solve the global optimization problem every fixed time window.

[0011] Furthermore, for the emergency intervention module, in the data aggregation decision module and the site coordination control module, when the model reports an error, the system will automatically generate an event report and notify the on-duty operation and maintenance personnel through text messages and APP notifications. The content of the generated event report includes the error code and number, the names of the sensors involved, and the site information involved in the error; The output module, after each adjustment and optimization of the site equipment parameters by the data summary decision module and the site coordination execution module, uses the initial state before equipment adjustment and relevant environmental data as the starting point for reasoning, records the rules and algorithms based on which the data summary decision module formulates the optimization strategy, and records each step in the process of the site coordination execution module implementing the optimization strategy, including the order of parameter adjustment, the adjustment range, and the feedback data of the equipment after each adjustment, and outputs a detailed text description of the complete optimization strategy reasoning process data.

[0012] An agent-based unattended intelligent operation and maintenance centralized control method for oilfield stations includes the following steps: S1. By deploying pressure, temperature, flow rate, liquid level, and water cut sensors at oilfield inter-well stations, transfer stations, and gathering stations, real-time collection of crude oil transportation equipment data at inter-well stations, transfer stations, and gathering stations is carried out, and real-time crude oil transportation equipment data of subordinate inter-well stations, transfer stations, and gathering stations is received. Based on the data changes of real-time sensors at inter-well stations and transfer stations, cross-site instant optimization adjustment strategies and target optimization adjustment strategies for transfer stations and gathering stations are generated through the LMM large model; S2. Based on the instant optimization adjustment strategies and target optimization adjustment strategies of each site, combined with the data optimization small models deployed at each site for in-depth optimization and fine control of specific equipment and functions, the corresponding adjustment strategies are respectively implemented for each site. At the same time, based on the data changes during the implementation of the small models of different sites, the large model optimization adjustment strategy is further fine-tuned according to the real-time monitoring data; S3. When the model reports an error, an event report is automatically generated and the on-duty operation and maintenance personnel are notified by means of text messages and APP notifications. The content of the generated event report includes the error code and number, the name of the involved sensor, and the site information involved in the error; S4. After each adjustment and optimization of the site equipment parameters, the initial state before equipment adjustment and relevant environmental data are used as the starting point for reasoning, the rules and algorithms based on which the optimization strategy is formulated are recorded, and each step in the process of the site implementing the optimization strategy is recorded, including the order of parameter adjustment, the adjustment range, and the feedback data of the equipment after each adjustment, and a detailed text description of the complete optimization strategy reasoning process data is output.

[0013] Compared with the prior art, the beneficial effects of the present invention are: 1. In the present invention, by counting the types of oilfield stations, based on the relevance between different stations, a hierarchical decision-making system of large model + small model is used to fuse, analyze, and make decisions on data, realizing the overall automated management of oilfield stations from daily operation and maintenance to emergency management. Based on the real-time changes in sensor data of inter-well stations and transfer stations, an instant cross-station optimization and adjustment strategy is generated, and at the same time, a target optimization and adjustment strategy is generated in combination with historical data, improving the overall operation efficiency of oilfield stations under unattended conditions; 2. In the present invention, through the cooperation of the site coordination execution module and the data summary and decision-making module, combined with the data optimization small model of each site, the macroscopic adjustment strategy of the site is refined and executed, and the large model strategy is fine-tuned according to the data changes during the execution of the small model, realizing the continuous optimization of key parameters such as combustion control, pump speed adjustment, and water injection ratio during the oilfield production process, so as to improve the actual crude oil production quality under different conditions; 3. In the present invention, the output module records the inference process of the optimization strategy of the large model cooperating with the small model each time, including the rule algorithm based on the decision-making basis, the execution steps, and the equipment feedback data, etc., providing complete and accurate data support for subsequent analysis and summary as well as problem tracing, facilitating the management personnel to understand the actual situation during the operation and maintenance production process, and facilitating the continuous improvement of the operation and maintenance management work; The entire agent-based unattended intelligent operation and maintenance centralized control method and system for oilfield stations can realize site correlation analysis, generation of instant optimization and adjustment strategies and target optimization and adjustment strategies, execution of site-specific parameters, feedback and fine-tuning of optimization and adjustment strategies, and recording of the decision-making process, closely combining the large model and the small model to form a hierarchical control architecture with upper and lower coordination, enhancing the practicability and functionality. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a block diagram of the agent-based unattended intelligent operation and maintenance centralized control system for oilfield stations of the present invention; Figure 2 It is a flowchart of the agent-based unattended intelligent operation and maintenance centralized control method for oilfield stations of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0015] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0016] Embodiment 1: As Figure 1As shown in the figure, the unattended intelligent operation and maintenance centralized control system for oilfield stations based on agents includes a data acquisition module, a data summary and decision-making module, a site coordination and execution module, an emergency intervention module, and an output module; The data acquisition module, by deploying pressure, temperature, flow, liquid level, and water cut sensors at oilfield inter-well stations, transfer stations, and joint stations, can collect real-time data of crude oil transportation equipment at inter-well stations, transfer stations, and joint stations; The emergency intervention module, in the data summary and decision-making module and the site coordination and control module, when the model reports an error, the system will automatically generate an event report and notify the on-duty operation and maintenance personnel through text messages and APP notifications. The content of the generated event report includes the error code and number, the name of the involved sensor, and the information of the site involved in the error; The output module, after each adjustment and optimization of the site equipment parameters by the data summary and decision-making module and the site coordination and execution module, uses the initial state of the equipment before adjustment and the relevant environmental data as the starting point of reasoning, records the rules and algorithms based on which the data summary and decision-making module formulates the optimization strategy, and records each step in the process of the site coordination and execution module implementing the optimization strategy, including the order of parameter adjustment, the adjustment range, and the feedback data of the equipment after each adjustment, and outputs a detailed text description of the complete optimization strategy reasoning process data; The data summary and decision-making module receives the real-time crude oil transportation equipment data of the subordinate inter-well stations, transfer stations, and joint stations obtained by the data acquisition module. Based on the data changes of the real-time sensors in the inter-well stations and transfer stations, it generates cross-site instant optimization adjustment strategies for the transfer stations and joint stations through a large model. At the same time, based on the natural language processing technology in the large model, it semantically analyzes the data in the oilfield production logs and operation instructions, and generates target optimization adjustment strategies for different sites, including combustion control, pump speed adjustment, and water injection ratio optimization, and transmits them to the site coordination and execution module; The specific steps for constructing the LMM large model for the subordinate inter-well stations, transfer stations, and joint stations are as follows: Statistical information of the subordinate inter-well stations, transfer stations, and joint stations of the system, including site locations and the upstream and downstream relationships of the sites, establish a relationship network of inter-well stations, transfer stations, and joint stations, and collect historical sensor data of different sites respectively, including pressure data, temperature data, flow data, liquid level data, and crude oil water content data, and calculate the mean value in the historical sensor data of the corresponding equipment at different sites And the standard value , through the 3sigma principle, set the normal range of sensor data for different sites ; It should be noted that the inter-well stations, transfer stations, and joint stations are modeled as directed graphs , where They respectively represent the stations and the upstream and downstream flow relationships, and at the same time construct an adjacency topology matrix , if the station and the station have a direct upstream and downstream relationship, then , otherwise , where is the total number of stations. For the inter-well stations, transfer stations, and combined stations, collect the pressure data, temperature data, flow data, liquid level data, and water content data of the crude oil of the equipment existing under the corresponding stations, including the heating furnace combustion chamber temperature, outlet crude oil temperature, ambient temperature, gas pipeline pressure, combustion chamber pressure, gas supply volume, crude oil flow rate, pump inlet pressure, pump outlet pressure, water injection pipeline pressure, mixed liquid pipeline pressure, pump liquid delivery flow rate, crude oil water content, injected water flow rate, total mixed liquid flow rate, storage tank liquid level, buffer tank liquid level, etc., so as to conduct immediate optimization and adjustment through the large model specifically

[0017] Based on the normal ranges of the sensor data of different stations, record the abnormal parameter situations of the sensor data of each station respectively, including those lower than and higher than the normal ranges, and record them as low anomalies and high anomalies respectively. And collect the adjustment strategies of the associated stations in the relationship network under the corresponding abnormal parameter situations, extract the data of the upstream and downstream stations of the abnormal stations in the same time window, form an associated adjustment strategy event set, combine the abnormal parameter situations with the adjustment strategies of the associated stations into training samples, each sample contains input and output, select an open-source large language model as the framework, and make specific modifications according to the characteristics of the oilfield station data, add an input layer to process the structured abnormal parameter information, and set an output layer to generate the adjustment strategy text under the corresponding input layer. Input the constructed training samples into the model batch by batch, perform forward propagation to calculate the model output, calculate the loss value according to the loss function, and update the model parameters through the Adam optimization algorithm. Repeat multiple rounds until the model converges or reaches the preset number of training rounds

[0018] It should be noted that the input of each sample is the abnormal parameter situation, such as the abnormal high pressure and low temperature at the inter-well station A. At the same time, the abnormal parameters are converted into natural language descriptions such as "the pressure at the inter-well station A is higher than the normal range" as the text prefix. Using the Seq2Seq structure, the output of each sample is the adjustment strategy for the associated station, such as increasing the valve opening by 10% at the transfer station B and decreasing the heating power by 10% at the combined station C. Among them, the input layer encodes the abnormal parameter situation into a vector form through one-hot encoding and inputs it into the model. For the classification task of the adjustment strategy, different types of adjustment strategies are regarded as different categories, and the cross-entropy loss function is used as the loss function. Based on the special adjustment behaviors of the oilfield stations, for the numerical parameters existing in the adjustment strategy, such as the adjustment ratio of the valve opening, the mean squared error loss is combined with the cross-entropy loss function to calculate the loss value, and the weight ratios are set to 0.5 and 0.5 respectively. The ratios of the training set, validation set, and test set are 70%, 15%, and 15%. The open-source large language models include GPT-Neo, DeepSeek, etc.

[0019] When the data aggregation and decision-making module generates the immediate optimization adjustment strategy and the target optimization adjustment strategy through the large model, it includes the following steps: Receive the real-time pressure, temperature, flow rate, liquid level, and water content data of the subordinate inter-well stations and transfer stations obtained by the data acquisition module. Based on the abnormal changes in the real-time data, identify the abnormal inter-well stations and transfer stations. According to the crude oil transportation relationship network of different abnormal inter-well stations and transfer stations, use the LMM large model to generate cross-station immediate optimization adjustment strategies for the relevant transfer stations and combined stations, including the following steps: Receive the real-time pressure, temperature, flow rate, liquid level, and water content data of the subordinate inter-well stations and transfer stations obtained by the data acquisition module, and calculate the mean value according to the time window. Based on the normal range of the sensor data of different stations, determine the mean value data that is not included in the range. When the sensor mean value data is less than the normal range of the current sensor data of the corresponding station, it is determined that the current sensor data of the current station is abnormally low. When the sensor mean value data is higher than the normal range of the current sensor data of the corresponding station, it is determined that the current sensor data of the current station is abnormally high; One-hot encode the abnormal parameter situations of the abnormal inter-well stations and transfer stations and the position information of the stations in the relationship network. Use the LMM large model to generate cross-station immediate optimization adjustment strategies for the relevant transfer stations and combined stations. Obtain the probability of each adjustment strategy through the softmax function, and transmit the immediate optimization adjustment strategy with the highest probability to the station coordination and execution module.

[0020] It should be noted that based on the LMM large model, the macroscopic adjustment strategy of associated sites is obtained. For example, if the input information is that the pressure of well - to - well station A is higher than the normal range, based on the relationship network between well - to - well station A and transfer station B, the output result is that transfer station B increases the valve opening, and the output result is transmitted to the site coordination execution module for further parameter adjustment within the site.

[0021] Based on the natural language processing technology in the large model, semantic parsing is performed on the data in the oilfield production logs and updated operation instructions of well - to - well stations, transfer stations, and joint stations. Then, combined with historical data according to the analysis model, target optimization adjustment strategies for different sites are generated, including combustion control, pump speed regulation, and water injection ratio optimization, and are transmitted to the site coordination execution module. The steps are as follows: Collect production log and operation instruction data, including operation scenarios and equipment information. Use the labeled production log and operation instruction data to train the BiLSTM - CRF model. Use the dependency syntax analysis tool to analyze the operation instructions, construct a dependency syntax tree to clarify the dependency relationship between words, and extract the cause, action, and target parameters in the causal relationship according to the dependency syntax tree and the identified entities to form a structured triple. The triple = (cause, action, target parameter). Match the target optimization adjustment strategy suitable for the subordinate sites according to the generated triple and transmit it to the site coordination execution module.

[0022] It should be noted that the pre - processed production log is input into the trained NER model to identify key parameters, operation actions, and equipment identifiers. The key parameters include temperature, pressure, flow rate, etc., the operation actions include opening, closing, adjusting, etc., and the equipment identifiers include pumps, valves, burners, etc. According to the generated triple and the preset rules, match the target optimization adjustment strategy suitable for the subordinate sites. For example, if the triple = (too high temperature, reduce, combustion power), the target optimization adjustment strategy generated according to the relevant rules is the combustion control strategy, and the combustion control strategy signal is transmitted into the site coordination execution module for further parameter refinement through a small model.

[0023] Embodiment 2: The site coordination execution module, based on the instant optimization adjustment strategies and target optimization adjustment strategies generated for each site by the data aggregation decision module, combines the data optimization small models deployed for each site for in - depth optimization and fine control of specific equipment and functions, executes the corresponding adjustment strategies for each site respectively, and at the same time, based on the data changes during the execution of the small models of different sites, further fine - tunes the large model optimization adjustment strategy according to the real - time monitoring data. The steps are as follows: For different subordinate stations, based on the historical data of the equipment in each station, data adjustment prediction models for the corresponding equipment are established separately through neural networks. According to the instant optimization adjustment strategy and the target optimization adjustment strategy transmitted by the data summary decision module, the optimization adjustment stations and the optimization adjustment equipment involved are determined. Based on the parameter fluctuation value of the impact source, the data adjustment value is calculated through the corresponding equipment data adjustment prediction model of the involved stations. The specific steps are as follows: For different equipment in different well - block stations, transfer stations, and gathering stations, long - short - term memory networks are used to construct the equipment parameter adjustment prediction models under the corresponding stations respectively with the mean square error as the loss function combined with the Adam optimizer. Based on the instant optimization adjustment strategy and the target optimization adjustment strategy of the data summary decision module, through perform strategy - equipment matching, where are the strategy feature vector and the equipment historical data feature vector respectively; It should be noted that represents the strategy feature vector, which represents the mathematical features of the optimization strategy. The vector dimensions include energy consumption, efficiency, safety, adjustment range, priority weight, etc. For example, when the optimization strategy is "reduce the energy consumption of the well - block station", the vector is expressed as [energy consumption weight, adjustment range, equipment type code], is the equipment feature vector, which represents the mathematical features of the equipment historical data and functions, including parameters such as equipment type, historical energy consumption, failure frequency, operating efficiency, etc. For example, the feature vector of a pump is expressed as [average energy consumption, maximum flow rate, vibration threshold], The closer it is to 1, the higher the matching degree between the current optimization strategy and the current equipment.

[0024] Determine the key parameter fluctuation source through sensitivity analysis, and calculate the parameter for the target gradient, where: ; Among them, represents the output value of the th hidden layer in the neural network, is the optimization target variable, is the equipment parameter, represents the total number of hidden layers of the neural network; It should be noted that the optimization target variables include energy consumption, flow rate, safety score, etc., and the equipment parameters include pump speed, combustion temperature, water - blending ratio. By calculating the gradient of the target with respect to the parameters, the sensitivity of the parameter to the target can be quantified. The larger the absolute value of the gradient, the more effective the adjustment is for achieving the target.

[0025] Take the strategy target and the current parameter In the input model, the parameter adjustment amount is obtained by solving , combined with the device safety threshold, where the input model is: ; Among them, is the parameter adjustment amount, that is, the value of the device parameter change that needs to be calculated, is the neural network prediction model, that is, the target value predicted after the input parameter , represents the regularization coefficient, and then a constrained optimization problem is constructed: ; Among them, respectively represent the physical safety thresholds of the device parameters, and the optimal adjustment amount is finally obtained by using the Lagrange multiplier method, and the parameters are adjusted by the optimal adjustment amount at the corresponding site.

[0026] Based on the data changes during the execution of the small models at different sites, the large model optimization adjustment strategy is further fine-tuned according to the real-time monitoring data, including the following steps: During the parameter adjustment process of the small model, the real-time data after the execution of each site's small model is uploaded to the central data pool, and by calculating the deviation between the actual value and the target value , respectively represent the actual value and the target value, and the deviation is decomposed into the contributions of each associated site through the chain rule: ; Among them, represents the key parameter of the th site, is the actual adjustment amount, and based on the real-time feedback data, the policy value function of the large model is updated, the influence matrix between sites is defined and the constrained optimization problem is reconstructed, and at the same time, the adjusted policy parameters are verified to ensure that the policy parameters are between the safety thresholds, where the policy parameters , if the policy parameters exceed the boundary, the projection method is used for correction; It should be noted that through: ; the policy value function of the large model is updated, respectively represent the state and the action, represents the transferred state and action after execution, respectively represent the learning rate, discount factor, and immediate reward. The update step size is controlled by the learning rate, the current and future rewards are balanced by the discount factor, and the immediate rewards are such as the energy consumption reduction rate, water injection ratio optimization parameter, combustion control parameter, etc. The influence matrix between sites is defined , where the element represents the coupling coefficient of site m to site n. By updating the global objective function, the multi-site coupling effect is incorporated: ; Among them, represents the total number of sites, represents the coupling penalty coefficient to suppress excessive cross-site interference, which is obtained by the projection method when the parameter exceeds the boundary : .

[0027] At the same time, record the policy parameters and the corresponding performance indicators after each adjustment to build a historical version library, and re-solve the global optimization problem every fixed time window.

[0028] It should be noted that the specific steps for re-solving the global optimization problem every fixed time window are as follows: ; Among them, represents the number of data points within the window, represents the actual values at time points.

[0029] Example 3: As Figure 2 shown, the unattended intelligent operation and maintenance centralized control method for oilfield stations based on agents includes the following steps: S1. By deploying pressure, temperature, flow rate, liquid level, and water cut sensors at oilfield inter-well stations, transfer stations, and joint stations, real-time collection of crude oil transportation equipment data at inter-well stations, transfer stations, and joint stations is carried out, and real-time crude oil transportation equipment data of the subordinate inter-well stations, transfer stations, and joint stations is received. Based on the data changes of the real-time sensors at the inter-well stations and transfer stations, through the LMM large model, cross-site instant optimization adjustment strategies and target optimization adjustment strategies for the transfer stations and joint stations are generated; S2. Based on the instant optimization adjustment strategies and target optimization adjustment strategies of each site, combined with the data optimization small models deployed at each site for in-depth optimization and fine control of specific equipment and functions, the corresponding adjustment strategies are executed for each site respectively. At the same time, based on the data changes during the execution of the small models at different sites, the large model optimization adjustment strategy is further fine-tuned according to the real-time monitoring data; S3. When the model reports an error, an event report is automatically generated and notified to the on-duty operation and maintenance personnel through text messages and APP notifications. The content of the generated event report includes the error code and number, the names of the involved sensors, and the information of the sites involved in the error; S4. After each adjustment and optimization of the site equipment parameters, use the initial state of the equipment before adjustment and the relevant environmental data as the starting point for reasoning. Record the rules and algorithms based on which the optimization strategy is formulated, and record each step in the process of the site implementing the optimization strategy, including the order of parameter adjustment, the adjustment range, and the feedback data of the equipment after each adjustment. Output a detailed text description of the complete optimization strategy reasoning process data.

[0030] The intelligent unattended intelligent operation and maintenance centralized control method and system for oilfield stations based on agents of the present invention, when in use, statistically analyze the types of oilfield stations, and based on the relevance between different stations, use a hierarchical decision-making system of large models + small models to fuse, analyze, and make decisions on data, so as to realize the overall automated management of oilfield stations from daily operation and maintenance to emergency management. Quickly generate cross-site instant optimization adjustment strategies based on the changes in real-time sensor data of well-to-well stations and transfer stations, and at the same time generate target optimization adjustment strategies in combination with historical data, so as to improve the overall operation efficiency of oilfield stations in the case of unattended operation.

[0031] In the embodiments provided by the present invention, it should be understood that the disclosed equipment, devices, and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation; the modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the method of this embodiment.

[0032] The above embodiments are only used to illustrate the technical method of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. The intelligent agent-based unmanned intelligent operation and maintenance centralized control system for oilfield stations is characterized by: It includes data collection module, data summary decision module, site coordination execution module, emergency intervention module and output module; The data acquisition module collects the crude oil transportation equipment data of the well station, transfer station and joint station in real time by deploying pressure, temperature, flow, liquid level and water content sensors at the well station, transfer station and joint station of the oil field; The data aggregation decision module receives the real-time crude oil transportation equipment data of the subordinate well stations, transfer stations, and joint stations obtained by the data acquisition module, generates the cross-site instant optimization adjustment strategy of the transfer station and the joint station through the big model based on the data changes of the real-time sensors in the well stations and transfer stations, and performs semantic analysis on the data in the oilfield production log and the operation instructions based on the natural language processing technology in the big model, and generates the target optimization adjustment strategy of different sites by combining the analysis model with the historical data, including combustion control, pump speed adjustment, and water mixing ratio optimization, and transmits it to the site coordination execution module; The site coordination execution module, based on the real-time optimization adjustment strategy and the target optimization adjustment strategy generated for each site by the data aggregation decision module, combines the small data optimization models deployed at each site for deep optimization and fine control of specific equipment and functions, and executes corresponding adjustment strategies for each site respectively. At the same time, based on the data changes of the small models of different sites during the execution process, the large model optimization adjustment strategy is further fine-tuned according to real-time monitoring data.

2. The agent-based unattended intelligent operation and maintenance centralized control system for oilfield stations according to claim 1 is characterized by: When the data aggregation decision module generates an immediate optimization adjustment strategy and a target optimization adjustment strategy through a large model, the following steps are included: The real-time pressure, temperature, flow, liquid level and water content data of the subordinate well stations and transfer stations obtained by the data acquisition module are received. The abnormal well stations and transfer stations are identified based on the abnormal changes in real-time data. According to the crude oil transportation relationship network of different abnormal well stations and transfer stations, the LMM large model is used to generate cross-site instant optimization adjustment strategies for related transfer stations and joint stations; Based on the natural language processing technology in the big model, semantic analysis is performed on the data in the oilfield production logs and updated operation instructions in the well stations, transfer stations, and joint stations. According to the analysis model and historical data, target optimization adjustment strategies for different sites are generated, including combustion control, pump speed adjustment, and water mixing ratio optimization, and transmitted to the site coordination execution module.

3. The agent-based unattended intelligent operation and maintenance centralized control system for oilfield stations according to claim 2 is characterized by: The specific steps for constructing the LMM large model for the subordinate well stations, transfer stations, and joint stations are as follows: The statistical system manages the information of well stations, transfer stations, and joint stations, including the location of the stations, the upstream and downstream relationships of the stations, and establishes a relationship network of well stations, transfer stations, and joint stations. It collects historical sensor data from different stations, including pressure data, temperature data, flow data, liquid level data, and crude oil water content data. By calculating the mean of the historical sensor data of the corresponding equipment at different stations With standard value , through the 3sigma principle, set the normal interval of sensor data at different sites ; Based on the normal interval of sensor data at different sites, the abnormal parameter conditions of sensor data at each site are recorded separately, including those below the normal interval and those above the normal interval, which are recorded as low abnormalities and high abnormalities respectively. The adjustment strategies of related sites in the relationship network under the corresponding abnormal parameter conditions are collected, and the data of upstream and downstream sites of the abnormal site in the same time window are extracted to form a set of related adjustment strategy events. The abnormal parameter conditions and the adjustment strategies of the related sites are combined into training samples. Each sample contains input and output. An open source large language model is selected as the framework, and specific modifications are made according to the characteristics of oilfield site data. An input layer is added to process structured abnormal parameter information, and an output layer is set to generate the adjustment strategy text under the corresponding input layer. The constructed training samples are input into the model in batches, and the model output is calculated by forward propagation. The loss value is calculated according to the loss function, and the model parameters are updated through the Adam optimization algorithm. Multiple rounds are repeated until the model converges or the preset number of training rounds is reached.

4. The agent-based unattended intelligent operation and maintenance centralized control system for oilfield stations according to claim 3 is characterized by: Based on abnormal changes in real-time data, abnormal well stations and transfer stations are identified. According to the crude oil transportation relationship network of different abnormal well stations and transfer stations, the LMM large model is used to generate cross-site instant optimization adjustment strategies for related transfer stations and joint stations, including the following steps: The real-time pressure, temperature, flow, liquid level and water content data of the subordinate well stations and transfer stations obtained by the data acquisition module are received, and the mean is calculated according to the time window. Based on the normal interval of the sensor data of different stations, the mean data not included in the interval is judged. When the sensor mean data is less than the normal interval of the current sensor data of the corresponding station, the current sensor data of the current station is judged to be low abnormal. When the sensor mean data is higher than the normal interval of the current sensor data of the corresponding station, the current sensor data of the current station is judged to be high abnormal. The abnormal parameters of abnormal well stations and transfer stations and the location information of the stations in the relationship network are uniquely encoded, and the LMM large model is used to generate cross-site instant optimization adjustment strategies for related transfer stations and joint stations. The probability of each adjustment strategy is obtained through the softmax function, and the instant optimization adjustment strategy with the maximum probability is transmitted to the site coordination execution module.

5. The agent-based unattended intelligent operation and maintenance centralized control system for oilfield stations according to claim 3 is characterized by: The method of generating target optimization adjustment strategies for different sites based on the natural language processing technology in the large model includes the following steps: Collect production logs and operation instruction data, including operation scenarios and equipment information, use the labeled production logs and operation instruction data to train the BiLSTM-CRF model, use the dependency syntax analysis tool to analyze the operation instructions, build a dependency syntax tree to clarify the dependency relationship between words, and extract the cause, action, and target parameters in the causal relationship based on the dependency syntax tree and the identified entities to form structured triples, triples = (cause, action, target parameter), match the target optimization and adjustment strategy suitable for the subordinate sites based on the generated triples, and transmit it to the site coordination execution module.

6. The agent-based unattended intelligent operation and maintenance centralized control system for oilfield stations according to claim 1 is characterized by: The instant optimization adjustment strategy and target optimization adjustment strategy generated by the data aggregation decision module for each site, and the corresponding adjustment strategy for each site are respectively executed by the small model, including the following steps: For different sites under its jurisdiction, based on the historical data of the equipment in each site, the data adjustment prediction model of the corresponding equipment is established through the neural network. According to the real-time optimization adjustment strategy and the target optimization adjustment strategy transmitted by the data summary decision module, the optimization adjustment site and optimization adjustment equipment involved are determined. Based on the parameter fluctuation value affecting the source, the data adjustment value is calculated through the corresponding equipment data adjustment prediction model of the site involved. The specific steps are as follows: For different equipment in different well stations, transfer stations, and joint stations, the equipment parameter adjustment prediction model for the corresponding stations is constructed by using the long short-term memory network with the mean square error as the loss function combined with the Adam optimizer. The instant optimization adjustment strategy based on the data summary decision module and the target optimization adjustment strategy are used. Perform policy device matching, where They are the strategy feature vector and the device history data feature vector respectively; Determine the source of fluctuation of key parameters through sensitivity analysis and calculate the parameters To target The gradient of , where: ； in, Represents the neural network The output value of the hidden layer is To optimize the target variable, is the device parameter, Represents the total number of hidden layers in the neural network; The strategic goal and current parameters Input the model and solve to obtain the parameter adjustment amount , combined with the equipment safety threshold, the input model is: ； in, is the parameter adjustment amount, that is, the device parameter change value that needs to be calculated, is the neural network prediction model, that is, the input parameter The target value of the prediction, Represents the regularization coefficient, and then constructs a constrained optimization problem: ； in, They represent the physical safety thresholds of equipment parameters respectively. The Lagrange multiplier method is used to finally obtain the optimal adjustment amount, and the parameters are adjusted at the corresponding site using the optimal adjustment amount.

7. The agent-based unattended intelligent operation and maintenance centralized control system for oilfield stations according to claim 6 is characterized by: The method further fine-tunes the optimization and adjustment strategy of the large model based on the data changes of the small models of different sites during the execution process according to the real-time monitoring data, including the following steps: During the parameter adjustment process of the small model, the real-time data of the small model at each site after execution is uploaded to the central data pool. Calculate the deviation between the actual value and the target value , Represent the actual value and the target value respectively, and decompose the deviation into the contribution of each associated site through the chain rule: ； in, Representative Key parameters of each site, The actual adjustment amount is based on real-time feedback data to update the strategy value function of the large model. , define the influence matrix between sites and reconstruct the constrained optimization problem, and verify the adjusted policy parameters to verify that the policy parameters are between the safety thresholds. ,If the strategy parameters are out of bounds, the projection method is used to correct them; At the same time, the strategy parameters and corresponding performance indicators after each adjustment are recorded to build a historical version library and re-solve the global optimization problem every fixed time window.

8. The agent-based unattended intelligent operation and maintenance centralized control system for oilfield stations according to claim 1 is characterized by: The emergency intervention module, in the data aggregation decision module and the site coordination control module, when an error occurs in the model, the system will automatically generate an event report and notify the on-duty operation and maintenance personnel via SMS and APP notification. The generated event report includes the error code and number, the name of the sensor involved, and the site information involved in the error; The output module uses the initial state of the equipment before adjustment and the relevant environmental data as the starting point of reasoning after each data aggregation decision module and the site coordination execution module perform site equipment parameter adjustment optimization, records the rules and algorithms used by the data aggregation decision module in formulating the optimization strategy, and records each step of the site coordination execution module in executing the optimization strategy, including the order of parameter adjustment, the amplitude of adjustment, and the feedback data of the equipment after each adjustment, and outputs a detailed text description of the complete optimization strategy reasoning process data.

9. An agent-based unmanned intelligent operation and maintenance centralized control method for oilfield stations, characterized in that: The method adopts the agent-based unmanned intelligent operation and maintenance centralized control system for oilfield stations as described in any one of claims 1 to 8, comprising the following steps: S1. By deploying pressure, temperature, flow, liquid level and water content sensors at oilfield well stations, transfer stations and joint stations, real-time data of crude oil transportation equipment at well stations, transfer stations and joint stations are collected, and real-time crude oil transportation equipment data of subordinate well stations, transfer stations and joint stations are received. Based on the data changes of real-time sensors in well stations and transfer stations, cross-site instant optimization adjustment strategies and target optimization adjustment strategies of transfer stations and joint stations are generated through LMM large model; S2. Based on the real-time optimization and adjustment strategies and target optimization and adjustment strategies of each site, combined with the data optimization small models deployed at each site for deep optimization and fine control of specific equipment and functions, the corresponding adjustment strategies are implemented for each site. At the same time, based on the data changes of the small models at different sites during the execution process, the optimization and adjustment strategies of the large model are further fine-tuned according to the real-time monitoring data; S3. When the model reports an error, it automatically generates an event report and notifies the on-duty operation and maintenance personnel via SMS or APP notification. The generated event report includes the error code and number, the name of the sensor involved, and the site information involved in the error. S4. After each site equipment parameter adjustment and optimization, the initial state of the equipment before adjustment and related environmental data are used as the starting point for reasoning. The rules and algorithms based on which the optimization strategy is formulated are recorded. Each step in the site's execution of the optimization strategy is recorded, including the order of parameter adjustment, the magnitude of adjustment, and the feedback data of the equipment after each adjustment. A complete text description of the optimization strategy reasoning process data is output.

Images