Complex geothermal system digital twinning online experiment method based on multi-source data deep fusion

Abstract

The invention provides a complex geothermal system digital twinning online experiment method based on multi-source data deep fusion, and solves the problems of precise construction of geothermal system digital twinning and low-delay visualization of virtual-real interaction. The method comprises the following steps: firstly, building a typical geothermal reservoir multi-source heterogeneous data parallel acquisition system, and carrying out fine inversion, high-precision identification and continuous correction on geothermal reservoir parameters by using a large amount of field experiment test data; then, a complete physical model of the geothermal system is established, online simulation is conducted on an actual system through high-performance scientific calculation, and mutual calibration is conducted on the actual system and test data; and finally, building a geothermal system digital twinning on-line experiment platform based on real-time acquired test data, a reliable mechanism model and an artificial intelligence method, and realizing accurate measurement and on-line monitoring of a geothermal exploitation process and iterative evolution of a geothermal reservoir exploitation scheme. According to the method, multi-source data of the geothermal system are deeply fused, high-precision online simulation and low-time-delay virtual-real interaction in the geothermal experiment process are achieved, and geothermal storage and exploitation details can be comprehensively transparentized. And furthermore, intelligent and rapid inversion is performed on the multi-scale seepage and heat transfer process of the geothermal reservoir by carrying out a virtual online experiment, so that the scientificity and efficiency of full-life-cycle geothermal exploitation are greatly improved. Meanwhile, a multi-agent collaborative optimization regulation and control framework provided by the method has relatively high expandability and flexibility, and an optimal well spacing design scheme and an optimal experiment scheme of the geothermal well group can be automatically excavated through adaptive optimization and iterative evolution of a system regulation and control strategy; and the method has a very strong application prospect in the distributed optimization field of the geothermal storage virtual experiment.

Description

Technical Field

[0001] The present invention relates to digital twin online testing and virtual experiment technology in the geothermal field, and in particular to a digital twin online experiment method for a complex geothermal system based on deep fusion of multi-source data. Background Art

[0002] The strong heterogeneity and multi-scale nature of geothermal reservoirs cause the flow in deep underground rock bodies to coexist in multiple modes. The temporal and spatial distribution of these flow modes is very complex, and the flow mechanisms are rich and diverse. It is a typical multi-scale multi-physics coupled flow. In addition, geothermal systems are generally in a geological environment where the seepage-stress-temperature-chemistry multi-physics field is coupled. There are many physical, mechanical and chemical processes, mainly heat transfer processes, fluid flow processes, medium stress and deformation processes, chemical reactions and solute migration processes. These processes always coexist and influence each other to varying degrees, which are specifically manifested in the characteristics of multi-phase, multi-medium, multi-scale and multi-mode. The geothermal system is accelerating the development of high efficiency, intelligence, low carbon and resilience, and has put forward higher requirements for the efficient perception, dynamic response, rapid analysis and optimization decision-making of the preliminary experimental testing process of geothermal reservoirs with complex storage environments. How to use experimental technology to accurately invert the flow and heat transfer processes is a challenging task. However, there are many problems with the current geothermal system experimental testing and exploration technology: on the one hand, the geothermal system experimental methods under the existing technical framework are relatively simple and extensive; on the other hand, the existing experimental testing methods often select typical sections for idealized analysis, and the authenticity needs to be improved. In addition, insufficient information under hidden geological conditions makes it difficult to guarantee the experimental testing results.

[0003] As an effective method to communicate the digital virtual world and the real physical world, digital twin technology has been widely used in different industries in recent years. Digital twin technology integrates multi-disciplinary, multi-physical quantity, multi-scale, and multi-probability simulation processes by making full use of data such as physical models, sensor updates, and operation history, so as to accurately simulate the real behavior of physical entities in virtual space. The online experimental method based on digital twin technology can effectively support the real feature perception, high-fidelity modeling, virtual-real mapping, and multi-physical simulation of geothermal systems, bringing a new technical paradigm for all-round digital and intelligent experimental testing in the geothermal field, and is becoming a new development direction for the construction of smart geothermal fields and digital transformation. However, the digital twin online experimental method is still in its initial stage, and its conceptual framework, key theories, and technical directions are still unclear. There are problems such as "cannot be measured", "cannot be fully measured", "cannot be measured accurately", "cannot be calculated", "cannot be calculated accurately", and "cannot be calculated quickly". These six problems permeate all aspects of the precise construction of geothermal storage digital twins and the collaborative interaction between virtual and real:

[0004] (1) System hardware testing: Due to factors such as the scale of the geothermal system, design capacity, geothermal reservoir geological environment monitoring methods, and data transmission costs, it is difficult to arrange sensors at all measurement points in the system, and thus truly, objectively, comprehensively, and accurately perceive the system experimental process. In addition, the downhole sensors used in geothermal reservoir geological experiments are of various types and are based on independent supporting host computer software. The diversity of similar sensor monitoring instruments at the geothermal reservoir experiment site leads to inconsistent data formats, making it difficult to complete synchronous data collection; finally, the data measured by sensors or other testing methods are often of poor quality and noisy, and the information density of the monitoring data is very low, which cannot reflect the true status of each component of the system from multiple angles, in all directions, and with multiple attributes.

[0005] (2) System software simulation: Geothermal systems exhibit complex dynamic characteristics of multiple fields, multiple scales, uncertainty and time-varying characteristics. Under the existing simulation technology system, it is almost impossible to build a global, cross-scale, multi-disciplinary and dynamic accurate digital twin, and it is difficult to mirror the characteristics, behaviors and performance of geothermal systems from the physical world to the digital world with high fidelity; on the one hand, there is a contradiction between resolution and scale in geothermal experimental test data. All high-resolution data are discontinuous, while the resolution of continuous data is limited. On the other hand, traditional geothermal reservoir geological models have problems such as poor simulation effects in multi-scale seepage simulation and parameter optimization. It is urgent to further explore the problem of coordinated integration of physical models and data-driven models to improve the simulation accuracy of geothermal digital twin systems. In addition, the digital model has a huge calculation scale, requires a lot of time and computing resources to solve, and may even be impossible to calculate, making it difficult to meet the timeliness requirements of building digital twins. Summary of the invention

[0006] In order to solve the above problems, the present invention can overcome the shortcomings of the existing technology by fully integrating field experimental tests, physical models and machine learning methods to construct a geothermal storage digital twin. Based on the precise digital twin, the transient response of the geothermal storage in the early experimental test phase is simulated, and the geothermal system well layout plan and intelligent control strategy are optimized on this basis, so as to realize the iterative evolution of the geothermal system control strategy under different mining conditions, thereby finding the optimal experimental plan.

[0007] In view of this, the present invention provides a digital twin online experimental method for a complex geothermal system based on deep fusion of multi-source data, which includes the following basic operating steps:

[0008] S1: Using signal conversion, real-time communication, multi-thread management and C / S programming technology, we developed a parallel and universal acquisition interface for geothermal system multi-source heterogeneous test data based on the Labview software platform to unify the monitoring data formats of various sensors, thereby realizing online reading, analysis, storage, conversion and transmission of a wide variety of monitoring data;

[0009] S2: Based on the inverse problem theory, machine learning and swarm intelligence optimization algorithm, we deeply integrate some parameters collected by sensors in real time, and build an intelligent testing method for accurate inversion and identification of global dynamic parameters of geothermal reservoirs, so as to solve the problem that many dynamically changing parameters such as porosity and permeability of geothermal reservoirs are difficult to invert online.

[0010] S3: Use limited sensor points to fully and objectively perceive the geothermal system, establish the optimal layout of complex measurement points, and then conduct comprehensive testing of the parameters that are "incompletely measured" by the digital twin model of the geothermal system;

[0011] S4: Establish multi-source data segmentation, noise reduction and calibration methods provided by multiple sensors to achieve accurate perception of the measured object from different angles and the same dimension and different angles and different dimensions, and then accurately test the "uncertain" parameters of the geothermal system digital twin model to overcome the problems of poor test data quality, high noise and low information density;

[0012] S5: Constructing a virtual-reality mapping method for low-latency visualization and real-time collaborative interaction of geothermal system digital twins;

[0013] S6: Summarize the hierarchical coupling characteristics between the subsystems of the geothermal system, clarify the functions of the agents in each subsystem, and then design a reasonable multi-agent system structure. Through the task interaction and coordination between the agents, a multi-agent distributed optimization and control experimental framework is constructed;

[0014] S7: Build an online digital twin experimental platform for the geothermal system, start the multi-agent collaborative evolution workflow to conduct virtual experiments online, optimize the well layout plan and intelligent control strategy of the geothermal system, realize the iterative evolution of the system regulation strategy under different mining conditions, and find the optimal experimental plan.

[0015] Furthermore, in step S1, a parallel and universal acquisition interface for geothermal system multi-source heterogeneous test data is developed to unify the monitoring data formats of various sensors, thereby realizing online reading, parsing, storage, conversion and sending of a wide variety of monitoring data, which specifically includes the following steps:

[0016] Step 1-1: To address the problem of inconsistent data formats caused by the diversity of sensors, a systematic database is established by standardizing data according to a custom protocol to reduce the complexity of the data interaction interface between the geothermal storage digital twin and the physical entity;

[0017] In step 1-2, all sensor measuring points are organically and cooperatively connected into a whole, and the monitoring data information is intelligently acquired based on the multi-source sensor data acquisition system, and then transmitted and processed online; the acquisition terminal database is constructed by integrating its own attribute data as well as sensor data and GPS data that characterize the status of the monitoring points.

[0018] Furthermore, in step S2, an intelligent testing method for accurately inverting and identifying global dynamic parameters of geothermal reservoirs is constructed to solve the problem that many dynamically changing parameters such as the porosity and permeability of geothermal reservoirs are difficult to invert online, including the following steps:

[0019] Step 2-1, using uniform sampling or Latin hypercube sampling to obtain a sample set of test variables, and using the obtained sample set to drive the geothermal system seepage forward problem calculation model;

[0020] Step 2-2: Compare some parameters collected by the sensor in real time with the system thermodynamic characteristics solved by the forward problem model, and estimate the optimal model parameters based on various inverse criteria such as the least squares criterion, maximum likelihood criterion, and minimum mean square error criterion, using machine learning and swarm intelligence optimization algorithms such as simulated annealing, genetic algorithm, and particle swarm optimization algorithm;

[0021] In step 2-3, the inverse parameters are substituted into the seepage dynamics analysis model to obtain information such as groundwater flow and geothermal field that are difficult to monitor inside complex geothermal reservoirs.

[0022] Furthermore, in step S3, a limited number of sensor measuring points are used to comprehensively and objectively sense the geothermal system, an optimal layout scheme of complex measuring points is established, and then the parameters that are "incompletely measured" in the digital twin model of the geothermal system are comprehensively tested, and the steps are as follows:

[0023] Step 3-1: evaluate the significance of sensor measurement points, select the measurement points that have a significant impact on the analysis results, and achieve optimal sensor layout;

[0024] Step 3-2 is to screen characteristic parameters that can comprehensively, accurately and reliably characterize the multi-scale seepage dynamics performance of the geothermal system, improve the density of sensor information and reduce the redundancy of information.

[0025] Furthermore, in step S4, in order to solve the problems of poor data quality, high noise and low information density of monitoring data measured by the geothermal system, a multi-source data segmentation, noise reduction and calibration method provided by multiple sensors is established to accurately test the "uncertain" parameters of the geothermal system digital twin model, which specifically includes the following steps:

[0026] Step 4-1: Use multi-source data segmentation and noise reduction technology to reduce the deviation caused by poor data quality from the source. Deeply integrate experimental data, fast Fourier transform, wavelet algorithm and machine learning to drive the artificial intelligence twin to perform reliability time-frequency analysis on geothermal reservoir measurement data;

[0027] Step 4-2, by arranging different types of sensors at the same measuring point and different types of sensors at different measuring points, the information density of the monitoring data can be effectively improved, and more accurate and detailed perception of the measured object can be achieved from different angles in the same dimension and from different angles in different dimensions.

[0028] Furthermore, in step S5, a virtual-reality mapping method for low-latency visualization and real-time collaborative interaction of the geothermal system digital twin is established, and the detailed steps are as follows:

[0029] Step 5-1: Through field testing and the above-mentioned reservoir inversion methods, the three-dimensional geological body boundary range, geological conditions (including aquifer structure and hydrogeological parameters, etc.), formation lithology change data, deep formation geothermal resource recharge and discharge conditions (including mining, lateral outflow, etc.) and experimental process wellbore injection and production conditions required for constructing a geothermal reservoir simulation model are obtained, and then a three-dimensional geothermal reservoir digital twin model is constructed to depict the distribution characteristics of aquifers and geothermal resources;

[0030] Step 5-2: Summarize the multi-field, multi-scale, uncertain, and time-varying dynamic characteristics of geothermal systems, and build a cross-scale, high-fidelity, stable and reliable machine learning modeling method that integrates online monitoring data with physical constraints to overcome the "unable to calculate" problem in geothermal system digital twin experiments;

[0031] Step 5-3: summarize the diversity distribution characteristics of the three reservoir space types in geothermal reservoirs, including pores, fractures and faults, and then summarize the characteristics of the multi-scale flow patterns of three properties inside the geothermal reservoir, including Darcy flow, mixed flow and pipe flow;

[0032] Step 5-4: Based on the specific characterization volume unit scale, a cross-scale flow online simulation method that nests multi-scale physical processes is directly constructed in discrete space, and the multi-scale flow characteristics inside the geothermal reservoir space are automatically and efficiently characterized according to the needs of thermal reservoir testing in the digital twin online experiment;

[0033] Step 5-5: sort out the differences between the geothermal storage digital twin model and the real system in characteristic parameters and characteristic scales, establish a method of "measurement and calculation fusion" of measured data and simulation data to improve the accuracy of simulation results, mirror the characteristics, behavior and performance of the geothermal system to the digital space with high fidelity, and improve the problem of "inaccurate calculation" of the geothermal system digital twin;

[0034] Steps 5-6: Establish a real-time and fast solution method based on model reduction and machine learning acceleration for the multi-phase, multi-media, multi-physics, and multi-scale simulation model for the seepage dynamics analysis of the actual geothermal system; evolve the system simulation process with huge computational overhead to a response efficiency of seconds, overcome the problem of "slow calculation" of the geothermal system digital twin, and thus provide real-time virtual-to-real mapping results for the geothermal system digital twin.

[0035] Furthermore, in step S6, the hierarchical coupling characteristics between the subsystems of the geothermal system are summarized to design a reasonable multi-agent system structure, and then a multi-agent distributed optimization and control experimental framework is constructed, and the steps are as follows:

[0036] Step 6-1: Design the central coordination agent, underground heat storage / ground energy station equipment group agent, and individual agent from the top to the bottom, clarify the functions and goals of each agent, and then establish a geothermal system hierarchical multi-agent structure to achieve efficient distributed online control;

[0037] In step 6-2, the multi-agent group structure of underground heat storage is divided into geothermal reservoirs and well groups. According to the characteristics of typical geothermal storage and based on the reservoir composition, it is divided into pore agents, fracture agents, fault zone agents, etc.; the surface equipment agent group is divided into geothermal water injection and production devices and circulation pumps; based on the direct information exchange and control action execution of single surface / underground agents, the system-level optimization and regulation of the geothermal extraction process is realized.

[0038] Furthermore, in step S7, a geothermal system digital twin online experimental platform is built, and a multi-agent collaborative evolution workflow is started to carry out virtual experiments online to achieve iterative evolution of system control strategies under different mining conditions to find the optimal experimental solution. The detailed steps are as follows:

[0039] In step 7-1, the digital twin model of the geothermal system is used as the research object, and the objective function, decision variables and constraints of the system control strategy optimization problem are determined through theoretical analysis. Among them, the objective function is determined as the experimental energy consumption index, economic index or multi-objective optimization function according to the optimization direction; the decision variables are determined as system control parameters such as the number of designed drilling holes, wellbore injection flow, etc.; the constraints are the upper and lower limits of the decision variables, or the equality constraints and inequality constraints on the decision variables;

[0040] Step 7-2: Based on the model input parameters (including real-time and historical data or other characteristic variables, etc.), the geothermal reservoir multi-scale seepage dynamics process prediction model is trained, and the interval estimation method is used to achieve real-time evaluation and prediction of the reliability of the prediction results, providing accurate geothermal reservoir seepage field as data support for the optimization of experimental schemes;

[0041] Step 7-3, based on the geothermal system online bidirectional virtual-real mapping method established in step S5, a high-fidelity simulation module of the geothermal subsystem is designed, and through the data interaction interface, a collaborative workflow of real-time simulation of the digital twin model and optimization of operating parameters is started;

[0042] Step 7-4: During the iteration process, the central coordination agent receives the underlying information and optimizes the objectives, domain prior knowledge and negotiation mechanism according to the experimental process. It uses dynamic programming and heuristic intelligent algorithms to establish a collaborative optimization strategy among agents and optimize the control parameters in the decision variables.

[0043] Step 7-5, the optimized control parameters are transferred into the geothermal reservoir multi-scale seepage dynamics model for simulation calculation, wherein the simulation calculation results are used to calculate the objective function to evaluate the effect of the optimized control;

[0044] Step 7-6, ending the optimization process of the central coordination agent, sending control signals to the underground heat storage / surface equipment group agent, and controlling the optimization logic of the individual agents at the bottom of the geothermal system;

[0045] In step 7-7, the iterative evolution of geothermal system design and geothermal mining process control strategy under different working conditions is achieved through multiple iterative optimization of decision variables.

[0046] Compared with the prior art, the complex geothermal system digital twin online experimental method based on deep fusion of multi-source data described in the present invention has the following advantages:

[0047] (1) An optimal layout scheme for complex measuring points was constructed to more accurately and meticulously sense the measured object from different angles and the same dimension, as well as from different angles and different dimensions, thereby greatly improving the density of sensor information and reducing information redundancy, and realizing the inversion of detailed information of the geothermal system using limited sensor measuring points;

[0048] (2) By deeply integrating real-time collected test data, reliable mechanism models and artificial intelligence methods, a "measurement and calculation fusion" method is constructed to mirror the characteristics, behavior and performance of the geothermal system in digital space with high fidelity, greatly improving the accuracy of the online simulation results of the digital twin experiment;

[0049] (3) Establish a real-time, fast solution method based on model reduction and machine learning acceleration, evolve the system simulation process with huge computational overhead into a response efficiency of seconds, and significantly improve the problem of "slow calculation" of geothermal system digital twins;

[0050] (4) Through task interaction and mutual coordination among the intelligent agents in the geothermal system, a multi-agent collaborative evolutionary control experimental framework is constructed, which can efficiently carry out virtual experiments, optimize the well layout plan and intelligent control strategy of the geothermal system online, and realize the iterative evolution of the system control strategy under different mining conditions, so as to find the optimal experimental plan. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is an online experimental method for digital twin of complex geothermal system according to an embodiment of the present invention.

[0052] Figure 2 This is a flow chart of the digital twin experiment of the multi-scale seepage process of sandstone thermal reservoirs with multi-source data fusion in the embodiment of the method of the present invention.

[0053] Figure 3 The invention provides a machine learning modeling method that integrates the physics of sandstone thermal reservoir seepage dynamics in an embodiment of the method.

[0054] Figure 4 This is a multi-intelligent architecture of a sandstone geothermal system in accordance with an embodiment of the method of the present invention.

[0055] Figure 5 This is an online virtual-reality interaction method for a sandstone thermal reservoir digital twin according to an embodiment of the method of the present invention.

[0056] Figure 6 This is an intelligent optimization method for the arrangement of sandstone thermal reservoir well groups and experimental schemes in an embodiment of the method of the present invention.

[0057] Figure 7 The method of the present invention is an intelligent agent collaborative optimization process based on dynamic programming and heuristic intelligent algorithms. DETAILED DESCRIPTION

[0058] The following is a detailed description of an embodiment of the present invention in conjunction with the accompanying drawings. This embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation method and a specific operation process are given, but the protection scope of the present invention is not limited to the following embodiments.

[0059] This embodiment is described based on a scenario of geothermal exploitation of a medium-deep sandstone thermal reservoir. Figure 1 The present invention provides a digital twin online experimental method for complex geothermal systems based on deep fusion of multi-source data. Figure 2 As shown, the following steps are included;

[0060] S1: Using signal conversion, real-time communication, multi-thread management and C / S programming technology, we developed a sandstone geothermal system multi-source heterogeneous test data parallelism and a universal acquisition interface based on the Labview software platform to unify the monitoring data formats of various sensors, thereby realizing online reading, analysis, storage, conversion and sending of a wide variety of monitoring data. The specific steps include:

[0061] Step 1-1: To address the problem of inconsistent data formats caused by sensor diversity, a systematic database is established by standardizing data according to a custom protocol to reduce the complexity of the data interaction interface between the digital twin of the sandstone thermal reservoir and the physical entity;

[0062] In step 1-2, all sensor measuring points are organically and cooperatively connected into a whole, and the monitoring data information is intelligently acquired based on the multi-source sensor data acquisition system, and then transmitted and processed online; the acquisition terminal database is constructed by integrating its own attribute data as well as sensor data and GPS data that characterize the status of the monitoring points.

[0063] S2: Based on the inverse problem theory, machine learning and swarm intelligence optimization algorithm, we deeply integrate some parameters collected by sensors in real time, and build an intelligent testing method for accurate inversion and identification of global dynamic parameters of geothermal reservoirs. This method solves the problem that many dynamically changing parameters such as porosity and permeability of sandstone thermal reservoirs are difficult to invert online. The method includes the following steps:

[0064] Step 2-1, uniform sampling or Latin hypercube sampling is used to obtain a sample set of test variables, and the obtained sample set is used to drive the seepage forward problem calculation model of the sandstone geothermal system;

[0065] Step 2-2, compare some parameters collected by sensors in real time with the thermodynamic characteristics of the sandstone geothermal system solved by the forward problem model, and estimate the optimal model parameters based on various inverse criteria such as the least squares criterion, maximum likelihood criterion, and minimum mean square error criterion, using machine learning and swarm intelligence optimization algorithms such as simulated annealing, genetic algorithm, and particle swarm optimization algorithm;

[0066] In step 2-3, the inverse parameters are substituted into the seepage dynamics analysis model to obtain information such as groundwater flow and geothermal field that are difficult to monitor inside the complex sandstone thermal reservoir.

[0067] S3: Use limited sensor measurement points to fully and objectively perceive the sandstone geothermal system, establish the optimal layout plan for complex measurement points, and then conduct a comprehensive test of the parameters that are "incompletely measured" in the digital twin model of the sandstone geothermal system. The steps are as follows:

[0068] Step 3-1: evaluate the significance of sensor measurement points, select the measurement points that have a significant impact on the analysis results, and achieve optimal sensor layout;

[0069] In step 3-2, characteristic parameters are screened that can comprehensively, accurately and reliably characterize the multi-scale seepage dynamics performance of the sandstone geothermal system, thereby improving the density of sensor information and reducing the redundancy of information.

[0070] S4: Establish multi-source data segmentation, noise reduction and calibration methods provided by multiple sensors to achieve accurate perception of the measured object from different angles and the same dimension and different angles and different dimensions, and then accurately test the "uncertain" parameters of the digital twin model of the sandstone geothermal system to overcome the problems of poor test data quality, high noise and low information density. Specifically, the following steps are included:

[0071] Step 4-1: Use multi-source data segmentation and noise reduction technology to reduce the deviation caused by poor data quality from the source. Deeply integrate experimental data, fast Fourier transform, wavelet algorithm and machine learning to drive the artificial intelligence twin to perform reliability time-frequency analysis on sandstone geothermal reservoir measurement data;

[0072] Step 4-2, by arranging different types of sensors at the same measuring point and different types of sensors at different measuring points, the information density of the monitoring data can be effectively improved, and more accurate and detailed perception of the measured object can be achieved from different angles in the same dimension and from different angles in different dimensions.

[0073] S5: Construct a virtual-reality mapping method for low-latency visualization and real-time collaborative interaction of the digital twin of the sandstone geothermal system. The detailed steps are as follows:

[0074] Step 5-1: Through field testing and the above-mentioned reservoir inversion methods, the three-dimensional geological body boundary range, geological conditions (including aquifer structure and hydrogeological parameters, etc.), formation lithology change data, deep formation geothermal resource recharge and discharge conditions (including mining, lateral outflow, etc.) and experimental process wellbore injection and production conditions required for constructing the sandstone thermal reservoir simulation model are obtained, and then a three-dimensional sandstone thermal reservoir digital twin model is constructed to depict the distribution characteristics of aquifers and geothermal resources;

[0075] Step 5-2 summarizes the multi-field, multi-scale, uncertain, and time-varying dynamic characteristics of the sandstone geothermal system, such as Figure 3 As shown in the figure, a cross-scale, high-fidelity, stable and reliable machine learning modeling method is constructed to integrate online monitoring data with physical constraints, overcoming the "unable to calculate" problem in the digital twin experiment of sandstone geothermal system;

[0076] Step 5-3: summarize the diversity distribution characteristics of the three reservoir space types of sandstone thermal reservoirs, including pores, cracks and fractures, and then summarize the characteristics of the multi-scale flow patterns of three properties inside the sandstone thermal reservoirs, including Darcy flow, mixed flow and pipe flow;

[0077] Step 5-4: Based on the specific characterization volume unit scale, a cross-scale flow online simulation method that nests multi-scale physical processes is directly constructed in discrete space, and the multi-scale flow characteristics inside the sandstone thermal reservoir space are automatically and efficiently characterized according to the needs of sandstone thermal reservoir testing in the digital twin online experiment;

[0078] Step 5-5: sort out the differences between the digital twin model of the sandstone geothermal reservoir and the real system in characteristic parameters and characteristic scales, establish a method of "measurement and calculation fusion" of measured data and simulation data to improve the accuracy of simulation results, and mirror the characteristics, behavior and performance of the sandstone geothermal system to the digital space with high fidelity, so as to improve the problem of "inaccurate calculation" of the digital twin of the sandstone geothermal system;

[0079] Steps 5-6: Establish a real-time and fast solution method based on model reduction and machine learning acceleration for the multi-phase, multi-media, multi-physics, and multi-scale simulation model for the seepage dynamics analysis of the actual sandstone geothermal system; evolve the system simulation process with huge computational overhead to a response efficiency of seconds, overcome the problem of "slow calculation" of digital twins, and provide real-time virtual-to-real mapping results for the digital twins of sandstone geothermal systems.

[0080] S6: Summarize the hierarchical coupling characteristics between the subsystems of the sandstone geothermal system, clarify the functions of the agents in each subsystem, and then design a reasonable multi-agent system structure. Through the task interaction and coordination between the agents, a multi-agent distributed optimization and control experimental framework is constructed. The steps are as follows:

[0081] Step 6-1, such as Figure 4 , from the top to the bottom, the central coordination agent, underground heat storage / ground energy station equipment group agent, and single agent are designed respectively, the functions and goals of each agent are clarified, and then the hierarchical multi-agent structure of the sandstone geothermal system is established to achieve efficient distributed online control;

[0082] In step 6-2, the multi-agent group structure of underground heat storage is divided into geothermal reservoirs and well groups. According to the characteristics of typical sandstone heat storage and based on the reservoir composition, it is divided into pore agents, fracture agents, fault zone agents, etc.; the surface equipment agent group is divided into geothermal water injection and production devices and circulation pumps; based on the direct information exchange and control action execution of single surface / underground agents, the system-level optimization and regulation of the geothermal extraction process is realized.

[0083] S7: Figure 5 As shown in the figure, a digital twin online experimental platform for the sandstone geothermal system is built, and a multi-agent collaborative evolution workflow is started to conduct virtual experiments online, optimize the well layout plan and intelligent control strategy of the sandstone geothermal system, and realize the iterative evolution of the system regulation strategy under different mining conditions to find the optimal experimental plan. The detailed steps are as follows:

[0084] Step 7-1, such as Figure 6 As shown in the figure, the digital twin model of the sandstone geothermal system is taken as the research object, and the objective function, decision variables and constraints of the system control strategy optimization problem are determined through theoretical analysis. Among them, the objective function is determined as the experimental energy consumption index, economic index or multi-objective optimization function according to the optimization direction; the decision variables are determined as system control parameters such as the number of designed drilling holes, wellbore injection flow, etc.; the constraints are the upper and lower limits of the decision variables, or the equality constraints and inequality constraints on the decision variables;

[0085] Step 7-2: Based on the model input parameters (including real-time and historical data or other characteristic variables, etc.), the multi-scale seepage dynamics process prediction model of sandstone thermal reservoir is trained, and the interval estimation method is used to achieve real-time evaluation and prediction of the reliability of the prediction results, so as to provide accurate sandstone thermal reservoir seepage field as data support for the optimization of experimental schemes;

[0086] Step 7-3, based on the online bidirectional virtual-real mapping method of the sandstone geothermal system established in step S5, a high-fidelity simulation module of the geothermal subsystem is designed, and through the data interaction interface, a collaborative workflow of real-time simulation of the digital twin model and optimization of operating parameters is started;

[0087] In step 7-4, during the iteration process, the central coordination agent receives the underlying information and optimizes the objectives, domain prior knowledge and negotiation mechanism according to the experimental process. It uses dynamic programming and heuristic intelligent algorithms to establish a collaborative optimization strategy among agents and optimize the control parameters in the decision variables, such as Figure 7 As shown,

[0088] Step 7-5, the optimized control parameters are transferred into the multi-scale seepage dynamics model of the sandstone thermal reservoir for simulation calculation, wherein the simulation calculation results are used to calculate the objective function to evaluate the effect of the optimized control;

[0089] Step 7-6, ending the optimization process of the central coordination agent, sending control signals to the underground heat storage / surface equipment group agent, and controlling the optimization logic of the individual agents at the bottom of the geothermal system;

[0090] In step 7-7, the iterative evolution of the sandstone geothermal system design scheme and the geothermal mining process control strategy under different working conditions is achieved through multiple iterative optimization of the decision variables.

[0091] The above-described embodiment is only a preferred solution of the present invention, but it is not intended to limit the present invention. A person skilled in the relevant technical field may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, any technical solution obtained by equivalent replacement or equivalent transformation falls within the protection scope of the present invention.

Claims

1. A digital twin online experimental method for complex geothermal systems based on deep fusion of multi-source data, characterized in that: The following steps are involved: S1: Using signal conversion, real-time communication, multi-thread management and C / S programming technology, we developed a parallel and universal acquisition interface for geothermal system multi-source heterogeneous test data based on the Labview software platform to unify the monitoring data formats of various sensors, thereby realizing online reading, analysis, storage, conversion and transmission of a wide variety of monitoring data; S2: Based on the inverse problem theory, machine learning and swarm intelligence optimization algorithm, we deeply integrate some parameters collected by sensors in real time, and build an intelligent testing method for accurate inversion and identification of global dynamic parameters of geothermal reservoirs, so as to solve the problem that many dynamically changing parameters such as porosity and permeability of geothermal reservoirs are difficult to invert online. S3: Use limited sensor measurement points to fully and objectively perceive the geothermal system, establish the optimal layout plan for complex measurement points, and then conduct comprehensive testing of the parameters that are "incompletely measured" by the digital twin model of the geothermal system; S4: Establish multi-source data segmentation, noise reduction and calibration methods provided by multiple sensors to achieve accurate perception of the measured object from different angles and the same dimension and different angles and dimensions, and then accurately test the "uncertain" parameters of the geothermal system digital twin model to overcome the problems of poor test data quality, high noise and low information density; S5: Constructing a virtual-reality mapping method for low-latency visualization and real-time collaborative interaction of geothermal system digital twins; S6: Summarize the hierarchical coupling characteristics between the subsystems of the geothermal system, clarify the functions of the agents in each subsystem, and then design a reasonable multi-agent system structure. Through the task interaction and coordination between the agents, a multi-agent distributed optimization and control experimental framework is constructed; S7: Build an online digital twin experimental platform for the geothermal system, start the multi-agent collaborative evolution workflow to conduct virtual experiments online, optimize the well layout plan and intelligent control strategy of the geothermal system, realize the iterative evolution of the system regulation strategy under different mining conditions, and find the optimal experimental plan.

2. According to claim 1, a digital twin online experimental method for complex geothermal systems based on deep fusion of multi-source data is characterized in that: In step S1, a parallel and universal acquisition interface for geothermal system multi-source heterogeneous test data is developed to unify the monitoring data formats of various sensors, thereby realizing online reading, parsing, storage, conversion and transmission of a wide variety of monitoring data, which specifically includes the following steps: Step 1-1: To address the problem of inconsistent data formats caused by the diversity of sensors, a systematic database is established by standardizing data according to a custom protocol to reduce the complexity of the data interaction interface between the geothermal storage digital twin and the physical entity; In step 1-2, all sensor measuring points are organically and cooperatively connected into a whole, and the monitoring data information is intelligently acquired based on the multi-source sensor data acquisition system, and then transmitted and processed online; the acquisition terminal database is constructed by integrating its own attribute data as well as sensor data and GPS data that characterize the status of the monitoring points.

3. According to claim 1, a digital twin online experimental method for a complex geothermal system based on deep fusion of multi-source data is characterized in that: In step S2, an intelligent testing method for accurately inverting and identifying global dynamic parameters of geothermal reservoirs is constructed to solve the problem that many dynamically changing parameters such as the porosity and permeability of geothermal reservoirs are difficult to invert online, including the following steps: Step 2-1, using uniform sampling or Latin hypercube sampling to obtain a sample set of test variables, and using the obtained sample set to drive the geothermal system seepage forward problem calculation model; Step 2-2: Compare some parameters collected by the sensor in real time with the system thermodynamic characteristics solved by the forward problem model, and estimate the optimal model parameters based on various inverse criteria such as the least squares criterion, maximum likelihood criterion, and minimum mean square error criterion, using machine learning and swarm intelligence optimization algorithms such as simulated annealing, genetic algorithm, and particle swarm optimization algorithm; In step 2-3, the inverse parameters are substituted into the seepage dynamics analysis model to obtain information such as groundwater flow and geothermal field that are difficult to monitor inside complex geothermal reservoirs.

4. According to claim 1, a digital twin online experimental method for complex geothermal systems based on deep fusion of multi-source data is characterized in that: In step S3, a limited number of sensor measuring points are used to comprehensively and objectively sense the geothermal system, and an optimal layout plan for complex measuring points is established, thereby comprehensively testing the "incompletely measured" parameters of the geothermal system digital twin model. The steps are as follows: Step 3-1: evaluate the significance of sensor measurement points, select the measurement points that have a significant impact on the analysis results, and achieve optimal sensor layout; Step 3-2 is to screen characteristic parameters that can comprehensively, accurately and reliably characterize the multi-scale seepage dynamics performance of the geothermal system, improve the density of sensor information and reduce the redundancy of information.

5. The method for online digital twin experiment of complex geothermal system based on deep fusion of multi-source data according to claim 1 is characterized in that: In step S4, in order to solve the problems of poor data quality, high noise and low information density of monitoring data measured by the geothermal system, a multi-source data segmentation, noise reduction and calibration method provided by multiple sensors is established to accurately test the "uncertain" parameters of the geothermal system digital twin model, which specifically includes the following steps: Step 4-1: Use multi-source data segmentation and noise reduction technology to reduce the deviation caused by poor data quality from the source. Deeply integrate experimental data, fast Fourier transform, wavelet algorithm and machine learning to drive the artificial intelligence twin to perform reliability time-frequency analysis on geothermal reservoir measurement data; Step 4-2, by arranging different types of sensors at the same measuring point and different types of sensors at different measuring points, the information density of the monitoring data can be effectively improved, and more accurate and detailed perception of the measured object can be achieved from different angles in the same dimension and from different angles in different dimensions.

6. The method for online digital twin experiment of complex geothermal system based on deep fusion of multi-source data according to claim 1 is characterized in that: In step S5, a virtual-real mapping method for low-latency visualization and real-time collaborative interaction of geothermal system digital twins is established, and the detailed steps are as follows: Step 5-1: Through field testing and the above-mentioned reservoir inversion methods, the three-dimensional geological body boundary range, geological conditions (including aquifer structure and hydrogeological parameters, etc.), formation lithology change data, deep formation geothermal resource recharge and discharge conditions (including mining, lateral outflow, etc.) and experimental process wellbore injection and production conditions required for constructing a geothermal reservoir simulation model are obtained, and then a three-dimensional geothermal reservoir digital twin model is constructed to depict the distribution characteristics of aquifers and geothermal resources; Step 5-2: Summarize the multi-field, multi-scale, uncertain, and time-varying dynamic characteristics of geothermal systems, and build a cross-scale, high-fidelity, stable and reliable machine learning modeling method that integrates online monitoring data with physical constraints to overcome the "unable to calculate" problem in geothermal system digital twin experiments; Step 5-3: summarize the diversity distribution characteristics of the three reservoir space types in geothermal reservoirs, including pores, fractures and faults, and then summarize the characteristics of the multi-scale flow patterns of three properties inside the geothermal reservoir, including Darcy flow, mixed flow and pipe flow; Step 5-4: Based on the specific characterization volume unit scale, a cross-scale flow online simulation method that nests multi-scale physical processes is directly constructed in discrete space, and the multi-scale flow characteristics inside the geothermal reservoir space are automatically and efficiently characterized according to the needs of thermal reservoir testing in the digital twin online experiment; Step 5-5: sort out the differences between the geothermal storage digital twin model and the real system in characteristic parameters and characteristic scales, establish a method of "measurement and calculation fusion" of measured data and simulation data to improve the accuracy of simulation results, mirror the characteristics, behavior and performance of the geothermal system in digital space with high fidelity, and improve the problem of "inaccurate calculation" of the geothermal system digital twin; Steps 5-6: Establish a real-time and fast solution method based on model reduction and machine learning acceleration for the multi-phase, multi-media, multi-physics, and multi-scale simulation model for the seepage dynamics analysis of the actual geothermal system; evolve the system simulation process with huge computational overhead to a response efficiency of seconds, overcome the problem of "slow calculation" of the geothermal system digital twin, and thus provide real-time virtual-to-real mapping results for the geothermal system digital twin.

7. The method for online digital twin experiment of complex geothermal system based on deep fusion of multi-source data according to claim 1 is characterized in that: In step S6, the hierarchical coupling characteristics between the subsystems of the geothermal system are summarized to design a reasonable multi-agent system structure, and then a multi-agent distributed optimization and control experimental framework is constructed. The steps are as follows: Step 6-1: Design the central coordination agent, underground heat storage / ground energy station equipment group agent, and individual agent from the top to the bottom, clarify the functions and goals of each agent, and then establish a geothermal system hierarchical multi-agent structure to achieve efficient distributed online control; In step 6-2, the multi-agent group structure of underground heat storage is divided into geothermal reservoirs and well groups. According to the characteristics of typical geothermal storage and based on the reservoir composition, it is divided into pore agents, fracture agents, fault zone agents, etc.; the surface equipment agent group is divided into geothermal water injection and production devices and circulation pumps; based on the direct information exchange and control action execution of single surface / underground agents, the system-level optimization and regulation of the geothermal extraction process is realized.

8. The method for online digital twin experiment of complex geothermal system based on deep fusion of multi-source data according to claim 1 is characterized in that: In step S7, a geothermal system digital twin online experimental platform is built, and a multi-agent collaborative evolution workflow is started to conduct virtual experiments online to achieve iterative evolution of system control strategies under different mining conditions, thereby finding the optimal experimental solution. The detailed steps are as follows: In step 7-1, the digital twin model of the geothermal system is used as the research object, and the objective function, decision variables and constraints of the system control strategy optimization problem are determined through theoretical analysis. Among them, the objective function is determined as the experimental energy consumption index, economic index or multi-objective optimization function according to the optimization direction; the decision variables are determined as system control parameters such as the number of designed drilling holes, wellbore injection flow, etc.; the constraints are the upper and lower limits of the decision variables, or the equality constraints and inequality constraints on the decision variables; Step 7-2: Based on the model input parameters (including real-time and historical data or other characteristic variables, etc.), the geothermal reservoir multi-scale seepage dynamics process prediction model is trained, and the interval estimation method is used to achieve real-time evaluation and prediction of the reliability of the prediction results, providing accurate geothermal reservoir seepage field as data support for the optimization of experimental schemes; Step 7-3, based on the geothermal system online bidirectional virtual-real mapping method established in step S5, a high-fidelity simulation module of the geothermal subsystem is designed, and through the data interaction interface, a collaborative workflow of real-time simulation of the digital twin model and optimization of operating parameters is started; Step 7-4: During the iteration process, the central coordination agent receives the underlying information and optimizes the objectives, domain prior knowledge and negotiation mechanism according to the experimental process. It uses dynamic programming and heuristic intelligent algorithms to establish a collaborative optimization strategy among agents and optimize the control parameters in the decision variables. Step 7-5, the optimized control parameters are transferred into the geothermal reservoir multi-scale seepage dynamics model for simulation calculation, wherein the simulation calculation results are used to calculate the objective function to evaluate the effect of the optimized control; Step 7-6, ending the optimization process of the central coordination agent, sending control signals to the underground heat storage / surface equipment group agent, and controlling the optimization logic of the individual agents at the bottom of the geothermal system; In step 7-7, the iterative evolution of geothermal system design and geothermal mining process control strategy under different working conditions is achieved through multiple iterative optimization of decision variables.

Images