Method for judging supporting stability of compressed air energy storage cave depot

Abstract

The invention provides a compressed air energy storage cavern support stability discrimination method, and relates to the technical field of cavern support, and the method specifically comprises the steps: collecting a cavern support sample and mechanical parameters and influence parameters of a to-be-discriminated support, constructing a current adversarial network, extracting pre-training model parameters in a database through transfer learning, and initializing the pre-training model parameters; simulation parameters are generated through training; building a heat-fluid-solid coupling finite element simulation model based on measured data, and calibrating the model by using simulation parameters; inputting influence parameters of a to-be-discriminated support into the calibrated model, obtaining mechanical parameters, and extracting characteristic parameters such as maximum stress, strain and anchor rod axial force; the safety coefficient is generated through data processing, and the supporting stability is comprehensively evaluated. According to the method, a multi-dimensional data evaluation system is constructed, comprehensive monitoring of complex physical phenomena such as thermal-mechanical coupling and seepage-stress hysteresis in the operation of the cave is achieved, the real stress state of the cave can be accurately evaluated, and potential risks can be pre-judged in advance.

Description

Technical Field

[0001] The present invention relates to the technical field of cavern support, and in particular to a method for judging the stability of compressed air energy storage cavern support. Background Art

[0002] Compressed air energy storage vaults are core facilities for storing compressed air, and the stability of their support structures is directly related to the safety and reliability of the energy storage system. During operation, the vaults not only bear static loads such as the weight of the surrounding rock and ground stress, but also face complex dynamic conditions such as thermal-mechanical coupling and seepage-stress coupling caused by the compressed air filling and deflating cycles. This makes the stress state of the support structure extremely complex, and existing technologies lack analysis and research on its stability.

[0003] In the prior art, publication number CN115203780A provides a method for determining the stability of the initial support of an underground gas storage chamber in a compressed gas storage power station. The method includes the following steps: considering the expansion force and softening properties of the hydrophilic surrounding rock of the underground gas storage chamber, plotting a surrounding rock response curve during humidification; obtaining the support parameters of the underground gas storage chamber, calculating and plotting a support structure characteristic curve; combining the surrounding rock response curve with the support characteristic curve, determining the coordinates of the intersection of the two curves, and obtaining the equilibrium support force when the hydrophilic surrounding rock and the support are in equilibrium based on the coordinates of the intersection; calculating the safety factor of the underground gas storage chamber based on the equilibrium support force and the ultimate support force; and determining the stability of the chamber based on the safety factor. This method fully considers the expansion force, softening property, and support force effects of the surrounding rock in the surrounding rock-support interaction relationship, and uses the safety factor as a stability indicator for the underground gas storage chamber. It is suitable for evaluating the stability of underground chambers with complex surrounding rock characteristics, and the determination method is accurate and efficient.

[0004] However, there are still the following deficiencies. As can be seen from the above statements, the existing technology is difficult to meet the stability judgment requirements of compressed air energy storage caverns under complex working conditions. First, the data dimension is single, focusing only on the expansion force and softening of hydrophilic surrounding rock, and not considering the key dynamic parameters in the operation process of compressed air energy storage caverns, such as the influence of temperature change amplitude, inflation / deflation cycle duration, etc. on the mechanical properties of the surrounding rock, as well as complex physical phenomena such as thermal-mechanical coupling and seepage-stress hysteresis, resulting in an inability to fully reflect the actual stress state of the cavern; Second, the analysis method is very limited. The judgment method based on curve combination and safety factor calculation is essentially a traditional analytical method, relying on simplified mechanical models and assumptions. It is difficult to accurately characterize the complex mechanical behavior of the support structure and surrounding rock under the coupling of multiple physical fields. When faced with nonlinear and dynamically changing working conditions, the accuracy of the judgment results is low; Third, there is a lack of model calibration and optimization mechanism, and no effective correlation with actual working condition data has been established. It is impossible to calibrate and optimize the judgment model based on real-time monitoring data or simulation results, and it is difficult to adapt to the support stability judgment requirements under different operating conditions.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0006] The purpose of the present invention is to provide a method for determining the stability of compressed air energy storage support to solve the problems raised in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions: A method for judging the support stability of a compressed air energy storage cavern comprises the following specific steps: S1. Collect support samples of the cavern and the mechanical parameters and influencing parameters of the support to be identified. Build a current adversarial network based on the mechanical parameters and influencing parameters of the cavern. Extract a pre-trained adversarial network from the database, transfer the model parameters to the current adversarial network, and complete the parameter initialization of the current adversarial network. Use the mechanical parameters and influencing parameters to train the current adversarial network to obtain simulated mechanical parameters and their corresponding simulated influencing parameters. The influencing parameters include the elastic modulus of the surrounding rock, the thermal expansion coefficient, the temperature variation amplitude of a single cycle, the duration of a single inflation / deflation cycle, the baseline consolidation time at room temperature, the influence coefficient of temperature on the consolidation time, the difference between the current working temperature and the room temperature, and the thermal-mechanical coupling risk coefficient and the seepage-stress hysteresis coefficient obtained based on the above parameters. S2. Construct a thermal-fluid-solid coupled finite element simulation model based on the mechanical parameters of the support specimen and their influencing parameters, and calibrate the simulation model based on the simulated mechanical parameters and their corresponding simulated influencing parameters; S3. Inputting the influencing parameters of the support to be identified into the calibrated simulation model to obtain the mechanical parameters of the support to be identified; S4. Extract characteristic parameters from the mechanical parameters of the support to be identified, and obtain the maximum stress value, the maximum strain value, and the maximum axial force value of the anchor; S5. Process the maximum stress, maximum strain, and maximum axial force of the support to be identified, respectively, to generate a stress safety factor, a strain safety factor, and an anchor axial force safety factor for the support to be identified; S6. Evaluate the stability status of the support to be identified based on the stress safety factor, strain safety factor and anchor axial force safety factor of the support to be identified.

[0008] Furthermore, the mechanical parameters include stress value, strain value and axial force value of the anchor rod.

[0009] Furthermore, we extract the pre-trained adversarial network from the database and transfer the model parameters to the current adversarial network. The specific steps are as follows: In the process of extracting pre-trained adversarial networks, a dynamic similarity matching algorithm is introduced to construct feature vectors based on the influencing parameters. Based on the feature vectors, the Euclidean distance between the influencing parameters of the pre-trained adversarial network and the current adversarial network is calculated to screen out the pre-trained adversarial networks with the top three similarities. The Euclidean distance between the influencing parameters of the pre-trained adversarial network and the current adversarial network is calculated based on the following formula: ; in, For the The Euclidean distance between the influencing parameters of the pre-trained adversarial network and the current adversarial network, that is, the similarity between the pre-trained adversarial network and the current adversarial network; Where, For the The first Influencing parameters, The current adversarial network Influencing parameters, is the index of the pre-trained adversarial network, is the index of the influencing parameter, , They correspond to the elastic modulus of the surrounding rock, the thermal expansion coefficient, the temperature variation range of a single cycle, the duration of a single inflation / deflation cycle, the benchmark consolidation time at room temperature, the influence coefficient of temperature on consolidation time, the difference between the current working temperature and the room temperature, the thermal-mechanical coupling risk coefficient, and the seepage-stress hysteresis coefficient. A weighted fusion migration method is used to assign corresponding weights to each pre-trained adversarial network according to the similarity between it and the current adversarial network, and the parameters of multiple pre-trained adversarial networks are superimposed and migrated to the current adversarial network according to the weights. The parameters of the pre-trained adversarial network are the weight matrix of the generator, the bias vector of the generator, the weight matrix of the discriminator, and the bias vector of the discriminator; The top three pre-trained adversarial networks in terms of similarity are pre-trained adversarial network A, pre-trained adversarial network B, and pre-trained adversarial network C. The weight coefficient of each pre-trained adversarial network is determined according to its similarity with the current adversarial network. The higher the similarity, the greater the weight coefficient. ; in, The current adversarial network parameters, is the 𝛿th parameter of the pre-trained adversarial network A, For the pre-trained adversarial network B parameters, is the first parameters, is the weight coefficient of the pre-trained adversarial network A, is the weight coefficient of the pre-trained adversarial network B, is the weight coefficient of the pre-trained adversarial network C, is the index of the pre-trained adversarial network parameters, , Corresponding to the weight matrix of the generator, the bias vector of the generator, the weight matrix of the discriminator and the bias vector of the discriminator, respectively, On the basis of .

[0010] Furthermore, the thermal-mechanical coupling risk coefficient is calculated according to the following formula: ; in, is the thermal-mechanical coupling risk coefficient; Where, is the elastic modulus of the support, is the coefficient of thermal expansion, is the temperature variation of a single cycle, The duration of a single inflation / deflation cycle.

[0011] Furthermore, the seepage-stress hysteresis coefficient is calculated according to the following formula: ; in, is the seepage-stress hysteresis coefficient; Where, is the benchmark consolidation time at room temperature, is the influence coefficient of temperature on consolidation time, It is the difference between the current operating temperature and the normal temperature.

[0012] Furthermore, the stability coefficient of the support to be identified is obtained according to the following formula: ; ; ; ; in, is the stability coefficient of the support to be identified; Where, is the stress safety factor, is the strain safety factor, is the anchor axial force safety factor, is the maximum stress that the support can withstand, is the maximum strain that the support can withstand, The maximum axial force allowed in anchor design. is the maximum stress value of the support, is the maximum strain value of the support, is the maximum axial force of the anchor rod.

[0013] Furthermore, the specific process of step S6 is as follows: The stability of the support is determined by the stress safety factor , strain safety factor , Anchor axial force safety factor The minimum value of determines that when any index does not meet the safety requirements, it will lead to support instability; when When , it means that the three safety factors are all greater than 1, then the ultimate bearing capacity of the support is greater than the actual load, and the support is stable; when When , it means that at least one of the three safety factors is equal to 1, then the actual load of the support reaches the ultimate bearing capacity and the support is on the verge of destruction; when When , it means that at least one of the three safety factors is less than 1, then the actual load of the support exceeds the ultimate bearing capacity and the support becomes unstable.

[0014] Among them, "1" is the mathematical representation that the ultimate capacity is equal to the actual load, and it is also the dividing point between stability and instability.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention collects multi-index mechanical parameters and their influencing parameters, including the elastic modulus of the surrounding rock, the thermal expansion coefficient, and the temperature variation amplitude of a single cycle. By calculating quantitative indicators such as the thermal-mechanical coupling risk coefficient and the seepage-stress hysteresis coefficient, a multi-dimensional data evaluation system is constructed. This improves and realizes the comprehensive monitoring of complex physical phenomena such as thermal-mechanical coupling and seepage-stress hysteresis during the operation of the cavern. It can accurately evaluate the actual stress state of the cavern and predict potential risks in advance. The thermal-fluid-solid coupling finite element simulation model is calibrated based on the simulated mechanical parameters and their corresponding simulated influencing parameters. The model can be continuously optimized according to actual working condition data, so that the model can adapt to the support stability judgment requirements under different geological conditions and operating conditions, thereby enhancing the versatility of the model. By constructing a current adversarial network, using transfer learning and adversarial training to obtain simulated mechanical parameters, and combining them with a thermal-fluid-solid coupled finite element simulation model for calibration, the adaptive ability of machine learning is combined with the physical modeling advantages of finite element analysis. This significantly improves the adaptability to nonlinear and dynamically changing working conditions, making the mechanical behavior analysis of the support structure and surrounding rock more in line with reality, and effectively improving the accuracy of the stability judgment results. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1Schematic diagram of the overall method flow of the present invention; Figure 2 is a fitting curve diagram of the elastic modulus and the thermal-mechanical coupling risk coefficient of the present invention; Figure 3 is a fitting curve diagram of the thermal expansion coefficient and the thermal-mechanical coupling risk coefficient of the present invention; Figure 4 This is a fitting curve diagram of the temperature change amplitude of a single cycle and the thermal-mechanical coupling risk coefficient of the present invention; Figure 5 This is a fitting curve diagram of the single inflation / deflation cycle duration and the thermal-mechanical coupling risk coefficient of the present invention; Figure 6 is a fitting curve diagram of the benchmark consolidation time and the seepage-stress hysteresis coefficient of the present invention; Figure 7 Graph showing the fitting curve of the duration of a single inflation / deflation cycle and the seepage-stress hysteresis coefficient of the present invention; Figure 8 : is a fitting curve diagram of the influence coefficient of temperature on consolidation time and seepage-stress hysteresis coefficient of the present invention; Figure 9 It is a fitting curve diagram of the difference between the current working temperature and the normal temperature and the seepage-stress hysteresis coefficient of the present invention. DETAILED DESCRIPTION

[0017] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to specific embodiments.

[0018] It should be noted that, unless otherwise defined, the technical or scientific terms used in the present invention should have the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0019] Example 1: See also Figures 1 to 9 , the present invention provides a technical solution: A method for judging the support stability of a compressed air energy storage cavern comprises the following specific steps: S1. Collect support samples of the cavern and the mechanical parameters and influencing parameters of the support to be identified. Build a current adversarial network based on the mechanical parameters and influencing parameters of the cavern. Extract a pre-trained adversarial network from the database, transfer the model parameters to the current adversarial network, and complete the parameter initialization of the current adversarial network. Use the mechanical parameters and influencing parameters to train the current adversarial network to obtain simulated mechanical parameters and their corresponding simulated influencing parameters. The influencing parameters include the elastic modulus of the surrounding rock, the thermal expansion coefficient, the temperature variation amplitude of a single cycle, the duration of a single inflation / deflation cycle, the baseline consolidation time at room temperature, the influence coefficient of temperature on the consolidation time, the difference between the current working temperature and the room temperature, and the thermal-mechanical coupling risk coefficient and the seepage-stress hysteresis coefficient obtained based on the above parameters. Based on the above embodiment, the mechanical parameters include stress value, strain value and axial force value of the anchor rod.

[0020] Based on the above embodiment, the stress value, strain value, axial force value of the anchor rod, elastic modulus of the surrounding rock, thermal expansion coefficient, temperature variation amplitude of a single cycle, duration of a single inflation / deflation cycle, reference consolidation time at room temperature, coefficient of influence of temperature on consolidation time, and difference between the current working temperature and the room temperature are obtained as follows: Stress refers to the internal force per unit area caused by external forces, temperature changes, or deformation constraints inside an object, reflecting the material's ability to resist deformation. Strain refers to the relative deformation of an object under stress, that is, the ratio of the deformation to the original size.

[0021] Bury strain gauges (such as vibrating wire strain gauges) on the surface or inside the support structure to directly measure the strain value and calculate the strain using Hooke's law , calculate the stress using Hooke's law , , is the elastic modulus of the support.

[0022] The axial force value of the anchor rod refers to the internal force that the anchor rod bears along the axial direction, which is mainly tensile force and is transmitted to the anchor rod by the deformation of the surrounding rock.

[0023] Install a dynamometer (such as a steel string dynamometer) at the tail of the anchor rod to directly measure the axial force value of the anchor rod.

[0024] The elastic modulus of surrounding rock refers to the proportional constant of the stress value and the strain value in the elastic deformation stage of surrounding rock. , reflecting the ability of rock mass to resist elastic deformation.

[0025] The coefficient of thermal expansion refers to the relative change in length or volume of a material caused by a unit temperature change when the temperature changes.

[0026] Apply a temperature gradient to the support material specimen (such as a concrete specimen), measure the length change, and calculate the thermal expansion coefficient ,in, is the length change, is the initial length, It is the difference between the current operating temperature and the normal temperature.

[0027] The temperature variation amplitude of a single cycle refers to the temperature fluctuation range of the surrounding rock during a single inflation / deflating cycle of the energy storage chamber (e.g., the difference between the highest and lowest temperatures). Temperature sensors (e.g., thermocouples, RTDs) are placed inside the chamber to record temperature changes during the cycle and obtain the temperature variation amplitude of a single cycle.

[0028] The duration of a single inflation / deflation cycle refers to the time required for the compressed air energy storage system to complete a "inflation energy storage-deflation energy release" process.

[0029] Directly obtain cycle duration data from the energy storage system control platform (e.g., 2 hours of inflation + 2 hours of deflation is one cycle, and the inflation / deflation cycle duration is 4 hours).

[0030] The benchmark consolidation time at room temperature refers to the time required for the rock mass to complete the main consolidation (the pore water pressure is basically dissipated) under given stress conditions at room temperature.

[0031] Consolidation test: A constant load is applied and the time required for the pore water pressure to dissipate to 90% is measured as the benchmark consolidation time.

[0032] The temperature effect coefficient on consolidation time refers to the degree to which temperature changes affect the consolidation time of a rock mass. Consolidation tests are conducted at different temperatures, and a curve is fitted showing the relationship between consolidation time and temperature. The temperature effect coefficient on consolidation time is the slope of the curve.

[0033] Current working temperature and normal temperature ( ) is the difference between the current operating temperature of the energy storage chamber and the normal temperature.

[0034] The current temperature in the cave is directly measured by the temperature sensor, and the difference is calculated by comparing it with the normal temperature.

[0035] Based on the above embodiments, the subsequently calculated stress values, strain values, axial force values ​​of the anchor rod, elastic modulus of the surrounding rock, thermal expansion coefficient, temperature variation amplitude of a single cycle, duration of a single inflation / deflation cycle, baseline consolidation time at normal temperature, coefficient of influence of temperature on consolidation time, and difference between the current working temperature and normal temperature are all data after averaging.

[0036] On the basis of the above embodiment, after collecting stress values, strain values, axial force values ​​of anchor rods, elastic modulus of surrounding rock, thermal expansion coefficient, temperature change amplitude of a single cycle, duration of a single inflation / deflation cycle, benchmark consolidation time at room temperature, influence coefficient of temperature on consolidation time, and difference between current working temperature and room temperature, the above parameters are respectively subjected to maximum-minimum normalization processing, and then the normalized data is used for subsequent analysis and processing, so that in the subsequent analysis and processing process, various data are analyzed and processed under the same dimension, avoiding the problem of some data being ignored due to different dimensions.

[0037] Based on the above embodiment, a pre-trained adversarial network is extracted from the database. The specific process is as follows: Select engineering scenario models similar to the current task from the database (such as the support stability model of traditional underground caverns and high-pressure gas storage), and ensure that their input and output parameters match the energy storage cavern; For example, if there is a "tunnel anchor support stability assessment model" in the database, whose input parameters include the elastic modulus and thermal expansion coefficient of the surrounding rock, it can be used as a pre-trained model; A pre-trained model usually consists of two parts: Generator: used to "manufacture" simulation parameters (e.g., generating possible stress values ​​based on the elastic modulus of the surrounding rock); Discriminator: used to "identify" whether the data is a real monitoring value or a generated simulated value; Directly extract the parameters of these two components (such as the weights and hierarchical structure of the neural network) without rebuilding the network framework; Copying the parameters of the pre-trained model (such as the connection strength of each layer in the generator) directly into the current network is equivalent to "borrowing" the learning results of the existing model; If the current task has more parameters (such as adding an "inflation / deflation cycle duration" input), you need to add a corresponding input layer to the front end of the model; If the number of parameters is smaller, delete irrelevant input layers in the pre-trained model (e.g., delete input related to "earthquake load"); Input simple data: Input a set of known and reasonable parameters (e.g., elastic modulus of surrounding rock = 20 GPa) into the migrated network and check whether the output is consistent with common sense (e.g., stress values ​​should be positive and anchor axial forces are within the design range); Calibrate output range: Ensure that the simulation parameters (such as stress and strain) output by the generator are within the physically feasible range (for example, stress does not exceed the compressive strength of the surrounding rock); Normalize parameters of different units; The collected measured data are divided into a training set and a validation set; Generate simulated mechanical parameters (such as stress and axial force of anchor bolts) based on input measured parameters (such as elastic modulus of surrounding rock and inflation cycle duration); Determine whether the input data is "real monitoring data" or "generated simulation data" and feed it back to the generator; The generator continuously adjusts its parameters to make the generated data as "fooling" as possible to the discriminator; The discriminator continuously improves its discrimination ability and distinguishes real data from simulated data; When the generated data is basically consistent with the real data in statistical characteristics (such as mean and distribution shape), stop training.

[0038] Based on the above embodiment, extracting a pre-trained adversarial network from a database has the following advantages: No need to train models from scratch; instead, you can leverage the learning results of pre-trained models in similar fields to shorten the R&D cycle. To address the problem of insufficient measured data for energy storage caverns (e.g., missing data for extreme temperature-pressure combination conditions), the generator can "create" reasonable simulation data to fill the data gaps. If new monitoring data is obtained later (such as new stress values ​​after the operation of the tunnel), the new data can be directly added to the training set to perform local fine-tuning on the model instead of retraining the entire network.

[0039] Based on the above embodiment, the pre-trained adversarial network is extracted from the database and the model parameters are transferred to the current adversarial network. The specific steps are as follows: In the process of extracting pre-trained adversarial networks, a dynamic similarity matching algorithm is introduced to construct feature vectors based on the influencing parameters. Based on the feature vectors, the Euclidean distance between the influencing parameters of the pre-trained adversarial network and the current adversarial network is calculated. The similarity between the two is measured by the Euclidean distance. The smaller the distance, the higher the similarity. The pre-trained adversarial networks with the top three similarities are selected according to the similarity ranking. The Euclidean distance between the influencing parameters of the pre-trained adversarial network and the current adversarial network is calculated based on the following formula: ; in, For the The Euclidean distance between the influencing parameters of the pre-trained adversarial network and the current adversarial network, that is, the similarity between the pre-trained adversarial network and the current adversarial network; Where, For the The first Influencing parameters, The current adversarial network Influencing parameters, is the index of the pre-trained adversarial network, is the index of the influencing parameter, , They correspond to the elastic modulus of the surrounding rock, the thermal expansion coefficient, the temperature variation range of a single cycle, the duration of a single inflation / deflation cycle, the benchmark consolidation time at room temperature, the influence coefficient of temperature on consolidation time, the difference between the current working temperature and the room temperature, the thermal-mechanical coupling risk coefficient, and the seepage-stress hysteresis coefficient. A weighted fusion migration method is used to assign corresponding weights to each pre-trained adversarial network according to the similarity between it and the current adversarial network, and the parameters of multiple pre-trained adversarial networks are superimposed and migrated to the current adversarial network according to the weights. The parameters of the pre-trained adversarial network are the weight matrix of the generator, the bias vector of the generator, the weight matrix of the discriminator, and the bias vector of the discriminator; The top three pre-trained adversarial networks with the highest similarity are pre-trained adversarial network A, pre-trained adversarial network B, and pre-trained adversarial network C. The weight coefficient of each pre-trained adversarial network is determined according to its similarity with the current adversarial network. The higher the similarity, the greater the weight coefficient. ; in, The current adversarial network parameters, is the 𝛿th parameter of the pre-trained adversarial network A, For the pre-trained adversarial network B parameters, is the first parameters, is the weight coefficient of the pre-trained adversarial network A, is the weight coefficient of the pre-trained adversarial network B, is the weight coefficient of the pre-trained adversarial network C, is the index of the pre-trained adversarial network parameters, , Corresponding to the weight matrix of the generator, the bias vector of the generator, the weight matrix of the discriminator and the bias vector of the discriminator, respectively, On the basis of .

[0040] Table 1. Changes in the thermal-mechanical coupling risk factor with elastic modulus, thermal expansion coefficient, temperature variation in a single cycle, and duration of a single inflation / deflation cycle

[0041] As shown in Table 1, the elastic modulus gradually increases from 10 to 22, and the thermal-mechanical coupling risk coefficient continues to rise from 1.13 to 67.61. This indicates that the larger the elastic modulus, the greater the stiffness of the cavern structure under thermal action, the more significant the stress accumulation effect, and the higher the thermal-mechanical coupling risk. Therefore, the thermal-mechanical coupling risk coefficient is positively correlated with the elastic modulus. As the thermal expansion coefficient increases from 0.1 to 0.82, the risk coefficient increases simultaneously, indicating that the stronger the thermal expansion characteristics of the geotechnical mass or structural material, the greater the thermal stress generated during temperature cycling, and the higher the coupling risk. Therefore, the thermal-mechanical coupling risk coefficient is positively correlated with the thermal expansion coefficient. The temperature variation in a single cycle increases from 1°C to 5.8°C, and the risk coefficient increases from 1.13 to 67.61. This indicates that the greater the temperature fluctuation, the more intense the alternating changes in thermal stress, the more significant the cyclic thermal load on the structure, and the higher the coupling risk. Therefore, the thermal-mechanical coupling risk coefficient is positively correlated with the temperature variation in a single cycle. The duration of a single inflation / deflation cycle was shortened from 8 hours to 0.6 hours, and the risk coefficient increased significantly from 1.13 to 67.61. This indicates that the shorter the cycle period, the higher the frequency of thermal effects, the denser the repeated loads the structure endures in a short period of time, the more obvious the cumulative effect of thermal-mechanical coupling, and the higher the risk. Therefore, the thermal-mechanical coupling risk coefficient is negatively correlated with the duration of a single inflation / deflation cycle.

[0042] according to Figure 2-Figure 5 It can be seen that the thermal-mechanical coupling risk coefficient is positively correlated with the elastic modulus, thermal expansion coefficient, and the temperature change amplitude of a single cycle (and there is an accelerated growth segment), and is negatively correlated with the duration of a single inflation / deflation cycle (there is also an accelerated decline segment).

[0043] Based on the above embodiment, the thermal-mechanical coupling risk coefficient is calculated according to the following formula: ; in, The thermal-mechanical coupling risk coefficient is used to evaluate the thermal-mechanical coupling risk of the energy storage cavern by combining four index parameters: elastic modulus, thermal expansion coefficient, temperature variation amplitude of a single cycle, and duration of a single inflation / deflation cycle. The larger the thermal-mechanical coupling risk coefficient, the higher the thermal-mechanical coupling risk of the energy storage cavern. Where, is the temperature variation of a single cycle, The duration of a single inflation / deflation cycle; On this basis, it should be noted that: When the elastic modulus When the thermal stress increases, according to Hooke's law, when the material expands and contracts due to temperature changes, the internal thermal stress increases. For high-rigidity materials, it is difficult to release thermal stress through deformation, resulting in an increased risk of stress concentration. Therefore, the thermal-mechanical coupling risk of the energy storage chamber increases, and thus the thermal-mechanical coupling risk coefficient increases. When the thermal expansion coefficient When the temperature increases, according to the formula of thermal deformation and thermal expansion coefficient of the material, the thermal deformation of the material increases. The greater deformation will lead to greater thermal stress inside the structure. Especially in confined spaces such as caverns, the obstructed deformation will aggravate the stress accumulation, making the thermal-mechanical coupling risk of the energy storage caverns increase, thereby increasing the thermal-mechanical coupling risk coefficient; When the temperature change amplitude of a single cycle When the temperature increases, the greater the temperature change, the more intense the material expansion / contraction. Severe temperature cycling will cause the material to repeatedly bear tensile and compressive stresses, increasing the risk of fatigue damage, increasing the thermal-mechanical coupling risk of the energy storage chamber, and increasing the thermal-mechanical coupling risk coefficient. When a single inflation / deflation cycle lasts When the temperature increases, the rate of temperature change decreases, and the heat does not have time to conduct, resulting in a large temperature difference between the inside and outside of the cavern, a significant temperature gradient, and additional thermal stress. In addition, the cycle frequency decreases, resulting in a decrease in the number of stress cycles per unit time, and fatigue damage is light, which reduces the thermal-mechanical coupling risk of the energy storage cavern and the thermal-mechanical coupling risk coefficient.

[0044] Therefore, the thermal-mechanical coupling risk coefficient is positively correlated with the elastic modulus, thermal expansion coefficient, and temperature change amplitude of a single cycle, and the thermal-mechanical coupling risk coefficient is negatively correlated with the duration of a single inflation / deflation cycle.

[0045] in addition, Reflects the material's ability to resist elastic deformation. Reflects the sensitivity of the material to temperature changes. The product of the two ( ) is the core item of thermal stress calculation (the combination of Hooke's law and thermal expansion formula), which directly determines the basic thermal stress level caused by temperature change; ( ) as the coefficient basis, indicating that these two parameters are inherent properties of the thermal-mechanical coupling risk and are directly related to the properties of the material itself.

[0046] Characterizes the intensity of temperature fluctuation in a single cycle, Reflects the cycle speed (the shorter the duration, the higher the frequency), the ratio of the two ( ) is essentially the temperature change rate (temperature fluctuation per unit time), which directly affects the uniformity of heat conduction and the frequency of fatigue damage; The dynamic correction of foundation thermal stress reflects the amplification or suppression of risk by the time effect of the cyclic process.

[0047] In summary, the above function form is used to express the functional relationship between the thermal-mechanical coupling risk coefficient and the elastic modulus, thermal expansion coefficient, temperature change amplitude of a single cycle, and duration of a single inflation / deflation cycle.

[0048] Based on the above embodiment, thermal-mechanical coupling will directly generate thermal stress. is the basic thermal stress term (similar to Hooke's law: stress = elastic modulus × strain, and strain is caused by temperature change); when When it increases, it indicates that the thermal stress risk is higher, that is, the thermal stress term is superimposed on the original mechanical stress.

[0049] The total strain is composed of mechanical strain and thermal strain. is positively correlated, in Reflects the sensitivity of the material to thermal deformation. Indirect assessment of contingency risk can be considered The larger it is, the greater the total strain (especially the contribution of thermal strain) is, and the higher the risk of structural deformation, that is, thermal strain is superimposed on mechanical strain.

[0050] The anchor rod bears the stress transfer of the support. When the thermal stress causes the structure to deform, the anchor rod will be subjected to additional force due to deformation coordination. For example, the structure is "pulled" by the anchor rod due to thermal expansion, and the axial force of the anchor rod may increase. The additional force caused by thermal stress needs to be considered in the original axial force calculation. Positive correlation.

[0051] It is a comprehensive indicator to measure the risk of thermal-mechanical coupling. The larger the value, the higher the risk of thermal stress, thermal strain and anchor additional force.

[0052] Table 2. Changes in the seepage-stress hysteresis coefficient with the baseline consolidation time at room temperature, the duration of a single inflation / deflation cycle, the influence coefficient of temperature on consolidation time, and the difference between the current working temperature and room temperature.

[0053] As shown in Table 2, as the benchmark consolidation time gradually increases from 2 to 26, the seepage-stress hysteresis coefficient continues to increase from 0.05 to 0.78. This indicates that at room temperature, the longer the benchmark consolidation time is, the longer the basic period for the rock and soil to complete consolidation is. In the dynamic coupling process of seepage and stress, the cumulative hysteresis effect is more prominent, resulting in a larger seepage-stress hysteresis coefficient and a more significant coupling hysteresis characteristic. Therefore, the seepage-stress hysteresis coefficient and the benchmark consolidation time are positively correlated.

[0054] The duration of a single inflation / deflation cycle gradually shortened from 8 to 0.6, and the seepage-stress hysteresis coefficient continued to increase from 0.05 to 0.78. This indicates that the shorter the cycle duration, the higher the frequency of alternating changes in seepage and stress. The pore water discharge and stress adjustment inside the rock and soil cannot keep up with the cycle rhythm, and the hysteresis effect is enhanced, resulting in an increase in the seepage-stress hysteresis coefficient. Therefore, the seepage-stress hysteresis coefficient is negatively correlated with the duration of a single inflation / deflation cycle.

[0055] The influence coefficient of temperature on consolidation time gradually decreased from 0.25 to 0.01, and the seepage-stress hysteresis coefficient continued to increase from 0.05 to 0.78, which means that the weaker the influence of temperature on consolidation time (the smaller the influence coefficient), the less the actual consolidation process is affected by temperature and is relatively more stable. However, since this coefficient is negatively correlated with the hysteresis coefficient, as the coefficient decreases, the hysteresis coefficient increases instead, reflecting that there is an inverse correlation between the temperature effect and the seepage-stress hysteresis effect. When the temperature effect is weak, the hysteresis effect is more likely to appear due to the relatively stable consolidation environment. Therefore, the seepage-stress hysteresis coefficient and the influence coefficient of temperature on consolidation time are negatively correlated.

[0056] The difference between the current operating temperature and normal temperature gradually decreases from 5.8 to 1, and the seepage-stress hysteresis coefficient continues to increase from 0.05 to 0.78. This shows that the closer the operating temperature is to normal temperature, the less the temperature change interferes with the consolidation time and the seepage-stress coupling process, and the more obvious the hysteresis effect is, which increases the seepage-stress hysteresis coefficient. This reflects the reverse regulation of the temperature difference on the seepage-stress hysteresis characteristics. The smaller the temperature difference, the larger the hysteresis coefficient. Therefore, the seepage-stress hysteresis coefficient and the difference between the current operating temperature and normal temperature are negatively correlated.

[0057] according to Figure 6-Figure 9 It can be seen that the benchmark consolidation time is positively correlated with the hysteresis coefficient. As the benchmark consolidation time increases, the hysteresis coefficient increases, indicating that the longer the consolidation time, the more significant the seepage-stress response hysteresis.

[0058] The duration of a single inflation / deflation cycle is negatively correlated with the hysteresis coefficient. The shorter the cycle duration, the larger the hysteresis coefficient, indicating that short-cycle cycles will enhance the seepage-stress hysteresis effect.

[0059] The influence coefficient of temperature on consolidation time is negatively correlated with the hysteresis coefficient. The smaller the influence coefficient, the larger the hysteresis coefficient, which reflects that when the temperature interference is weak, the hysteresis characteristic is more prominent.

[0060] The difference between the current operating temperature and the normal temperature is negatively correlated with the hysteresis coefficient. The smaller the difference, the larger the hysteresis coefficient, which means that when the temperature is stable, the hysteresis effect is more likely to appear.

[0061] Based on the above embodiment, the seepage-stress hysteresis coefficient is calculated according to the following formula: ; in, is the seepage-stress hysteresis coefficient. The seepage-stress hysteresis coefficient is used to evaluate the dynamic coupling hysteresis characteristics of the seepage field and stress field in the energy storage chamber by combining four index parameters: the benchmark consolidation time at room temperature, the duration of a single inflation / deflation cycle, the influence coefficient of temperature on consolidation time, and the difference between the current working temperature and the room temperature. The larger the seepage-stress hysteresis coefficient, the more significant the hysteresis effect of the seepage and stress response, and the higher the coupling risk. Where, is the benchmark consolidation time at room temperature, is the influence coefficient of temperature on consolidation time, is the difference between the current working temperature and the normal temperature; On this basis, it should be noted that: When the benchmark consolidation time at room temperature When it increases, it indicates that the rock mass has low permeability or high pore water viscosity, and the pore water pressure dissipates slowly. During the inflation / degassing cycle of the energy storage chamber, the stress field changes before the seepage field responds. The stress change has occurred, but the seepage field cannot adjust in time due to slow consolidation, resulting in the dynamic response of the two being out of sync. The hysteresis effect of seepage and stress response increases, the coupling risk increases, and the seepage-stress hysteresis coefficient increases; When a single inflation / deflation cycle lasts When the pressure increases, the stress cycle frequency decreases (such as slow inflation / deflation), the stress change amplitude per unit time decreases, the seepage field has more sufficient time to respond to stress changes, the hysteresis effect of seepage and stress response is reduced, the coupling risk is reduced, and the seepage-stress hysteresis coefficient is reduced; When the temperature influences the consolidation time When the temperature increases, it means that the permeability of the medium is improved more significantly when the temperature rises, and the pore water pressure dissipates faster. Even if the difference between the current working temperature and the normal temperature is the same, the larger the It will further shorten the actual consolidation time, enable the seepage field to respond faster to stress field changes, reduce the hysteresis effect of seepage and stress response, reduce the coupling risk, and reduce the seepage-stress hysteresis coefficient; When the difference between the current working temperature and the normal temperature When it increases, the viscosity of water decreases, and the thermal expansion of the rock mass may open microcracks, both of which will increase permeability, shorten the consolidation time, and make the seepage field respond faster to the stress field. The hysteresis effect of seepage and stress response is reduced, the coupling risk is reduced, and the seepage-stress hysteresis coefficient is reduced.

[0062] Therefore, the seepage-stress hysteresis coefficient is positively correlated with the benchmark consolidation time at room temperature, and the seepage-stress hysteresis coefficient is negatively correlated with the duration of a single inflation / deflation cycle, the influence coefficient of temperature on consolidation time, and the difference between the current working temperature and room temperature.

[0063] in addition,( ) is the ratio of the benchmark consolidation time to the cycle time, which is used to quantify the relative relationship between the “natural response speed of the seepage field” and the “stress field change speed”: when , the response speed of the seepage field is much slower than the change speed of the stress field, resulting in a significant hysteresis effect; when , the seepage field has enough time to respond to the stress change, resulting in a weak hysteresis effect; As a scaling factor for temperature effects: When the temperature rises ( ), The overall coefficient is reduced, reflecting the weakening effect of temperature on the hysteresis effect, and ( ) jointly determine the final lag degree; When the temperature rises ( ), The overall coefficient increases, reflecting the enhanced effect of temperature on the hysteresis effect, and ( ) superposition; therefore,( ) is the core proportional term of the formula. By comparing the “consolidation velocity of the seepage field” with the “circulation velocity of the stress field”, it establishes the benchmark level of the hysteresis effect. On this basis, the degree of hysteresis is further adjusted through the influence of temperature on permeability, and finally a comprehensive evaluation of the seepage-stress coupling hysteresis characteristics is formed.

[0064] In summary, the above function form is used to express the functional relationship between the seepage-stress hysteresis coefficient and the benchmark consolidation time at room temperature, the duration of a single inflation / deflation cycle, the influence coefficient of temperature on consolidation time, and the difference between the current working temperature and the room temperature.

[0065] Based on the above embodiment, the delayed response of the seepage field will cause the pore water pressure to be unable to dissipate in time. According to the Terzaghi effective stress principle, if the pore water pressure dissipates with a delay, the change of the effective stress will lag behind the total stress, thereby affecting the mechanical behavior of the rock mass (such as strain and anchor force). The hysteresis effect may cause additional stress concentration or deformation difference inside the rock mass, which needs to be solved through Correct the traditional mechanical calculation results.

[0066] Seepage-stress hysteresis coefficient ( ) By affecting the dissipation rate of pore water pressure, the effective stress and rock deformation are indirectly changed, thereby producing a correction effect on the stress, strain and anchor axial force in the traditional thermal-mechanical coupling calculation. The core idea is to As a quantitative indicator of the hysteresis effect, the linear or nonlinear coefficients (such as ) are superimposed on the original mechanical calculation results to reflect the additional risks brought about by the asynchrony between seepage and stress fields.

[0067] S2. Construct a thermal-fluid-solid coupled finite element simulation model based on the mechanical parameters of the support specimen and their influencing parameters, and calibrate the simulation model based on the simulated mechanical parameters and their corresponding simulated influencing parameters; On the basis of the above embodiment, a thermal-fluid-solid coupling finite element simulation model is constructed based on the mechanical parameters of the support sample and its measured influencing parameters. The specific process is as follows: Prepare data and determine the model scope. Data requirements: Mechanical parameters: stress, strain, and axial force of anchor bolts; Influencing parameters: Elastic modulus of surrounding rock, thermal expansion coefficient, inflation / deflation cycle duration, and consolidation time; Model range: With the chamber as the center, take the rock mass three times the chamber diameter; Meshing and material definition: Meshing: The rock mass around the chamber is meshed densely, and the far field is roughly meshed; the support structure (anchor) uses one-dimensional rod elements, and the rock mass uses two-dimensional plane elements; Material definition: Rock mass: Assign elastic modulus, thermal expansion coefficient, and permeability coefficient; Anchor: Elastic material, only bears tensile and compressive loads; Apply boundary conditions: Thermal boundary: Inner wall of the chamber: cyclic temperature load; Far away from the rock mass: fixed temperature; Seepage boundary: Rock mass bottom: fixed pore water pressure; Chamber wall: zero seepage flux (assuming support structure is watertight); Mechanical boundary: Initial ground stress: calculated based on rock mass deadweight; Inside the chamber: cyclic air pressure load (simulating inflation / deflation); Set up multi-field coupling relationships: Thermal-mechanical coupling: temperature changes cause thermal expansion of the rock mass, generating thermal stress; Seepage-mechanical coupling: pore water pressure affects rock stress through the effective stress principle; Run the model and analyze the results: First, calculate the temperature field to obtain the temperature distribution at each point; use the temperature field as input to calculate thermal stress and rock deformation; finally, calculate the seepage field and analyze the pore water pressure distribution.

[0068] On the basis of the above embodiment, the simulation model is calibrated based on the simulated mechanical parameters and their corresponding simulated influencing parameters. The specific process is as follows: Simulation data preprocessing and feature extraction Parameter initialization: Initialize the material properties and boundary conditions in the finite element model based on the mechanical parameters of the support sample (stress value, strain value, axial force value of the anchor) and influencing parameters (elastic modulus of the surrounding rock, thermal expansion coefficient, inflation / deflation cycle duration, consolidation time); Run the simulation model: Use the initialization parameters to run the thermal-fluid-solid coupling model to simulate the multi-field response of the energy storage chamber during the inflation / deflation cycle, and obtain the mechanical parameters output by the model, including stress, strain, and axial force of the anchor bolt. Comparison of simulated and actual parameters: Compare the simulated mechanical parameters output by the model with the simulated mechanical parameters obtained by training the adversarial network and their corresponding simulated influence parameters, and analyze the differences between the two; Adjust model parameters: Based on the comparison results, use optimization algorithms or manual adjustments to adjust the parameters in the finite element simulation model. For example, when the simulated stress value is inconsistent with the actual simulation parameters, the rock damping coefficient can be increased or the elastic modulus can be reduced. Iterative optimization: Repeat the steps of comparing simulation and actual parameters and adjusting model parameters, and continuously adjust the model parameters until the difference between the simulated mechanical parameters output by the model and the given simulation parameters reaches an acceptable range, thus completing the calibration of the finite element simulation model.

[0069] S3. Inputting the influencing parameters of the support to be identified into the calibrated simulation model to obtain the mechanical parameters of the support to be identified; S4. Extract characteristic parameters from the mechanical parameters of the support to be identified, and obtain the maximum stress value, the maximum strain value, and the maximum axial force value of the anchor; Based on the above embodiment, the maximum stress value, the maximum strain value and the maximum axial force value of the anchor rod are obtained according to the following formula: The support was obtained through the finite element simulation model The stress value of each measuring point, , calculate the maximum stress value of the support: ; in, is the maximum stress value of the support, For the support The stress value of each measuring point, is the index of the measurement point, is the number of support measurement points; The support was obtained through the finite element simulation model The strain value of each measuring point, , calculate the maximum strain value of the support: ; in, is the maximum strain value of the support, For the support The strain value of each measuring point; The finite element simulation model The axial force of the root anchor, ; Calculate the maximum axial force of the anchor rod: ; in, is the maximum axial force of the anchor rod, For the The axial force value of the anchor rod, is the index of the anchor, , is the number of anchor rods.

[0070] S5. Process the maximum stress, maximum strain, and maximum axial force of the support to be identified, respectively, to generate a stress safety factor, a strain safety factor, and an anchor axial force safety factor for the support to be identified; Based on the above embodiment, the stability coefficient of the support to be identified is obtained according to the following formula: ; ; ; ; in, The stability coefficient of the support to be identified is used to take the minimum value from the stress safety factor, strain safety factor, and anchor axial force safety factor to evaluate the stability of the support to be identified. The larger the stability coefficient, the higher the stability of the support to be identified. Where, is the stress safety factor, is the strain safety factor, is the anchor axial force safety factor, is the maximum stress that the support can withstand, is the maximum strain that the support can withstand, The maximum axial force allowed in anchor design.

[0071] S6. Evaluate the stability status of the support to be identified based on the stress safety factor, strain safety factor, and anchor axial force safety factor of the support to be identified.

[0072] Based on the above embodiment, the specific process of step S6 is as follows: The stability of the support is determined by the stress safety factor , strain safety factor , Anchor axial force safety factor The minimum value of determines that when any index does not meet the safety requirements, it will lead to support instability; when When , it means that the three safety factors are all greater than 1, then the ultimate bearing capacity of the support is greater than the actual load, and the support is stable; when When , it means that at least one of the three safety factors is equal to 1, then the actual load of the support reaches the ultimate bearing capacity and the support is on the verge of destruction; when When , it means that at least one of the three safety factors is less than 1, then the actual load of the support exceeds the ultimate bearing capacity and the support becomes unstable.

[0073] Among them, "1" is the mathematical representation that the ultimate capacity is equal to the actual load, and it is also the dividing point between stability and instability.

[0074] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0075] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by computer software, electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software depends on the specific application and design constraints of the technical solution.

[0076] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment as needed.

[0077] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A method for judging the support stability of a compressed air energy storage cavern, characterized by: The specific steps include: S1. Collect support samples of the cavern and the mechanical parameters and influencing parameters of the support to be identified. Build a current adversarial network based on the mechanical parameters and influencing parameters of the cavern. Extract a pre-trained adversarial network from the database, transfer the model parameters to the current adversarial network, and complete the parameter initialization of the current adversarial network. Use the mechanical parameters and influencing parameters to train the current adversarial network to obtain simulated mechanical parameters and their corresponding simulated influencing parameters. The influencing parameters include the elastic modulus of the surrounding rock, the thermal expansion coefficient, the temperature variation amplitude of a single cycle, the duration of a single inflation / deflation cycle, the baseline consolidation time at room temperature, the influence coefficient of temperature on the consolidation time, the difference between the current working temperature and the room temperature, and the thermal-mechanical coupling risk coefficient and the seepage-stress hysteresis coefficient obtained based on the above parameters. S2. Construct a thermal-fluid-solid coupled finite element simulation model based on the mechanical parameters of the support specimen and their influencing parameters, and calibrate the simulation model based on the simulated mechanical parameters and their corresponding simulated influencing parameters; S3. Inputting the influencing parameters of the support to be identified into the calibrated simulation model to obtain the mechanical parameters of the support to be identified; S4. Extract characteristic parameters from the mechanical parameters of the support to be identified, and obtain the maximum stress value, the maximum strain value, and the maximum axial force value of the anchor; S5. Process the maximum stress, maximum strain, and maximum axial force of the support to be identified, respectively, to generate a stress safety factor, a strain safety factor, and an anchor axial force safety factor for the support to be identified; S6. Evaluate the stability status of the support to be identified based on the stress safety factor, strain safety factor and anchor axial force safety factor of the support to be identified.

2. The method for determining the support stability of a compressed air energy storage cavern according to claim 1 is characterized in that: Mechanical parameters include stress value, strain value and axial force value of anchor rod.

3. The method for determining the support stability of a compressed air energy storage cavern according to claim 1 is characterized in that: Extract the pre-trained adversarial network from the database and transfer the model parameters to the current adversarial network. The specific steps are as follows: In the process of extracting pre-trained adversarial networks, a dynamic similarity matching algorithm is introduced to construct feature vectors based on the influencing parameters. Based on the feature vectors, the Euclidean distance between the influencing parameters of the pre-trained adversarial network and the current adversarial network is calculated to screen out the pre-trained adversarial networks with the top three similarities. The Euclidean distance between the influencing parameters of the pre-trained adversarial network and the current adversarial network is calculated based on the following formula: ； in, For the The Euclidean distance between the influencing parameters of the pre-trained adversarial network and the current adversarial network, that is, the similarity between the pre-trained adversarial network and the current adversarial network; Where, For the The first Influencing parameters, The current adversarial network Influencing parameters, is the index of the pre-trained adversarial network, is the index of the influencing parameter, , They correspond to the elastic modulus of the surrounding rock, the thermal expansion coefficient, the temperature variation range of a single cycle, the duration of a single inflation / deflation cycle, the benchmark consolidation time at room temperature, the influence coefficient of temperature on consolidation time, the difference between the current working temperature and the room temperature, the thermal-mechanical coupling risk coefficient, and the seepage-stress hysteresis coefficient. A weighted fusion migration method is used to assign corresponding weights to each pre-trained adversarial network according to the similarity between it and the current adversarial network, and the parameters of multiple pre-trained adversarial networks are superimposed and migrated to the current adversarial network according to the weights. The parameters of the pre-trained adversarial network are the weight matrix of the generator, the bias vector of the generator, the weight matrix of the discriminator, and the bias vector of the discriminator; The top three pre-trained adversarial networks in terms of similarity are pre-trained adversarial network A, pre-trained adversarial network B, and pre-trained adversarial network C. The weight coefficient of each pre-trained adversarial network is determined according to its similarity with the current adversarial network. The higher the similarity, the greater the weight coefficient. ； in, The current adversarial network parameters, is the 𝛿th parameter of the pre-trained adversarial network A, For the pre-trained adversarial network B parameters, is the first parameters, is the weight coefficient of the pre-trained adversarial network A, is the weight coefficient of the pre-trained adversarial network B, is the weight coefficient of the pre-trained adversarial network C, is the index of the pre-trained adversarial network parameters, , Corresponding to the weight matrix of the generator, the bias vector of the generator, the weight matrix of the discriminator and the bias vector of the discriminator, respectively, On the basis of .

4. The method for determining the support stability of a compressed air energy storage cavern according to claim 1 is characterized in that: The thermal-mechanical coupling risk coefficient is calculated according to the following formula: ； in, is the thermal-mechanical coupling risk coefficient; Where, is the elastic modulus of the support, is the coefficient of thermal expansion, is the temperature variation of a single cycle, The duration of a single inflation / deflation cycle.

5. The method for determining the support stability of a compressed air energy storage cavern according to claim 2 is characterized in that: The seepage-stress hysteresis coefficient is calculated based on the following formula: ； in, is the seepage-stress hysteresis coefficient; Where, is the benchmark consolidation time at room temperature, is the influence coefficient of temperature on consolidation time, It is the difference between the current operating temperature and the normal temperature.

6. The method for determining the support stability of a compressed air energy storage cavern according to claim 5 is characterized in that: The stability coefficient of the support to be identified is obtained according to the following formula: ； ； ； ； in, is the stability coefficient of the support to be identified; Where, is the stress safety factor, is the strain safety factor, is the anchor axial force safety factor, is the maximum stress that the support can withstand, is the maximum strain that the support can withstand, The maximum axial force allowed in anchor design. is the maximum stress value of the support, is the maximum strain value of the support, is the maximum axial force of the anchor rod.

7. The method for determining the support stability of a compressed air energy storage cavern according to claim 6 is characterized in that: The specific process of step S6 is as follows: The stability of the support is determined by the stress safety factor , strain safety factor , Anchor axial force safety factor The minimum value of determines that when any index does not meet the safety requirements, it will lead to support instability; when When , it means that the three safety factors are all greater than 1, then the ultimate bearing capacity of the support is greater than the actual load, and the support is stable; when When , it means that at least one of the three safety factors is equal to 1, then the actual load of the support reaches the ultimate bearing capacity and the support is on the verge of destruction; when When , it means that at least one of the three safety factors is less than 1, then the actual load of the support exceeds the ultimate bearing capacity and the support becomes unstable; Among them, "1" is the mathematical representation that the ultimate capacity is equal to the actual load, and it is also the dividing point between stability and instability.

Images