Compressed air energy storage underground cavern site selection method based on multi-scale stability evaluation

Abstract

The invention discloses a compressed air energy storage underground cavern site selection method based on multi-scale stability evaluation. The method comprises six main steps of establishing a multi-scale evaluation system, evaluating the stability of a regional structure, identifying and analyzing a safety island, analyzing the stability of a site in detail, comprehensively integrating multi-scale evaluation results and verifying the evaluation results. According to the method, a multi-scale structure stability evaluation system is established, a safety island concept and an identification method thereof are provided, an innovative scale conversion and data fusion algorithm is developed, comprehensive and multi-level evaluation from macroscopic to microscopic is realized, and the accuracy and reliability of CAES underground cavern site selection are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of compressed air energy storage, and in particular to a method for selecting a site for an underground cavern for compressed air energy storage based on multi-scale stability evaluation. More specifically, it is a method for selecting a site for an underground cavern for a compressed air energy storage power station based on multi-scale stability evaluation. Background Art

[0002] Compressed Air Energy Storage (CAES) is a high-capacity, long-cycle energy storage technology that has received widespread attention in recent years. The site selection of underground gas storage, the core component of the CAES system, directly affects the safety, economy and long-term stability of the power station. However, the complex geological structure environment poses a huge challenge to the site selection of CAES underground caverns, especially in tectonic active areas.

[0003] At present, the CAES underground cavern site selection evaluation methods mainly include the following categories: 1) Qualitative evaluation methods based on geological conditions: such as the geological suitability evaluation method proposed by the Electric Power Research Institute (EPRI), which mainly considers factors such as lithology, geological structure, and hydrogeology, but lacks quantitative analysis, which reduces the accuracy of site selection. 2) Single-scale quantitative evaluation methods: such as the multi-factor evaluation method based on GIS, although quantitative indicators are introduced, it is limited to regional scale evaluation, which reduces the accuracy of site selection. 3) Evaluation methods based on numerical simulation: such as the cavern stability evaluation method based on finite element analysis, which can perform detailed stress analysis, but it is difficult to consider the influence of large-scale geological factors, which reduces the accuracy of site selection.

[0004] In summary, these existing CAES underground cavern site selection evaluation methods have the following main problems: 1) The evaluation scale is single, which makes it difficult to fully reflect the multi-level structural characteristics; 2) The evaluation index system is not perfect and lacks systematicity; 3) Qualitative evaluation is mainly used, and there is a lack of quantitative evaluation methods; 4) Static evaluation is mainly used, which makes it difficult to reflect the dynamic change characteristics of structural stability; 5) Scale conversion and data fusion problems need to be solved urgently. The existing CAES underground cavern site selection evaluation technology is insufficient in data collection, analysis methods, optimization decision-making and automation level, which limits the efficiency of site selection. Specific problems include: difficult data fusion and long collection cycle; reliance on manual experience and lack of multi-scale analysis; low efficiency of multi-objective optimization and imperfect data conflict processing; lack of automation and dynamic update mechanism; complex site selection process and long decision cycle.

[0005] It can be seen that the problems existing in the existing CAES underground cavern site selection evaluation method seriously restrict the site selection efficiency and accuracy of CAES projects, and increase engineering risks and costs. Therefore, it is urgent to develop a CAES underground cavern site selection evaluation method that can reduce manual intervention, improve data fusion capabilities, and achieve real-time response and optimization, improve the site selection efficiency and accuracy of CAES projects, and reduce engineering risks and costs. Summary of the invention

[0006] The purpose of the present invention is to provide a compressed air energy storage underground cavern site selection method based on multi-scale stability evaluation, which is a compressed air energy storage power station underground cavern site selection method based on multi-scale structural stability evaluation. It comprehensively considers multi-scale geological factors, and can effectively perform data fusion and scale conversion to improve the site selection accuracy of CAES projects. It introduces intelligent data analysis, automation systems and multi-scale dynamic models to improve the site selection efficiency of CAES projects, thereby reducing manual intervention, improving data fusion capabilities, and realizing real-time response and optimization, reducing engineering risks and costs.

[0007] In order to achieve the above-mentioned purpose, the technical solution of the present invention is: a method for selecting a site for an underground cavern for compressed air energy storage based on multi-scale stability evaluation, characterized in that it comprises the following steps:

[0008] Step 1: Establish a multi-scale evaluation system, including regional scale, mesoscale and microscale, and formulate data requirements and analysis framework for multi-scale evaluation;

[0009] Step 2: Conduct regional structural stability assessment;

[0010] Step 3: Identify and analyze safe islands;

[0011] Step 4: Conduct detailed site stability analysis;

[0012] Step 5: Comprehensive integration of multi-scale evaluation results;

[0013] Step 6: Verify the evaluation results;

[0014] The verification of the evaluation results at least includes cross-validation, field verification and historical case comparison.

[0015] In the above technical solution, in step 1, the scale range of the regional scale in the multi-scale evaluation system is 10 4 -10 5 km 2; The contents of regional scale include: regional tectonic pattern analysis to identify large-scale tectonic trends in geology, identification of potential active structures, assessment of crustal movement characteristics and regional stress field analysis; regional scale uses remote sensing images, seismic and GPS data using improved DS evidence theory and wavelet transform fusion algorithm for data fusion and scale conversion;

[0016] The scale range of the mesoscale is 10 2 -10 3 km 2 ; The contents of the mesoscale include safety island identification, long-term stability assessment, and earthquake resistance analysis; After the long-term stability of the safety island is assessed at the mesoscale, the improved DS evidence theory and wavelet transform fusion algorithm are used for data fusion and scale conversion;

[0017] The microscopic scale ranges from 1-10 km 2 ; The contents of micro-scale include detailed engineering address mapping, high-precision geophysical detection, site rock mass quality assessment and ground stress state analysis; after conducting on-site investigation and high-precision detection at the micro-scale, the improved DS evidence theory and wavelet transform fusion algorithm are used for data fusion and scale conversion;

[0018] After data fusion and scale conversion at the regional scale, mesoscale and microscale, the optimal site selection scheme is formed. Each level in the above multi-scale evaluation system (such as regional scale, mesoscale and microscale) has corresponding evaluation criteria. Different projects have different evaluation criteria, which can be based on the following sources:

[0019] National or industry standards: such as national standards or industry standards in the fields of geological exploration, rock mass quality classification, fault stability analysis, etc.

[0020] Project-specific standards: Evaluation standards set according to specific project needs, such as setting specific ground stress, seismic performance and rock integrity standards in the site selection of underground caverns for compressed air energy storage;

[0021] Expert experience and historical data: Based on the experience data of similar projects, combined with expert judgment, specific evaluation criteria are formed.

[0022] Therefore, evaluation standards of different scales are formulated according to project requirements and data conditions to ensure the accuracy and operability of the evaluation system.

[0023] In the above technical solution, in step 2, regional structural stability evaluation is performed, including the following steps:

[0024] Step 2.1: Regional structural pattern analysis; Step 2.1 completes the specific collection, processing and application of remote sensing images required in step 1. Therefore, the remote sensing images in step 2.1 correspond to the framework planning of step 1 and are implemented in the execution process; the present invention involves new methods such as multi-scale data fusion, introduction of fractal theory, and deep learning algorithm-assisted optimization. The present invention improves the depth, accuracy and speed of regional structural stability evaluation by creatively introducing deep learning algorithms (using machine learning algorithms (such as convolutional neural networks) to analyze earthquake source data and improve the model's stress distribution prediction ability), dynamically monitoring and predicting regional stress fields (combining InSAR monitoring data to achieve dynamic updating and prediction of stress fields), and performing multi-scale data fusion and nonlinear solution (using improved wavelet transform algorithms or Dempster-Shafer evidence theory to fuse different data sources and process uncertainty between data).

[0025] The regional structural pattern analysis method is: using deep learning algorithms (such as U-Net) to extract linear structural features in remote sensing images, and combining geological maps and digital elevation models to build a regional structural pattern model;

[0026] The construction of the regional tectonic pattern model is not only a macroscopic grasp of regional geology, but also provides key support for the entire multi-scale site selection and evaluation process; although its method is based on existing remote sensing and geological analysis technology, it improves the accuracy and efficiency of structural feature extraction by introducing innovative methods such as deep learning. Therefore, the construction method of this model plays a vital and fundamental role in the entire evaluation system;

[0027] The construction of the regional tectonic pattern model in the present invention combines deep learning algorithms (such as U-Net) with traditional geological analysis methods. In terms of deep learning and multi-source data integration, this method introduces new data processing and model construction technologies, which improves the depth and accuracy of the evaluation. The specific construction method includes the following steps:

[0028] Step 2.1.1: Remote sensing image data acquisition and processing:

[0029] Use satellite imagery (such as Sentinel or Landsat) to obtain high-resolution imagery.

[0030] The linear structure features in the image are identified by using the improved Canny edge detection algorithm or Hough transform.

[0031] Step 2.1.2: Deep learning model application:

[0032] The U-Net deep learning algorithm is introduced to train and analyze remote sensing images and automatically extract complex linear structural features.

[0033] A large number of manually annotated geological samples are used when training the model to ensure that the model has high recognition accuracy.

[0034] Step 2.1.3: Digital Elevation Model (DEM) analysis:

[0035] The terrain relief information of the region is extracted in combination with DEM data to further support the structural pattern analysis.

[0036] Step 2.1.4: Fusion of geological map and remote sensing data:

[0037] The structural features extracted from remote sensing images are combined with geological map information to form a comprehensive structural model.

[0038] Step 2.1.5: Construction of regional tectonic pattern model:

[0039] Based on the above data, an overall regional tectonic pattern model is constructed, and the main tectonic belts, fault zones and crustal deformation areas are marked.

[0040] The construction method of the regional tectonic pattern model introduces deep learning technology into traditional geological analysis, which improves the accuracy and efficiency of structural feature extraction. Compared with the traditional structural model construction method that relies on manual analysis, this comprehensive analysis method that integrates remote sensing, DEM and deep learning is more automated and accurate.

[0041] Step 2.2: Activity structure identification;

[0042] Active faults are identified by combining InSAR deformation monitoring, seismicity analysis and geological surveys; the InSAR deformation monitoring, seismicity analysis and geological survey data are mainly from existing data provided by the official website; this step integrates these existing data resources and provides a scientific basis for the structural stability evaluation of the site through in-depth analysis;

[0043] Step 2.3: Assessment of regional crustal movement characteristics and stress field;

[0044] Using GPS observation data and earthquake focal mechanism solutions, combined with finite element numerical simulation, a regional stress field model is constructed;

[0045] Methods for constructing regional stress field models include:

[0046] Step 2.3.1: Data acquisition and preprocessing:

[0047] Collect GPS observation data to determine the velocity field of crustal movement.

[0048] The stress direction of regional seismic activity is obtained using earthquake focal mechanism solution data.

[0049] Apply existing finite element numerical simulation tools to integrate crustal deformation data.

[0050] Step 2.3.2: Data fusion and model building:

[0051] The GPS data and earthquake focal mechanism solution are integrated to generate the stress field distribution in the area through interpolation or Kriging method.

[0052] Finite element simulations were used to further optimize the stress field model, taking into account the fracture surface, boundary conditions and crustal mechanical parameters.

[0053] Step 2.3.3: Model calibration and validation:

[0054] The accuracy of the model is verified by comparing the historical stress field observation data with the model results.

[0055] The stress data from field measurements (such as hydraulic fracturing) are used to further calibrate the model parameters. The construction method of the regional stress field model introduces deep learning, dynamic prediction or multi-data fusion, which plays a key role in the site selection process of the underground cavern of the compressed air energy storage power station, assists in identifying risk areas, supports the design optimization of the underground cavern of the compressed air energy storage power station, and provides a scientific basis for long-term stability evaluation.

[0056] Step 2.4: Establish a regional scale evaluation index system and determine the weights;

[0057] The evaluation indicators include active fault density, seismicity, crustal deformation rate, tectonic stress intensity and topographic relief, and the weights are determined by using the improved analytic hierarchy process (AHP);

[0058] The specific methods for establishing a regional scale evaluation index system and determining its weights include:

[0059] Introduce machine learning algorithms to determine weights: Use support vector machines (SVM) or random forest algorithms to automatically analyze historical data and optimize indicator weights;

[0060] Weight adjustment based on multi-scale data fusion: Dempster-Shafer evidence theory is used to handle the fusion and conflict of multi-source data and dynamically adjust the indicator weights;

[0061] Dynamic weight update: According to the real-time data of InSAR monitoring and seismic activity, the weights are adjusted in real time so that the model can dynamically respond to geological changes;

[0062] Application of multi-objective optimization models:

[0063] Use multi-objective decision analysis (such as TOPSIS) to comprehensively consider engineering geology, economic cost and construction difficulty to provide the optimal solution for regional stability evaluation;

[0064] The present invention introduces new methods such as machine learning, multi-objective optimization or dynamic data fusion in the index selection and weight determination methods of establishing a regional scale evaluation index system and determining the weights, and systematically conducts quantitative analysis on various factors related to structural stability, providing scientific and reliable support for engineering site selection.

[0065] The above steps 2.1 to 2.4 are interconnected and executed in sequence. The results of the previous step are the basis for the analysis of the next step. Each step is gradually deepened to ensure that the stability evaluation of the regional structure is comprehensively covered from macro to specific, from structure to mechanics. The progressive relationship between the above steps improves the depth and accuracy of the evaluation.

[0066] In the above technical solution, in step 3, safety island identification and analysis are performed, including the following steps:

[0067] Step 3.1: Identify relatively stable structural blocks;

[0068] Based on the principles of structural integrity, lithological consistency and deformation degree similarity, combined with multi-source geophysical data, safe islands are identified;

[0069] Step 3.2: Assess the long-term stability and earthquake resistance of the safety island;

[0070] Probabilistic seismic hazard analysis and dynamic finite element analysis are used to assess the long-term stability of the safety island;

[0071] Step 3.3: Establish a mesoscale evaluation indicator system and determine the weights;

[0072] On the basis of regional scale indicators, rock mass integrity, groundwater conditions and other indicators are added, and the weights are determined by fuzzy analytic hierarchy process;

[0073] The above steps 3.1 to 3.3 are interconnected and executed in sequence. The latter step is based on the analysis results of the previous step, and gradually deepens the identification, stability analysis and comprehensive evaluation of the structural blocks. Finally, after the progressive analysis of the above three steps, it is ensured that the selected safety island reaches the optimal stability, seismic resistance and comprehensive geological conditions, providing scientific support for the project site selection.

[0074] In the above technical solution, in step 4, a detailed analysis of site stability is conducted, including the following steps:

[0075] Step 4.1: Detailed engineering geological survey;

[0076] Use digital field systems and 3D laser scanning technology to obtain high-precision geological information;

[0077] Step 4.2: High-precision geophysical exploration;

[0078] Combine high-resolution 3D seismic, electrical exploration and geological radar detection to build a detailed geological structure model;

[0079] Step 4.3: Site rock mass quality assessment;

[0080] Assess the site's rock mass quality, fracture development and ground stress status;

[0081] An improved GSI system is used to assess rock mass quality and a variety of geostress measurement methods are combined to determine the initial geostress state;

[0082] Step 4.4: Establish a micro-scale evaluation index system and determine the weights;

[0083] Engineering scale indicators are added, such as rock deformation modulus and initial geostress state, and the weights are determined using a combined weighting method.

[0084] In the above technical solution, in step 5, comprehensive integration of multi-scale evaluation results is performed, including the following steps:

[0085] Step 5.1: Multi-scale data fusion;

[0086] Multi-scale data fusion includes data standardization processing, multi-source data fusion algorithm and data quality control;

[0087] The multi-scale data fusion adopts the improved Dempster-Shafer evidence theory and wavelet transform fusion algorithm;

[0088] Step 5.2: Scale conversion;

[0089] The scale conversion adopts a scale conversion model based on fractal theory;

[0090] Step 5.3: Comprehensive evaluation;

[0091] An improved analytic hierarchy process (AHP) is used for comprehensive evaluation: fuzzy theory is introduced to deal with uncertainty and improve the reliability of evaluation results;

[0092] A multi-objective decision analysis method is used to comprehensively consider engineering geological conditions, construction difficulty and long-term stability to optimize and compare site selection options.

[0093] In the above technical solution, in step 5.1, the multi-scale data fusion includes the following steps:

[0094] Step 5.1.1, data standardization:

[0095] Unit unification: Establish a unit conversion matrix to take into account the differences between different measurement standards;

[0096] Dimension normalization: Z-score normalization method is used;

[0097] Coordinate system conversion: unify the spatial reference systems of different data sources;

[0098] Step 5.1.2, multi-source data fusion algorithm:

[0099] For the same type of data, the weighted average method is used;

[0100] For uncertain data, Dempster-Shafer evidence theory is used;

[0101] For multi-resolution data, wavelet transform fusion is used;

[0102] Step 5.1.3, data quality control:

[0103] Outlier detection and processing: Use the 3σ criterion to identify outliers;

[0104] Data integrity check: Multiple imputation was used to handle missing values;

[0105] Consistency check: Establish data priority rules to resolve data conflicts.

[0106] In the above technical solution, in step 5.2, the scale conversion algorithm specifically includes:

[0107] 1) Upscaling method:

[0108] Simple statistical method: select arithmetic mean, geometric mean or weighted mean according to the characteristics of the data;

[0109] Geostatistical methods: Block Kriging was used to consider spatial correlation;

[0110] 2) Downscaling method:

[0111] Dynamic downscaling: using high-resolution numerical models to take into account the influence of local topography and geological factors;

[0112] Statistical downscaling: establishing statistical relationships between large and small scale predictors, such as multivariate regression models;

[0113] 3) Scale effect treatment of rock mechanics parameters:

[0114] Intensity parameters: the modified Hoek-Brown criterion is adopted, and the GSI value and disturbance factor D are introduced;

[0115] Deformation modulus: using the modified Hoek-Diederichs empirical formula;

[0116] Permeability: Using fractal theory, the characteristic length and fractal dimension of the fracture network are considered.

[0117] In the above technical solution, in step 3.1, the identification criteria of the safety island include:

[0118] (1) Internal fault development degree: fault density is less than 0.5 km / km 2 ;

[0119] (2) Boundary fault activity: The average slip rate over the past 100,000 years is less than 0.1 mm / year;

[0120] (3) Earthquake activity level: Annual average earthquake frequency (M≥3.0) is less than 1 / 100km 2 ;

[0121] (4) Terrain undulation: less than 200m / 10km 2 .

[0122] In the above technical solution, in step 3.1, multi-source data fusion and machine learning methods are used to identify the safety island. The specific steps are as follows:

[0123] Step 3.1.1: Data preparation and multi-source data collection;

[0124] Collect multi-source data, including high-resolution remote sensing images, digital elevation models (DEMs), geological maps, earthquake catalogs, GPS observations, and InSAR deformation monitoring data;

[0125] Step 3.1.2: Feature extraction;

[0126] (1) Extracting linear structural features from remote sensing images and DEM:

[0127] Use the improved Canny edge detection algorithm to identify linear structures in images; apply Hough transform to identify the direction of linear features;

[0128] (2) Calculate terrain relief;

[0129] Using the moving window method, the maximum elevation difference of each 10km×10km grid was calculated;

[0130] (3) Calculate active fault density;

[0131] The total length of the fault was counted in a sliding window of 100 km × 100 km;

[0132] (4) Calculate seismicity;

[0133] Count the frequency and maximum magnitude of historical earthquakes, and analyze the distribution of fault zones;

[0134] Step 3.1.3: Construct feature vector;

[0135] For each 10km×10km grid, construct the feature vector:

[0136] X=[x1,x2,x3,x4,x5]

[0137] Among them: x1: linear structure density; x2: terrain relief; x3: active fault density; x4: seismic activity index; x5: crustal deformation rate;

[0138] Step 3.1.4: Machine learning classification;

[0139] (1) Use the support vector machine (SVM) algorithm for binary classification, including:

[0140] Positive category: potential safe island;

[0141] Negative category: unstable area;

[0142] (2) SVM decision function, that is, using the SVM model for training and classification to output potential safe island areas;

[0143] f(X)=sign(∑(α i y i K(X i ,X))+b);

[0144] Where K is the radial basis function (RBF) kernel;

[0145] K(X i ,X)=exp(-γ||X i X|| 2 )

[0146] Step 3.1.5: Optimize the safety island boundary;

[0147] Use the Active Contour Model to refine the safety island boundary. ACM refines the boundary through the energy function, where (C(s)) is the contour curve;

[0148] The energy function is:

[0149] E=∫[α|C'(s)| 2 +β|C”(s)| 2 +Eext(C(s))]ds

[0150] Where C(s) is the contour curve, Eext is the external energy term;

[0151] Step 3.16: Multi-scale verification;

[0152] Repeat steps 3.12 to 3.15 at different spatial scales (e.g. 5km, 20km, 50km) and use majority voting to determine the final safe island range;

[0153] The method for identifying safety islands in the above step 3.1 is based on existing geophysical analysis and geological survey technologies (such as multi-source data fusion, fault structure analysis, seismic activity monitoring, etc.). It has been innovated through multi-source data fusion, machine learning classification and boundary optimization algorithms, making the identification of safety islands more accurate and efficient, and providing a scientific basis and important support for project site selection.

[0154] The beneficial effects of the present invention are:

[0155] (1) A multi-scale structural stability evaluation system is established. The multi-scale structural stability evaluation system of the present invention is intended to comprehensively analyze the geological conditions of the project site selection area from multiple levels: macroscopic (regional scale), mesoscopic (structural block or "safety island" level), and microscopic (local rock mass characteristics and ground stress state); the structure and method of the system fully realize the all-round and multi-level evaluation from macroscopic to microscopic, significantly improving the comprehensiveness and systematicness of the evaluation, and ensuring the comprehensiveness, accuracy and reliability of the site selection results;

[0156] (2) The concept of safety island and its identification and analysis methods are proposed, which provides a new idea for the evaluation of large-scale structural stability. In the site selection method of the present invention, the concept of safety island not only involves the local stability of the project area, but also covers the comprehensive analysis and evaluation of larger scales (such as regional structural pattern and crustal movement characteristics). Therefore, the identification process of safety island is an important part of multi-scale analysis. By evaluating the structural stability at the macro, meso and micro scales, the comprehensiveness, safety and scientificity of the project site selection are ensured.

[0157] (3) Developed innovative scale conversion and data fusion methods, effectively solved the problem of integrating information at different scales, and improved the accuracy and reliability of evaluation results;

[0158] (4) A site selection decision process based on multi-scale evaluation was proposed, which provided an operational method for engineering practice and improved the scientificity and efficiency of site selection decisions;

[0159] (5) Verified by actual engineering cases, compared with the traditional single-scale method, the present invention has significant advantages in site selection accuracy and risk control: the site selection accuracy is improved by about 75%, the engineering risk is reduced by about 40%, and the overall cost-effectiveness ratio is better than the traditional method. BRIEF DESCRIPTION OF THE DRAWINGS

[0160] Figure 1 It is a schematic diagram of the framework of the multi-scale structural stability evaluation system of the present invention.

[0161] Figure 2 This is a flow chart of the multi-scale evaluation method of the present invention.

[0162] Figure 3 This is a framework diagram of the multi-scale data fusion technology of the present invention.

[0163] Figure 4 This is a schematic diagram of the application of the improved analytic hierarchy process (AHP) in multi-scale evaluation.

[0164] Figure 5 This is a comparison chart of the evaluation result verification method of the present invention.

[0165] Figure 6 This is a flow chart of the multi-scale data fusion algorithm of the present invention.

[0166] Figure 7 This is a flow chart of the safety island identification algorithm of the present invention. DETAILED DESCRIPTION

[0167] The following is a detailed description of the implementation of the present invention in conjunction with the accompanying drawings, but they do not constitute a limitation of the present invention and are only given as examples. At the same time, the advantages of the present invention are made clearer and easier to understand through the description.

[0168] Referring to the attached drawings, it can be seen that a method for selecting a site for an underground cavern of a compressed air energy storage power station based on multi-scale structural stability evaluation comprises the following steps:

[0169] Step 1: Establish a multi-scale evaluation system, including regional scale (10 4 -10 5 km 2 ), mesoscale (10 2 -10 3 km 2 ) and microscopic scale (1-10km 2 ) three levels; The purpose of establishing a multi-scale evaluation system (step one) in the present invention is to provide a systematic and comprehensive geological stability assessment framework for the siting of underground caverns for compressed air energy storage (CAES) power stations. By dividing the evaluation into three levels: regional scale, mesoscale and microscale, it ensures that the site selection process can not only evaluate the suitability of the regional geological environment from a macro perspective, but also analyze the stability of specific locations in detail at the micro level. The core purpose of establishing a multi-scale evaluation system is to provide a logically clear and interlocking process for subsequent steps, advancing from macro to micro levels, to achieve precision and risk control in underground cavern site selection;

[0170] Step 2: Conduct a regional structural stability assessment, preliminarily screen suitable areas from a macroscopic perspective, and determine the range that is stable and has low risk at the geological structural level. This stage focuses on large-scale geological structures and stress distribution to ensure that subsequent more detailed analysis can be concentrated on areas with relatively favorable geological conditions, thereby improving the efficiency and accuracy of site selection. The core purpose of step two is to ensure the structural stability of the site selection range and lay the foundation for subsequent safety island identification and detailed analysis. Through large-scale screening and analysis, it significantly improves the efficiency and accuracy of site selection, reduces the complexity of subsequent analysis, and ensures the long-term stability and safety of the project;

[0171] Step 3: Identify and analyze the safety island. By accurately identifying and analyzing the safety island, scientific and safe candidate areas are provided for project site selection. The step of identifying and analyzing the safety island plays a connecting role in the overall site selection process, that is, it takes over the regional stability evaluation results and lays the foundation for detailed analysis of the site. It also provides high-quality data support for subsequent multi-scale data fusion and verification.

[0172] The safety island in the present invention refers to an innovative concept proposed to meet the specific needs of the site selection process of the underground cavern of the compressed air energy storage power station. It essentially refers to an area with stable geological structure and strong earthquake resistance, providing a priority area for subsequent detailed site selection analysis;

[0173] Step 4: Conduct a detailed site stability analysis, comprehensively analyze the stability of the candidate area at a microscopic scale, and provide a detailed scientific basis for the final site selection and engineering design. It plays a role in fine verification and data deepening in the entire process, not only supporting the comprehensive integration of multi-scale data, but also providing a solid foundation for subsequent result verification. Through this link, the accuracy of site selection and the feasibility of the project can be greatly improved, ensuring the long-term safety and stability of the project;

[0174] Step 5: Comprehensively integrate the multi-scale evaluation results, systematically integrate the analysis results of different scales, and provide comprehensive decision support for the site selection plan. It plays a role of coordination and optimization in the entire process, and provides a scientific and reliable basis for project implementation through data fusion, conflict resolution and comprehensive evaluation. This link ensures the coordination and consistency of analysis results from macro to micro, providing a strong guarantee for the final site selection;

[0175] Step 6: Verify the evaluation results. Through the systematic verification process, ensure the scientific nature of the evaluation model and the reliability of the evaluation results. It is the final link in the entire site selection process and provides an important guarantee for the safe implementation of the project. Discovering and correcting potential problems through verification will help to further optimize the site selection plan, reduce engineering risks, and ensure the long-term stability of the project under complex geological conditions. The verification standards for the above-mentioned verification and evaluation are derived from national standards, project requirements, historical experience, etc.; when there are no clear verification standards, the reliability of the verification results can be alternatively verified through methods such as expert review, historical comparison, field tests and sensitivity analysis. The present invention ensures the eligibility of the evaluation results of the method of the present invention by verifying the evaluation results, thereby providing a reliable basis for engineering site selection and design.

[0176] As a preferred embodiment of the present invention, step 2 comprises:

[0177] Step 2.1, regional structural pattern analysis: use deep learning algorithms (such as U-Net) to extract linear structural features in remote sensing images, and build a regional structural pattern model by combining geological maps and digital elevation models;

[0178] Step 2.2, identifying active structures: combining InSAR deformation monitoring, seismicity analysis and geological surveys to identify active faults;

[0179] Step 2.3, evaluate the regional crustal movement characteristics and stress field distribution: construct a regional stress field model using GPS observation data and earthquake focal mechanism solutions combined with finite element numerical simulation;

[0180] Step 2.4, establish a regional scale evaluation index system and determine the weights: the evaluation indexes include active fault density, seismicity, crustal deformation rate, tectonic stress intensity and terrain relief, and the weights are determined using the improved analytic hierarchy process (AHP).

[0181] As a preferred embodiment of the present invention, step 3 comprises:

[0182] Step 3.1, identify relatively stable structural blocks: Based on the principles of structural integrity, lithological consistency and deformation degree similarity, combined with multi-source geophysical data, identify "safety islands";

[0183] Step 3.2, evaluate the long-term stability and seismic resistance of the safety island: use the probabilistic seismic hazard analysis method and dynamic finite element analysis to evaluate the long-term stability of the safety island;

[0184] Step 3.3, establish a meso-scale evaluation index system and determine the weights: add indicators such as rock mass integrity and groundwater conditions on the basis of regional scale indicators, and use fuzzy hierarchical analysis method to determine the weights.

[0185] As a preferred embodiment of the present invention, step 4 comprises:

[0186] Step 4.1, conduct detailed engineering geological survey: use digital field system and 3D laser scanning technology to obtain high-precision geological information;

[0187] Step 4.2, conduct high-precision geophysical exploration: combine high-resolution three-dimensional seismic, electrical exploration and geological radar detection to build a fine geological structure model; at the same time, introduce convolutional neural network (CNN) to analyze seismic wave reflection data for deep learning-assisted modeling to improve modeling accuracy and efficiency; adopt multi-scale data fusion algorithm, use Dempster-Shafer evidence theory, fuse seismic, electrical, and radar data at different scales, and handle data conflicts; use Dempster-Shafer evidence theory to fuse seismic, electrical, and radar data at different scales, and handle data conflicts; introduce sensitivity analysis methods to automatically adjust modeling parameters and optimize model results; by introducing multi-source data fusion, deep learning, real-time update and other methods, not only detailed geological information is provided to support project site selection and construction optimization, but also geological risks can be predicted to ensure the safety and economy of the project;

[0188] Step 4.3, evaluate the rock mass quality, fracture development degree and geostress state of the site: use the improved GSI system to evaluate the rock mass quality, and combine multiple geostress measurement methods to determine the initial geostress state; at the same time, use the Dempster-Shafer evidence theory or Bayesian method to integrate the results of field survey, remote sensing monitoring and numerical simulation, handle data uncertainty, and realize multi-source data fusion; use machine learning models (such as random forests or SVM) to assist in analysis, and predict the rock mass quality and geostress distribution based on existing project data; use convolutional neural networks (CNN) to analyze remote sensing images to improve the automation of fracture identification; combine InSAR and GPS real-time monitoring data to achieve dynamic updating of stress fields, predict potential risks of crustal movement, and realize real-time geostress monitoring; the present invention combines multi-source data fusion, machine learning, real-time monitoring and other evaluations to play a key role in project site selection, construction design and geological risk prediction, ensuring the safety and economy of the project;

[0189] The above-mentioned determination of the initial geostress state is used to provide basic data for engineering design, support the optimization of construction plans, prevent geological disasters, and ensure the long-term stability and safety of the project. Determining the initial geostress state plays a key role in construction safety, cost control and risk management, and is an indispensable part of underground engineering site selection and construction.

[0190] Step 4.4, establish a micro-scale evaluation index system and determine the weights: add engineering scale indicators, such as rock deformation modulus and initial geostress state, and use a combined weighting method to determine the weights. In the process of establishing a micro-scale evaluation index system, the present invention uses a random forest algorithm or a neural network, and automatically optimizes the weights of evaluation indicators based on historical data through intelligent optimization and machine learning assisted analysis; combines InSAR monitoring data and field measurement results to achieve real-time data integration and dynamic update of evaluation indicators; integrates the micro-scale evaluation results with the meso- and macro-scale evaluation systems, and adjusts the micro-index system through a feedback mechanism to achieve multi-scale fusion and feedback adjustment; uses Bayesian networks or Dempster-Shafer evidence theory to handle uncertainties in multi-source data and improve the robustness of the model; the above-mentioned establishment of a micro-scale evaluation index system introduces innovative means of intelligent analysis, dynamic update and multi-scale fusion, which plays a key role in engineering design and construction management, and guides the optimization of construction plans, reduces risks, and improves the economy and safety of projects through refined evaluation indicators;

[0191] As a preferred embodiment of the present invention, step 5 comprises:

[0192] Step 5.1, using multi-scale data fusion technology: including data standardization processing, multi-source data fusion algorithm and data quality control; the role of multi-scale data fusion in the present invention is to integrate and optimize data from different spatial scales (such as macro scale, meso scale, micro scale) to comprehensively and accurately describe the geological environment and stress state of the engineering area, provide comprehensive and accurate geological evaluation results, and support the optimization of engineering decisions. Data at different scales have different focuses, and fusing this information can make up for the shortcomings of a single data source and ensure the scientific nature of the site selection and construction plan;

[0193] Step 5.2, use the improved analytic hierarchy process (AHP) to conduct comprehensive evaluation: introduce fuzzy theory to deal with uncertainty and improve the reliability of evaluation results;

[0194] Step 5.3, optimize and compare the site selection options: adopt a multi-objective decision analysis method to comprehensively consider engineering geological conditions, construction difficulty and long-term stability.

[0195] As a preferred solution of the present invention, the multi-scale data fusion technology includes the following steps:

[0196] Step 5.1.1: Data collection and classification:

[0197] Collect various types of data according to the classification of macro scale, meso scale and micro scale:

[0198] Macro scale: regional tectonic pattern, earthquake catalog, InSAR monitoring data.

[0199] Mesoscale: fault development and tectonic block analysis.

[0200] Micro scale: rock mass quality, ground stress state, drilling data.

[0201] Step 5.1.2, data standardization: The data units and formats of different scales are different, so standardization processing (such as normalization and data interpolation) is used to ensure the comparability of data;

[0202] Unit unification: Establish a unit conversion matrix to take into account the differences between different measurement standards;

[0203] Dimension normalization: Z-score normalization method is used;

[0204] Coordinate system conversion: unify the spatial reference systems of different data sources;

[0205] Step 5.1.3, multi-source data fusion algorithm:

[0206] Dempster-Shafer evidence theory: handles the uncertainty of multi-source data and fuses conflicting data.

[0207] Weighted average method and analytic hierarchy process (AHP): assign weights to data of different scales to achieve comprehensive analysis.

[0208] Bayesian network: integrates prior information and observation data at multiple levels to improve the accuracy of prediction;

[0209] Step 5.1.4, construction of multi-scale model:

[0210] Through the fused data, a multi-scale geological model is established to visualize the geological environment and stress state at different spatial scales;

[0211] Step 5.1.5, dynamic update and feedback mechanism:

[0212] By integrating monitoring data (such as InSAR and GPS), the model can be dynamically updated to reflect geological changes in a timely manner.

[0213] Establish a feedback mechanism to adjust model weights based on real-time data and optimize data fusion results.

[0214] Step 5.1.6, verification and optimization of fusion results:

[0215] Compare the fusion results with the actual observation data (such as drilling survey data) to verify and optimize the results to ensure the accuracy of the evaluation model;

[0216] Step 5.1.7, data quality control:

[0217] Outlier detection and processing: Use the 3σ criterion to identify outliers;

[0218] Data integrity check: Multiple imputation was used to handle missing values;

[0219] Consistency check: Establish data priority rules to resolve data conflicts. The present invention achieves efficient integration of data of different scales through the above-mentioned data standardization, intelligent fusion algorithm and dynamic update mechanism. This process not only reduces the inconsistency between data, improves the accuracy of the model, but also enhances the ability of engineering monitoring and control.

[0220] As a preferred solution of the present invention, the method for comprehensive integration of multi-scale evaluation results also includes a scale conversion algorithm. The role of scale conversion is to seamlessly connect and transform data between different spatial scales (macro, meso, micro), ensuring that data at all levels can be compatible and used in a coordinated manner during comprehensive analysis and decision-making. Due to differences in data granularity, resolution, and objectives at different scales, scale conversion helps optimize data integration, reduce errors, and improve the scientific nature of project site selection and construction design;

[0221] Scale conversion algorithm, including:

[0222] 1) Upscaling method:

[0223] Simple statistical method: select arithmetic mean, geometric mean or weighted mean according to the characteristics of the data;

[0224] Geostatistical methods: Block Kriging was used to consider spatial correlation;

[0225] 2) Downscaling method:

[0226] Dynamic downscaling: using high-resolution numerical models to take into account the influence of local topography and geological factors;

[0227] Statistical downscaling: establishing statistical relationships between large and small scale predictors, such as multivariate regression models;

[0228] 3) Scale effect treatment of rock mechanics parameters:

[0229] Intensity parameters: the modified Hoek-Brown criterion is adopted, and the GSI value and disturbance factor D are introduced;

[0230] Deformation modulus: using the modified Hoek-Diederichs empirical formula;

[0231] Permeability: Using fractal theory, the characteristic length and fractal dimension of the fracture network are considered.

[0232] As a preferred solution of the present invention, the safety island identification algorithm adopts multi-source data fusion and machine learning methods, and the specific steps are as follows:

[0233] Step 3.1.1: Data preparation

[0234] Collect the following data: high-resolution remote sensing images (such as Sentinel-2 multispectral images), digital elevation models (DEMs), geological maps, earthquake catalogs, GPS observation data, and InSAR deformation monitoring data;

[0235] Step 3.1.2: Feature extraction;

[0236] (1) Extracting linear structural features from remote sensing images and DEM:

[0237] Use the improved Canny edge detection algorithm;

[0238] Apply Hough transform to identify linear features;

[0239] (2) Calculate terrain relief:

[0240] Using the moving window method, calculate the maximum elevation difference for each 10km×10km grid

[0241] (3) Calculation of active fault density:

[0242] The total length of the fault is calculated in a sliding window of 100 km × 100 km.

[0243] (4) Calculation of seismicity:

[0244] Statistics on the frequency and maximum magnitude of historical earthquakes;

[0245] Step 3.1.3: Construct feature vectors; for each 10km×10km grid, construct feature vectors: X=[x1,x2,x3,x4,x5]; where: x1: linear structure density; x2: terrain relief; x3: active fault density; x4: seismicity index; x5: crustal deformation rate;

[0246] Step 3.1.4: Machine learning classification;

[0247] (1) Use the support vector machine (SVM) algorithm for binary classification:

[0248] Positive category: potential safe island;

[0249] Negative category: unstable area;

[0250] (2) SVM decision function:

[0251] f(X)=sign(∑(α i yi K(X i ,X))+b)

[0252] Where K is the radial basis function (RBF) kernel:

[0253] K(X i ,X)=exp(-γ||X i X|| 2 )

[0254] Step 3.1.5: Optimize the safety island boundary;

[0255] Use the Active Contour Model to refine the "safety island" boundary.

[0256] The energy function is:

[0257] E=∫[α|C'(s)| 2 +β|C”(s)| 2 +Eext(C(s))]ds

[0258] Where C(s) is the contour curve and Eext is the external energy term. The energy function is used in the Active Contour Model (ACM) or Snake algorithm, and is mainly used in image segmentation, edge detection and shape recognition. It describes the energy state of a curve (C(s)) in the image, and minimizes the energy to make the curve fit the target feature boundary as closely as possible, thereby detecting the target contour;

[0259] Step 3.1.6: Multi-scale verification;

[0260] Repeat steps 3.12 to 3.15 at different spatial scales (e.g., 5 km, 20 km, 50 km) to ensure comprehensive coverage of geological features, reduce the error of single-scale analysis, and optimize the site selection range;

[0261] At the same time, the majority voting method is used to determine the final scope of the safe island. The majority voting method is used to automatically integrate the results of different scales to ensure the coordination and consistency between multi-scale analyses. The process is executed through computer algorithms rather than human intervention to achieve scientific and objective decision-making.

[0262] As a preferred solution of the present invention, the present invention adopts a scale conversion model based on fractal theory for scale conversion, which is mainly used for processing the scale effect of rock mass mechanical parameters. The following is the mathematical derivation process of the scale conversion model:

[0263] (1) Establish the relationship between fractal dimension and scale;

[0264] Assuming that the rock mass fracture network obeys fractal distribution, its fractal dimension D and the observation scale L have the following relationship:

[0265] N(L)=cL^(-D)

[0266] Where N(L) is the number of cracks with a characteristic size greater than L, and c is a constant.

[0267] (2) deriving scale effects of intensity parameters;

[0268] Based on the Hoek-Brown criterion, the relationship between rock mass strength σ and scale L can be expressed as:

[0269] σ(L)=σc(mb(L)σci / σc+s)^a

[0270] Where σc is the intact rock strength, σci is the indoor test strength, mb, s and a are rock mass constants.

[0271] Express mb as a function of scale L:

[0272] mb(L)=miexp((GSI100) / (2814D(L)))

[0273] Where mi is the rock constant and GSI is the geological strength index.

[0274] (3) derive the scale effect of deformation modulus;

[0275] Based on the modified Hoek-Diederichs empirical formula:

[0276] E(L)=E0(0.02+(1D(L) / 2) / (1+exp((60+15D(L)GSI) / 11)))

[0277] Where E0 is the deformation modulus of intact rock.

[0278] (4) Derive the scale effect of permeability coefficient;

[0279] Assume that the permeability coefficient K of the rock mass fracture network has a power law relationship with the scale L:

[0280] K(L)=K0(L / L0)^(D(L)-2)

[0281] Where K0 is the permeability coefficient at the reference scale L0.

[0282] (5) Comprehensive expression of scale conversion model;

[0283] Given the initial scale L1 and the target scale L2, the scale conversion of each parameter can be expressed as:

[0284] Strength: σ(L2)=σ(L1)*[mb(L2) / mb(L1)]^(a / 2)

[0285] Deformation modulus: E(L2) = E(L1)*[(1-D(L2) / 2) / (1-D(L1) / 2)]

[0286] Permeability coefficient: K(L2) = K(L1)*(L2 / L1)^(D(L2)-D(L1))

[0287] where D(L) can be obtained by inverting the relationship in step 1.

[0288] This scale conversion model takes into account the heterogeneity and anisotropy of the rock mass, can effectively deal with the changes in the mechanical parameters of the rock mass at different scales, and provides a reliable theoretical basis for the multi-scale structural stability evaluation.

[0289] Example

[0290] The present invention is now described in detail by taking the application of the present invention to a certain CAES project for selecting the site for an underground cavern of a compressed air energy storage power station as an example. The present invention can also serve as a guide for applying the present invention to other CAES projects for selecting the site for an underground cavern of a compressed air energy storage power station.

[0291] Example 1: The present invention is applied to a CAES project in a mountainous area to select a site for an underground cavern of a compressed air energy storage power station. The specific method is as follows:

[0292] Step 1: Establish a multi-scale evaluation system; Regional scale: 50,000km 2 ; Mesoscale: 500km 2 ; Microscopic scale: 5km 2 ;

[0293] Step 2: Regional structural stability assessment;

[0294] Step 2.1: Regional structural pattern analysis:

[0295] Using Landsat 8OLI multispectral images, a pre-trained U-Net model was used to extract linear structures. The model was fine-tuned on 5,000 manually annotated image segments, and the F1-score reached 0.85.

[0296] Step 2.2, active structure identification:

[0297] By processing the Sentinel-1SAR data from 2015 to 2023, three active faults were identified with annual slip rates of 0.5-2.5 mm / year.

[0298] Step 2.3: Regional crustal movement characteristics and stress field assessment:

[0299] Analysis of 10 years of observation data from 10 continuous GPS stations showed that the horizontal movement rate was 0.5-2mm / year.

[0300] The regional stress field was inverted using 50 high-quality focal mechanism solutions, and the maximum principal stress direction was NE40°±5°.

[0301] Step 2.4: Regional scale evaluation index system: Active fault density: 0.02-0.1km / km 2 ; Seismicity: Annual average earthquake frequency (M≥3.0) 0.2-1 times / 100km 2 , the largest historical earthquake is M6.5; crustal deformation rate: horizontal strain rate (2-8)×10 -9 / year; tectonic stress intensity: maximum principal stress value 15-25MPa; terrain relief: 200-800m;

[0302] Step 3: Identification and analysis of safety islands;

[0303] Step 3.1: Identify 5 potential "safety islands" with an area of ​​50-200km 2 .

[0304] Step 3.2, long-term stability assessment:

[0305] Probabilistic earthquake hazard analysis shows that the peak acceleration of an earthquake with a 50-year probability of exceeding 10% is 0.15g.

[0306] Dynamic finite element model analysis shows that under the design earthquake, the maximum shear strain is <0.1%.

[0307] Step 3.3, Mesoscale evaluation index system:

[0308] Rock mass integrity index (GSI): 55-75; groundwater level depth: 50-150m; terrain slope: 15°-25°; distance to the nearest road: 2-10km;

[0309] Step 4: Detailed analysis of site stability; selection of two optimal safety islands for detailed analysis.

[0310] Step 4.1: Detailed engineering geological survey; using Trimble TDC150 system, the accuracy is better than 1 meter.

[0311] Leica RTC360 3D laser scanning, point cloud density > 1000 points / m 2 .

[0312] Step 4.2: High-precision geophysical detection:

[0313] High-resolution 3D seismic exploration: detection depth 800m, resolution 2m.

[0314] Resistivity tomography: detection depth 200m, resolution 1m.

[0315] Geological radar detection: detection depth 30m, resolution 0.1m.

[0316] Step 4.3: Site rock mass quality assessment:

[0317] Improved GSI system score: GSI = 65 + 5-2 = 68;

[0318] The maximum principal stress measured by hydraulic fracturing method is 22MPa, with the direction NE35°.

[0319] Step 4.4: Micro-scale evaluation index system:

[0320] Rock deformation modulus: 25-35GPa;

[0321] Initial ground stress state: K0 = 1.2-1.5;

[0322] Permeability coefficient: 1×10 -7 -5×10 -6 m / s;

[0323] Step 5: Comprehensive integration of multi-scale evaluation results;

[0324] Step 5.1: Multi-scale data fusion: Use the improved DS evidence theory to fuse the evaluation results of different scales, with the conflict factor < 0.1.

[0325] Step 5.2, scale conversion: Apply the fractal scale conversion model, and the deviation between the indoor test strength and the field test results is <15%.

[0326] Step 5.3: Comprehensive evaluation: Finally determine the best site area, with an area of ​​about 2km 2 .

[0327] In summary, the specific process of finally determining the most optimal address area is as follows:

[0328] (1) Preliminarily screen out 10-20 candidate regions at the regional scale.

[0329] (2) Meso- and micro-analysis further narrows the scope and screens out 3-5 high-quality candidate areas (i.e., safe islands).

[0330] (3) Through multi-objective optimization and voting decision-making, the final optimal region is determined.

[0331] Step 6: Verification of evaluation results;

[0332] Step 6.1, cross validation: RMSE = 0.12, R 2 =0.85.

[0333] Step 6.2: Field verification:

[0334] Three boreholes were drilled with a depth of 500-800m and a core recovery rate of >98%.

[0335] The deviation between the field test results and the evaluation model prediction was <10%. The verification results showed that the optimal site selection area determined in step 5.3 met the actual site requirements.

[0336] Step 6.3, historical case comparison:

[0337] Compared with three existing CAES projects of the same region and specifications, the site selection accuracy of this embodiment using the method of the present invention is improved by about 75%, and the engineering risk is reduced by about 40%. It can be seen that this embodiment uses the method of the present invention to comprehensively consider multi-scale geological factors, and can effectively perform data fusion and scale conversion, thereby improving the accuracy of the CAES project and reducing engineering risks.

[0338] Example 2: The present invention is applied to the site selection of a CAES project in Gansu Province for a compressed air energy storage power station underground cavern. The specific method is as follows:

[0339] Step 1: Establish a multi-scale evaluation system; Regional scale: 100,000km 2 (covering the study area); mesoscale: 1,000km 2 ; Microscopic scale: 10km 2 ;

[0340] Step 2: Regional structural stability assessment;

[0341] Step 2.1: Regional structural pattern analysis:

[0342] Using Sentinel-2 multispectral images and ALOS PALSAR-2 radar data, the improved DeepLab v3+ model was used to extract linear structures, with a model accuracy of 0.91. The Qilian Mountain Fold Belt and the Altyn Fault Belt were particularly identified.

[0343] Step 2.2, active structure identification:

[0344] Combining InSAR and GPS data, three major active faults were identified, including the Shule River Fault, with an annual slip rate of 0.5-3 mm / year.

[0345] Step 2.3: Regional crustal movement characteristics and stress field assessment:

[0346] Analysis of 10 years of observation data from 12 continuous GPS stations showed that the horizontal movement rate was 1-4 mm / year.

[0347] A three-dimensional stress field model was constructed using 60 focal mechanism solutions and oilfield drilling stress measurement data. The maximum principal stress direction is NE20°±10°.

[0348] Step 2.4: Regional scale evaluation index system:

[0349] Active fault density: 0.03-0.15km / km 2 ;

[0350] Seismicity: Annual average earthquake frequency (M≥3.0) 0.3-1.2 times / 100km 2 , the largest historical earthquake was M7.1 (1932 Changma earthquake);

[0351] Crustal deformation rate: horizontal strain rate (3-10)×10 -9 / Year;

[0352] Tectonic stress intensity: maximum principal stress value 20-35MPa;

[0353] Terrain relief: 300-1500m;

[0354] Step 3: Identification and analysis of safety islands;

[0355] Step 3.1: Identify four potential safe islands (i.e. further narrow the scope through meso- and micro-analysis and screen out four high-quality candidate areas) with an area of ​​60-250km 2 . It is mainly distributed in the alluvial fan area in front of the mountain south of the project area.

[0356] Step 3.2, long-term stability assessment:

[0357] The improved probabilistic earthquake hazard analysis method was adopted, taking into account the fact that strong earthquakes rarely occur in arid areas. The results showed that the peak acceleration of an earthquake with a probability of exceeding 10% in 50 years is 0.20g.

[0358] A three-dimensional dynamic finite element model was established using FLAC3D software to evaluate the stability of the safety island under the design earthquake.

[0359] Step 3.3, Mesoscale evaluation index system:

[0360] The indicators of weathering crust thickness and annual groundwater level variation were added to adapt to the geological characteristics of arid areas.

[0361] Step 4: Detailed analysis of site stability;

[0362] Step 4.1: Detailed engineering geological survey:

[0363] The UAV laser radar (LiDAR) technology is used to obtain a digital surface model (DSM) with centimeter-level accuracy. The focus is on the unique landforms in arid areas, such as the Gobi Desert and Yadan landforms.

[0364] Step 4.2: High-precision geophysical detection:

[0365] High-resolution 3D seismic exploration: detection depth 1000m, resolution 1.5m.

[0366] Electromagnetic sounding: detection depth 500m, resolution 2m.

[0367] Geological radar detection: detection depth 40m, resolution 0.05m.

[0368] Step 4.3: Site rock mass quality assessment:

[0369] Considering the influence of arid climate on rock weathering, an improved GSI system is used to evaluate rock quality.

[0370] GSI = 70 + 3-5 = 68 (taking into account the deduction for weathering and the bonus for dry environment);

[0371] The maximum principal stress measured by hydraulic fracturing method is 28MPa, with the direction NE25°.

[0372] Step 4.4: Micro-scale evaluation index system:

[0373] Rock deformation modulus: 30-40GPa

[0374] Initial ground stress state: K0 = 1.3-1.7

[0375] Permeability coefficient: 5×10 -8 -2×10 -6 m / s

[0376] The rock mass weathering resistance and thermophysical properties indicators have been added.

[0377] Step 5: Comprehensive integration of multi-scale evaluation results

[0378] Step 5.1, multi-scale data fusion: An improved deep belief network (DBN) is used to fuse multi-source heterogeneous data, with special consideration given to the sparsity and uncertainty of geological data in arid areas.

[0379] Step 5.2, scale conversion: Develop a scale conversion model based on the combination of fractal theory and porous media theory to adapt to the heterogeneity and anisotropy of rock masses in arid areas.

[0380] Step 5.3: Comprehensive evaluation:

[0381] The analytic hierarchy process (AHP) combined with the entropy weight method was used to comprehensively consider geological stability, engineering feasibility, economy and environmental impact. Finally, an area of ​​about 3km2 was selected in the south of a city. 2 The optimal address area (that is, this embodiment determines the final optimal area through multi-objective optimization and voting decision).

[0382] Step 6: Verification of evaluation results;

[0383] Step 6.1, cross validation:

[0384] Using 10-fold cross validation, the root mean square error (RMSE) of the model was 0.15 and the coefficient of determination (R 2 ) is 0.87.

[0385] Step 6.2: Field verification:

[0386] Five holes were drilled with a depth of 100-300m and a core recovery rate of >95%.

[0387] On-site pumping tests and gas injection tests were carried out to verify the hydraulic properties and air tightness of the rock mass; the verification results showed that the optimal site area determined in step 5.3 met the actual requirements of the site.

[0388] Step 6.3: Historical case comparison:

[0389] Compared with gas storage projects in the same region, this embodiment adopts the method of the present invention to show obvious advantages in site selection efficiency and long-term stability prediction: the site selection period is shortened by about 40%; the accuracy of geological disaster risk assessment is improved by about 60%; the project cost is reduced by about 25%. It can be seen that this embodiment adopts the method of the present invention to comprehensively consider multi-scale geological factors, and can effectively perform data fusion and scale conversion, thereby improving the site selection efficiency and accuracy of the CAES project and reducing engineering risks and costs.

[0390] Through the case, the applicability and superiority of the method of the present invention in the complex geological environment of the arid region of Northwest China are fully demonstrated. The method can effectively deal with the unique geological conditions of arid regions, such as strong weathering and groundwater scarcity, and provide scientific and reliable technical support for the site selection of CAES projects. At the same time, this case also reflects the flexibility of the method under different geological conditions, and can adjust the evaluation indicators and technical routes according to the specific situation, so as to obtain the optimal site selection plan.

[0391] This example demonstrates the applicability and flexibility of this method under complex geological conditions in the arid region of northwest my country.

[0392] The multi-scale structural stability evaluation method of the present invention is not only applicable to the site selection of underground caverns of compressed air energy storage power stations, but can also be extended to other types of large-scale underground engineering projects, such as underground oil storage depots, nuclear waste disposal depots, and underground laboratories. By adjusting the evaluation index system and weight distribution, this method can meet the special needs of different types of underground projects and provide scientific and reliable decision-making support for the site selection of underground projects under complex geological conditions.

[0393] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

[0394] Other parts not described belong to the prior art.

Claims

1. A method for selecting underground caverns for compressed air energy storage based on multi-scale stability evaluation, characterized in that: The following steps are included: Step 1: Establish a multi-scale evaluation system, including regional scale, mesoscale and microscale; The scale range of the regional scale is 10 4 -10 5 km 2 ; Regional-scale content includes regional tectonic pattern analysis, active tectonic identification, crustal movement characteristics assessment, and regional stress field analysis; The scale range of the mesoscale is 10 2 -10 3 km 2 ; The mesoscale content includes safety island identification, long-term stability assessment, and earthquake resistance analysis; The microscopic scale ranges from 1-10 km 2 ; Micro-scale content includes detailed engineering site mapping, high-precision geophysical detection, site rock mass quality assessment and ground stress state analysis; Step 2: Conduct regional structural stability assessment; Step 3: Identify and analyze safe islands; Step 4: Conduct detailed site stability analysis; Step 5: Comprehensive integration of multi-scale evaluation results; Step 6: Verify the evaluation results; The verification of the evaluation results at least includes cross-validation, field verification and historical case comparison.

2. The method for selecting a site for an underground cavern for compressed air energy storage based on multi-scale stability evaluation according to claim 1 is characterized in that: In step 2, the regional structural stability assessment is carried out, including the following steps: Step 2.1: Regional structural pattern analysis; A deep learning algorithm is used to extract linear structural features from remote sensing images, and a regional structural pattern model is constructed by combining geological maps and digital elevation models; Step 2.2: Activity structure identification; Combine InSAR deformation monitoring, seismicity analysis and geological surveys to identify active faults; Step 2.3: Assessment of regional crustal movement characteristics and stress field; Using GPS observation data and earthquake focal mechanism solutions, combined with finite element numerical simulation, a regional stress field model is constructed; Step 2.4: Establish a regional scale evaluation index system and determine the weights; The evaluation indicators include active fault density, seismicity, crustal deformation rate, tectonic stress intensity and terrain relief, and the weights are determined using an improved hierarchical analysis method.

3. The method for selecting a site for an underground cavern for compressed air energy storage based on multi-scale stability evaluation according to claim 1 or 2, characterized in that: In step three, safety island identification and analysis are performed, including the following steps: Step 3.1: Identify relatively stable structural blocks; based on the principles of structural integrity, lithological consistency and deformation degree similarity, combined with multi-source geophysical data, identify safe islands; Step 3.2: Assess the long-term stability and earthquake resistance of the safety island; Probabilistic seismic hazard analysis and dynamic finite element analysis are used to assess the long-term stability of the safety island; Step 3.3: Establish a mesoscale evaluation index system and determine the weights; add rock mass integrity and groundwater condition indicators on the basis of regional scale indicators, and use fuzzy hierarchical analysis method to determine the weights.

4. The method for selecting a site for an underground cavern for compressed air energy storage based on multi-scale stability evaluation according to claim 3 is characterized in that: In step 4, a detailed site stability analysis is conducted, including the following steps: Step 4.1: Detailed engineering geological mapping; use digital field systems and 3D laser scanning technology to obtain high-precision geological information; Step 4.2: High-precision geophysical survey; combining high-resolution 3D seismic, electrical prospecting and geological radar survey to build a detailed geological structure model; Step 4.3: Site rock mass quality assessment: assess the site rock mass quality, fracture development and ground stress state; An improved GSI system is used to assess rock mass quality and a variety of geostress measurement methods are combined to determine the initial geostress state; Step 4.4: Establish a micro-scale evaluation index system and determine the weights; add engineering scale indicators and use the combined weighting method to determine the weights.

5. The method for selecting a site for an underground cavern for compressed air energy storage based on multi-scale stability evaluation according to claim 4 is characterized in that: In step five, the comprehensive integration of multi-scale evaluation results is carried out, including the following steps: Step 5.1: Multi-scale data fusion; Multi-scale data fusion includes data standardization processing, multi-source data fusion algorithm and data quality control; The multi-scale data fusion adopts the improved Dempster-Shafer evidence theory and wavelet transform fusion algorithm; Step 5.2: scale conversion; the scale conversion adopts a scale conversion model based on fractal theory; Step 5.3: Comprehensive evaluation; use the improved analytic hierarchy process to conduct comprehensive evaluation: introduce fuzzy theory to deal with uncertainty and improve the reliability of evaluation results; A multi-objective decision analysis method is used to comprehensively consider engineering geological conditions, construction difficulty and long-term stability to optimize and compare site selection options.

6. The method for selecting a site for an underground cavern for compressed air energy storage based on multi-scale stability evaluation according to claim 5 is characterized in that: In step 5.1, the multi-scale data fusion includes the following steps: Step 5.1.1, data standardization: Unit unification: Establish a unit conversion matrix to take into account the differences between different measurement standards; Dimension normalization: Z-score normalization method is used; Coordinate system conversion: unify the spatial reference systems of different data sources; Step 5.1.2, multi-source data fusion algorithm: For the same type of data, the weighted average method is used; For uncertain data, Dempster-Shafer evidence theory is used; For multi-resolution data, wavelet transform fusion is used; Step 5.1.3, data quality control: Outlier detection and processing: Use the 3σ criterion to identify outliers; Data integrity check: Multiple imputation was used to handle missing values; Consistency check: Establish data priority rules to resolve data conflicts.

7. The method for selecting a site for an underground cavern for compressed air energy storage based on multi-scale stability evaluation according to claim 6 is characterized in that: In step 5.2, the scale conversion algorithm specifically includes: 1) Upscaling method: Simple statistical method: select arithmetic mean, geometric mean or weighted mean according to the characteristics of the data; Geostatistical methods: Block Kriging was used to consider spatial correlation; 2) Downscaling method: Dynamic downscaling: using high-resolution numerical models to take into account the influence of local topography and geological factors; Statistical downscaling: establishing statistical relationships between large and small scale predictors; 3) Scale effect treatment of rock mechanics parameters: Intensity parameters: the modified Hoek-Brown criterion is adopted, and the GSI value and disturbance factor D are introduced; Deformation modulus: using the modified Hoek-Diederichs empirical formula; Permeability: Using fractal theory, the characteristic length and fractal dimension of the fracture network are considered.

8. The method for selecting a site for an underground cavern for compressed air energy storage based on multi-scale stability evaluation according to claim 7 is characterized in that: In step 3.1, the identification criteria of the safety island include: (1) Internal fault development degree: fault density is less than 0.5 km / km 2 ; (2) Boundary fault activity: The average slip rate over the past 100,000 years is less than 0.1 mm / year; (3) Earthquake activity level: Annual average earthquake frequency (M≥3.0) is less than 1 / 100km 2 ; (4) Terrain undulation: less than 200m / 10km 2 .

9. The method for selecting a site for an underground cavern for compressed air energy storage based on multi-scale stability evaluation according to claim 8, characterized in that: In step 3.1, multi-source data fusion and machine learning methods are used to identify safety islands. The specific steps are as follows: Step 3.1.1: Data preparation; Collect the following data: high-resolution remote sensing images, digital elevation models, geological maps, earthquake catalogs, GPS observation data, and InSAR deformation monitoring data; Step 3.1.2: Feature extraction; (1) Extracting linear structural features from remote sensing images and DEM: Use the improved Canny edge detection algorithm or Hough transform to identify linear structural features in images; (2) Calculate terrain relief; Using the moving window method, the maximum elevation difference of each grid is calculated; (3) Calculate the density of active faults; count the total length of faults within the sliding window; (4) Calculate seismic activity; calculate the frequency and maximum magnitude of historical earthquakes; Step 3.1.3: Construct feature vectors; for each 10km×10km grid, construct feature vectors: X=[x1,x2,x3,x4,x5] Among them: x1: linear structure density; x2: terrain relief; x3: active fault density; x4: seismic activity index; x5: crustal deformation rate; Step 3.1.4: Machine learning classification; (1) Use the support vector machine algorithm for binary classification, including: positive class: potential safe island; negative class: unstable area; (2) SVM decision function; f(X)=sign(∑(α i y i K(X i ,X))+b); Where K is the radial basis function kernel; K(X i ,X)=exp(-γ||X i X|| 2 ) Step 3.1.5: Optimize the safety island boundary; use the active contour model to refine the safety island boundary, and the energy function is: E∫[α|C'(s)|2+β|C”(s)| 2 +Ext(C(s))]ds Where C(s) is the contour curve, Eext is the external energy term; Step 3.16: Multi-scale verification; repeat steps 3.12-3.15 at different spatial scales; The final safe island range is determined using majority voting.

Images