Bayesian Hierarchical Probability Prediction Method for Small-Strain Shear Modulus of Offshore Wind Farms Based on CPT

Abstract

The present invention discloses a Bayesian hierarchical probability prediction method for the small-strain shear modulus Gmax of soil in an offshore wind farm based on CPT, including: obtaining Gmax data at different CPT positions in the offshore wind farm to construct training samples; constructing a Bayesian regression model, including constructing a fully mixed model and a partially mixed model; performing Markov chain Monte Carlo (MCMC) simulation on the constructed Bayesian regression model to obtain the posterior distribution of model parameters, and further evaluating the fitting effect of the model to obtain a target model that meets the requirements of the prediction effect; according to the target model, dividing multiple scenarios to predict the Gmax profile, and further determining whether the CPT position and soil unit are in the calibration dataset to complete the probability prediction. The embodiments of the present invention can more effectively utilize CPT data, can generate a more accurate representative profile of soil parameters, improve the accuracy and reliability in designing wind turbine foundations, and can be widely applied to the field of offshore wind power technology.

Description

Technical Field

[0001] The present invention relates to the technical field of offshore wind power, and particularly relates to a Bayesian hierarchical probability prediction method for the small-strain shear modulus of soil masses in an offshore wind farm based on CPT. Background Art

[0002] With the continuous expansion of the scale of offshore wind farms, it has become crucial to accurately determine the representative or characteristic profile of soil parameters for the design of wind turbine foundations. Modern geotechnical limit state design (LSD) standards require specifying representative values of soil parameters at a certain confidence level or exceedance probability to reflect the uncertainty in the characterization of soil parameters. In the design of offshore wind farms, a combination of static cone penetration test (CPT) and in-situ testing of soil samples obtained by drilling is usually adopted to determine the spatial distribution of soil parameters. CPT can provide continuous, rapid, and indirect in-situ soil parameter measurements, while soil samples obtained by drilling are used to obtain direct measurements of soil parameters for calibrating empirical models based on CPT, so as to obtain the probability characterization profiles of soil parameters based on their CPT measurement curves at different wind turbine positions, such as the five representative ABCDE profiles reflecting different probability levels commonly used in the field of marine geotechnical investigation. Among them, profiles A and E represent the low and high estimates of G max (the boundaries of the 90% prediction interval), and profiles C, B, and D represent the best estimate, low estimate, and high estimate of the mean value of G max (the boundaries of the 90% confidence interval), respectively. The small-strain shear modulus G max is a key parameter for the design of offshore wind turbine foundations, and is used to estimate the structural characteristic frequency and foundation stiffness (under fatigue and serviceability limit states). In the site investigation of offshore wind farms, seismic static cone penetration test (SCPT) is widely used to obtain G max data. This method derives G max by measuring the in-situ shear wave velocity, avoiding the disturbance of soil layer sampling in laboratory tests and being able to obtain a relatively large amount of data.

[0003] Although the traditional method of obtaining soil samples by drilling can directly obtain soil samples for laboratory tests, it has many limitations. Drilling and sampling in the marine environment are costly and time-consuming. For large-scale offshore wind farm projects, soil samples can only be obtained for in-situ tests at specific depths at limited wind turbine positions, and the data shows significant sparsity, which cannot meet the accurate characterization of the spatial distribution of soil parameters at all wind turbine positions. Moreover, soil samples are easily disturbed during the process of drilling and sampling in the seabed and transportation, which will cause a certain deviation between the test results and the true characteristics of in-situ soil. As an in-situ testing technology, CPT has been widely used in the site investigation of offshore wind farms. It can continuously and rapidly obtain the cone tip resistance (q t), lateral friction (fs), and pore water pressure (u), thereby indirectly reflecting the engineering properties of the soil. CPT testing has the advantages of being relatively non-invasive, low-cost, and highly efficient. It can provide a large amount of test data in a short period of time, making it possible to fully understand the site soil parameters.

[0004] Currently, based on CPT data, G max The conventional method is the fully mixed method. This method assumes that the empirical relationship based on CPT is constant throughout the wind farm, ignoring the significant differences that may exist between different CPT locations and between different soil elements. The fully mixed model uses all the data from the entire site to calibrate a single CPT-G max The empirical model does not take into account the hierarchical structure and spatial variability of the data. This leads to the possibility of misleading probability predictions and unreasonable G in practical applications. max Representative profiles can affect the accuracy and safety of wind turbine foundation design. For example, when selecting a representative profile of soil property parameters for design (such as Profile B, which provides a cautiously low estimate of its mean), the fully mixed model may be overly optimistic due to ignoring the variation between data groups. This can lead to a significant deviation between the selected estimates and the actual situation, thus affecting the design reliability of the wind turbine foundation. Summary of the invention

[0005] The main purpose of the embodiment of the present invention is to propose a Bayesian layered probability prediction method for the small strain shear modulus of offshore wind farms based on CPT, which can more effectively utilize CPT data, generate more accurate representative profiles of soil layers, and improve the accuracy and reliability of wind turbine foundation design.

[0006] To achieve the above objectives, an embodiment of the present invention provides a Bayesian hierarchical probability prediction method for small strain shear modulus of offshore wind farms based on CPT, comprising the following steps:

[0007] Obtain soil small strain shear modulus data at different CPT locations in offshore wind farms to construct training samples;

[0008] Constructing a Bayesian regression model based on the training samples, including constructing a complete mixture model and a partial mixture model;

[0009] Perform Markov Chain Monte Carlo (MCMC) simulation on the constructed Bayesian regression model to obtain the posterior distribution of the model parameters, and then evaluate the model fitting effect to obtain the target model that meets the prediction effect requirements;

[0010] According to the target model, multiple scenarios are divided into predictions G max The profile is then used to determine whether the CPT position and soil unit are in the calibration data set, completing the probabilistic prediction.

[0011] In some embodiments, obtaining the small-strain shear modulus data of the soil mass at different CPT positions in an offshore wind farm and constructing a training sample includes the following steps:

[0012] Collect geological, geophysical, and geotechnical data of the offshore wind farm from a specified data source, including the cone tip resistance, pore water pressure data obtained from static cone penetration tests at multiple CPT positions, and the in-situ G max value measured by seismic static cone penetration test SCPT or other methods;

[0013] Calculate the effective vertical stress according to the site geological conditions and the measurement depth, specifically: calculate the effective vertical stress corresponding to each data point by multiplying the depth below the seabed by the average effective stress gradient;

[0014] Convert the G max value from the original measurement unit to MPa, classify it according to the CPT position and soil unit, construct a cross-classified data structure, count the number of samples in each classification group, form a data summary, and construct the training sample.

[0015] In some embodiments, constructing a fully mixed model in the step of constructing a Bayesian regression model according to the training sample includes the following steps:

[0016] Based on the empirical formula Construct a fully mixed regression model, where q t represents the total cone tip resistance of the static cone penetration test (CPT) corrected by pore pressure; p a represents the atmospheric pressure; σ′ v represents the effective vertical stress; α0, α1, α2 represent the three unknown parameters of the model;

[0017] Set a prior distribution for the model parameters, and use a weak-information prior to reflect the lack of specific prior knowledge;

[0018] Characterize the preliminary estimate of the possible value range of the parameters according to the prior distribution.

[0019] In some embodiments, constructing a partially mixed model in the step of constructing a Bayesian regression model according to the training sample includes the following steps:

[0020] Considering the variation between the CPT position and the soil unit, construct a cross-classified hierarchical model based on the set exchangeable intercept.

[0021] In some embodiments, performing Markov chain Monte Carlo MCMC simulation on the constructed Bayesian regression model to obtain the posterior distribution of the model parameters includes the following steps:

[0022] The Bayesian regression model constructed is fitted using the probabilistic programming language Stan in combination with the Markov chain Monte Carlo simulation method;

[0023] Multiple Markov chains are run in parallel, and each chain performs the corresponding number of simulated samplings to obtain the posterior distribution of the model parameters;

[0024] Among them, during the simulation process, the convergence of the posterior distribution of the parameters is judged by checking the statistical indicators within and between the chains. If the posterior distribution does not converge, the simulation parameters are adjusted and the simulation is run again until a stable posterior distribution is obtained; among them, the simulation parameters include increasing the number of iterations and adjusting the prior distribution.

[0025] In some embodiments, evaluating the fitting effect of the model to obtain a target model that meets the prediction effect requirements includes the following steps:

[0026] Analyze the posterior distribution of the model parameters, calculate statistical indicators such as the posterior mean, standard deviation, and 95% confidence interval, and evaluate the fitting effect of the model on the data;

[0027] Calculate the residual standard deviation τ ε , and judge the goodness of fit of the model by comparing the τ ε values of the fully mixed model and the cross-classified hierarchical model;

[0028] Calculate the intraclass correlation coefficient, and the formula is [[ID=2 / 4]] Measure the degree of similarity between data groups through the intraclass correlation coefficient; among them, ICC represents the intraclass correlation coefficient; represents the variance of the local group (such as position j) (representing G max the variance between different CPT positions); represents the variance of the superior group (such as unit k) (G max the variance between different soil body units); represents G max the variance of the residuals;

[0029] Use the leave-one-out cross-validation method to compare the expected log predictive densities of different models and evaluate the prediction accuracy of the models;

[0030] Select the target model with the best prediction performance according to the values of the predictive density, its differences, and the standard errors.

[0031] In some embodiments, according to the target model, divide multiple scenarios to predict the G max profile, and then judge whether the CPT position and the soil body unit are in the calibration dataset to complete the probability prediction, including the following steps:

[0032] Predict G according to the CPT position and whether the soil element is in the calibration dataset, in the following scenarios max Profile:

[0033] Existing position and existing element with observed data: Use the Estimated group-specific β0 in the calibration data and substitute it into the model formula (3a) to predict G max Profile; where β 0(jk) Represents the intercept of CPT position j and soil element k; Represents the overall mean of the intercept β0; Represents the local deviation of position j; Represents the local deviation of element k; β0 represents the overall basic intercept;

[0034] Existing position and existing element without observed data: Use the estimated β 0(jk) Calculate β0 for prediction;

[0035] Existing position and new element: Sample β0 from And then substitute it into the model formula to predict G max Profile, where And Are the corresponding hyperparameters and parameter estimates;

[0036] New position and existing element: Sample β0 from And then substitute it into the model formula to predict G max Profile, where And Are the corresponding hyperparameters and parameter estimates;

[0037] New position and new element: Sample β0 from And then substitute it into the model formula to predict G max Profile, where And Are the corresponding hyperparameters and parameter estimates;

[0038] Perform leave-one-group (LOGO) analysis. Each time, leave the data of one soil element at one CPT position as the validation set, fit the model with the remaining data and perform prior prediction, and then add the validation data for posterior prediction; Compare the prior and posterior predicted G max Profile changes, analyze the update of the model after observing new data, and analyze the improvement of the target model after obtaining new data by comparing these profile changes.

[0039] Another aspect of the embodiments of the present invention also provides a Bayesian hierarchical probability prediction system for the small-strain shear modulus of an offshore wind farm based on CPT, including:

[0040] The first module is used to obtain the small-strain shear modulus data of the soil at different CPT positions in the offshore wind farm and construct a training sample;

[0041] The second module is used to construct a Bayesian regression model according to the training sample, including constructing a fully mixed model and a partially mixed model;

[0042] The third module is used to perform Markov chain Monte Carlo (MCMC) simulation on the constructed Bayesian regression model to obtain the posterior distribution of the model parameters, and then evaluate the fitting effect of the model to obtain a target model that meets the prediction effect requirements;

[0043] The fourth module is used to divide multiple scenarios to predict the G max profile, and then determine whether the CPT position and the soil unit are in the calibration dataset to complete the probability prediction.

[0044] To achieve the above object, on the other hand, an embodiment of the present invention provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the foregoing method is implemented.

[0045] To achieve the above object, on the other hand, an embodiment of the present invention provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the foregoing method is implemented.

[0046] An embodiment of the present invention also discloses a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device can read the computer instructions from the computer-readable storage medium, and when the processor executes the computer instructions, the computer device executes the foregoing method.

[0047] The embodiments of the present invention at least include the following beneficial effects: The present invention provides a Bayesian hierarchical probability prediction method for the small-strain shear modulus of an offshore wind farm based on CPT. This solution obtains the small-strain shear modulus data of the soil at different CPT positions in the offshore wind farm and constructs a training sample; constructs a Bayesian regression model according to the training sample, including constructing a fully mixed model and a partially mixed model; performs Markov chain Monte Carlo (MCMC) simulation on the constructed Bayesian regression model to obtain the posterior distribution of the model parameters, and then evaluates the fitting effect of the model to obtain a target model that meets the prediction effect requirements; divides multiple scenarios to predict the G maxCross-section, and then determine whether the CPT position and soil unit are in the calibration dataset to complete the probability prediction. The embodiments of the present invention can utilize CPT data more effectively, generate a more accurate representative soil layer profile, and improve the accuracy and reliability in designing wind turbine foundations. Description of the Drawings

[0048] Figure 1 is a schematic diagram of an implementation environment provided by an embodiment of the present invention;

[0049] Figure 2 is a flowchart of the overall steps provided by an embodiment of the present invention;

[0050] Figure 3 is an overview map of the TNW wind farm with CPT borehole positions provided by an embodiment of the present invention;

[0051] Figure 4 is a directed acyclic graph of the Bayesian regression model provided by an embodiment of the present invention;

[0052] Figure 5 is a posterior estimate map of the CPT position and soil unit effect provided by an embodiment of the present invention;

[0053] Figure 6 is a comparison map of the estimated value of β0 provided by an embodiment of the present invention;

[0054] Figure 7 is a posterior standardized residual map provided by an embodiment of the present invention;

[0055] Figure 8 is a LOGO analysis result map provided by an embodiment of the present invention;

[0056] Figure 9 is a final prediction example of the entire wind farm provided by an embodiment of the present invention. Detailed Embodiments

[0057] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the embodiments of the present invention. They are only examples of devices and methods consistent with some aspects of the embodiments of the present invention as detailed in the appended claims.

[0058] It will be understood that the terms "first", "second", etc. used in the present invention may be used herein to describe various concepts, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of the present invention, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, as used herein, the words "if", "when" may be interpreted as "when...", "when...", or "in response to determining".

[0059] The terms "at least one", "a plurality of", "each", "any one", etc. used in the present invention, at least one includes one, two or more than two, a plurality of includes two or more than two, each refers to each of the corresponding plurality, and any one refers to any one of the plurality.

[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs. The terms used herein are only for the purpose of describing the embodiments of the present invention and are not intended to limit the present invention.

[0061] The Bayesian hierarchical probability prediction method for the small-strain shear modulus of an offshore wind farm based on CPT provided by the embodiments of the present invention relates to the technical field of offshore wind power. The Bayesian hierarchical probability prediction method for the small-strain shear modulus of an offshore wind farm based on CPT provided by the embodiments of the present invention can be applied to a terminal, can also be applied to a server, or can also be software running on a terminal or a server. In some embodiments, the terminal may be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, a vehicle-mounted terminal, etc., but is not limited thereto; the server side may be configured as an independent physical server, may also be configured as a server cluster or a distributed system composed of multiple physical servers, or may also be configured as a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server may also be a node server in a blockchain network; the software may be an application for implementing the Bayesian hierarchical probability prediction method for the small-strain shear modulus of an offshore wind farm based on CPT, etc., but is not limited to the above forms.

[0062] The present invention can be used in numerous general-purpose or special-purpose computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and so on. The present invention can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present invention can also be practiced in a distributed computing environment where tasks are executed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0063] As Figure 1 shown, it is a schematic diagram of an implementation environment provided by an embodiment of the present invention. Referring to Figure 1 , this implementation environment includes at least one terminal 102 and a server 101. The terminal 102 and the server 101 can be network-connected in a wireless or wired manner to complete data transmission and exchange.

[0064] The server 101 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. It can also be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms.

[0065] In addition, the server 101 can also be a node server in a blockchain network. Among them, blockchain is a new application mode of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, and encryption algorithms.

[0066] The terminal 102 can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc. Among them, the terminal 102 can also be an in-vehicle terminal of various device types exemplified above, but is not limited thereto. The terminal 102 and the server 101 can be directly or indirectly connected through wired or wireless communication methods, and the embodiments of the present invention do not limit this here.

[0067] Exemplarily based on Figure 1In the implementation environment shown, an embodiment of the present invention provides a Bayesian hierarchical probability prediction method for the small-strain shear modulus of an offshore wind farm based on CPT. Taking the application of this Bayesian hierarchical probability prediction method for the small-strain shear modulus of an offshore wind farm based on CPT to server 101 as an example for illustration, it can be understood that this method can also be applied to terminal 102.

[0068] Referring to Figure 2 , Figure 2 FIG. is a flowchart of a Bayesian hierarchical probability prediction method for the small-strain shear modulus of an offshore wind farm based on CPT applied to a server provided by an embodiment of the present invention. The execution subject of this method can be any of the aforementioned computer devices (including a server or a terminal). Referring to Figure 2 , this method may include the following steps:

[0069] Obtain the small-strain shear modulus data of the soil mass at different CPT positions in the offshore wind farm and construct a training sample;

[0070] Construct a Bayesian regression model according to the training sample, including constructing a full mixture model and a partial mixture model;

[0071] Perform Markov chain Monte Carlo (MCMC) simulation on the constructed Bayesian regression model to obtain the posterior distribution of the model parameters, and then evaluate the fitting effect of the model to obtain a target model that meets the prediction effect requirements;

[0072] According to the target model, divide multiple scenarios to predict the G max profile, and then determine whether the CPT position and the soil unit are in the calibration dataset to complete the probability prediction.

[0073] In some embodiments, the step of obtaining the small-strain shear modulus data of the soil mass at different CPT positions in the offshore wind farm and constructing a training sample includes the following steps:

[0074] Collect geological, geophysical, and geotechnical data of the offshore wind farm from a specified data source, including the cone tip resistance, pore water pressure data obtained from static cone penetration tests at multiple CPT positions, and the in-situ G max value measured by seismic cone penetration test (SCPT) or other methods;

[0075] Calculate the effective vertical stress according to the site geological conditions and the measurement depth. Specifically, the effective vertical stress corresponding to each data point is calculated by multiplying the depth below the seabed by the average effective stress gradient;

[0076] Take G maxThe values are converted from the original measurement units to MPa, classified according to the CPT positions and soil elements, a cross-classification data structure is constructed, the sample numbers of each classification group are counted to form a data summary, and the training samples are constructed accordingly.

[0077] In some embodiments, constructing a fully mixed model in the step of constructing a Bayesian regression model based on the training samples includes the following steps:

[0078] Based on the empirical formula Construct a fully mixed regression model, where q t represents the total cone tip resistance of the cone penetration test (CPT) after pore pressure correction; p a represents the atmospheric pressure; σ′ v represents the effective vertical stress; α0, α1, α2 represent three unknown parameters of the model;

[0079] Set a prior distribution for the model parameters, and use a weak-information prior to reflect the lack of specific prior knowledge;

[0080] Characterize the preliminary estimate of the possible value range of the parameters according to the prior distribution.

[0081] In some embodiments, constructing a partially mixed model in the step of constructing a Bayesian regression model based on the training samples includes the following steps:

[0082] Consider the variation between CPT positions and soil elements, and construct a cross-classification hierarchical model based on the set exchangeable intercept.

[0083] In some embodiments, performing a Markov chain Monte Carlo (MCMC) simulation on the constructed Bayesian regression model to obtain the posterior distribution of the model parameters includes the following steps:

[0084] Use the probabilistic programming language Stan combined with the Markov chain Monte Carlo simulation method to fit the constructed Bayesian regression model;

[0085] Run multiple Markov chains in parallel, and each chain performs the corresponding number of simulation samplings to obtain the posterior distribution of the model parameters;

[0086] Among them, during the simulation process, check the statistical indicators within and between the chains to determine whether the posterior distribution of the parameters converges. If the posterior distribution does not converge, adjust the simulation parameters and re-run the simulation until a stable posterior distribution is obtained; among them, the simulation parameters include increasing the number of iterations and adjusting the prior distribution.

[0087] In some embodiments, evaluating the fitting effect of the model to obtain a target model that meets the prediction effect requirements includes the following steps:

[0088] Analyze the posterior distribution of the model parameters, calculate statistical indicators such as the posterior mean, standard deviation, and 95% confidence interval, and evaluate the fitting effect of the model on the data;

[0089] Calculate the residual standard deviation τ ε , and judge the goodness of fit of the model by comparing the τ ε values of the fully mixed model and the cross-classified hierarchical model;

[0090] Calculate the intra-class correlation coefficient, and the formula is Measure the degree of similarity between data groups through the intra-class correlation coefficient; where, ICC represents the intra-class correlation coefficient; represents the variance of the local group (such as position j) (representing G max the variance between different CPT positions); represents the variance of the superior group (such as unit k) (G max the variance between different soil units); represents G max the variance of the residuals;

[0091] Adopt the leave-one-out cross-validation method to compare the expected log predictive density of different models, and evaluate the prediction accuracy of the models;

[0092] Select the target model with the best prediction performance according to the values of the predictive density, its differences, and the standard errors.

[0093] In some embodiments, according to the target model, divide multiple scenarios to predict the G max profile, and then judge whether the CPT position and the soil unit are in the calibration dataset to complete the probability prediction, including the following steps:

[0094] Predict the G max profile according to whether the CPT position and the soil unit are in the calibration dataset, in the following scenarios:

[0095] Existing position and existing unit with observed data: Use the estimated group-specific β0 calculated from the calibration data and substitute it into the model formula (3a) to predict the G max profile; where, β 0(jk) represents the intercept of the CPT position j and the soil unit k; represents the overall mean of the intercept β0; represents the local deviation of position j; represents the local deviation of unit k; β0 represents the overall basic intercept;

[0096] Existing position and existing unit but without observed data: Use the estimated β 0(jk) calculated from the calibration data to make predictions;

[0097] Existing location and new unit: Sample β0 from , then substitute it into the model formula to predict G max profile, where and are the corresponding hyperparameters and parameter estimates;

[0098] New location and existing unit: Sample β0 from , then substitute it into the model formula to predict G max profile, where and are the corresponding hyperparameters and parameter estimates;

[0099] New location and new unit: Sample β0 from , then substitute it into the model formula to predict G max profile, where and are the corresponding hyperparameters and parameter estimates;

[0100] Perform leave-one-LOGO analysis. Each time, leave the data of one soil unit at one CPT location as the validation set, fit the model with the remaining data and perform prior prediction, then add the validation data for posterior prediction; compare the G max profile changes, analyze the update of the model after observing new data, and analyze the improvement degree of the target model after obtaining new data by comparing the changes in these profiles.

[0101] Next, taking a specific application scenario as an example, the specific implementation process of the embodiment of the present invention will be described in detail:

[0102] Aiming at the problems existing in the prior art, the present invention relates to a method for predicting soil layer profiles for the design of offshore wind farm foundations, especially a high-precision prediction model for the small-strain shear modulus G max . With the expansion of the scale of offshore wind farms, it is usually dependent on CPT and its empirical statistical models to deduce soil layer profiles. However, traditional statistical modeling methods fail to make full use of the widely increasing amount of G max data, ignoring the differences between CPT locations and soil layers, which may lead to misleading probability predictions. The present invention proposes a Bayesian cross-classification hierarchical model that takes into account the differences between CPT locations and soil layers, so as to achieve "high-resolution" prediction of the G max profile at different locations within the wind farm. Practical case applications have verified the significant improvement of the model in terms of probability prediction and profile rationality, and have confirmed effective posterior update and uncertainty reduction after observing new data through leave-one analysis.

[0103] Specifically, a method for predicting the small-strain shear modulus G of soil based on CPT data max A construction method for a Bayesian cross-classification hierarchical model for high-resolution prediction, comprising the following steps:

[0104] 1. Obtain the small-strain shear modulus data of the soil at different CPT positions in the offshore wind farm:

[0105] Collect geological, geophysical, and geotechnical data of the offshore wind farm (such as the Ten noorden van de Waddeneilanden (TNW) wind farm in the Dutch North Sea) from the specified data source, including the cone tip resistance q obtained from the static cone penetration test at multiple CPT positions (at least 162 CPT positions) t , pore water pressure data, and the in-situ G max value measured by seismic static cone penetration test SCPT or other relevant methods.

[0106] Calculate the effective vertical stress σ′ according to the site geological conditions and the measurement depth v , specifically, the effective vertical stress corresponding to each data point is calculated by multiplying the depth below the seabed by the average effective stress gradient (such as the average effective stress gradient in the North Sea waters is 10 kPa / m).

[0107] Convert the G max value from the original measurement unit to MPa, and the conversion formula is G max =ρV s 2 (where ρ is the soil density, determined according to the soil type and actual measurement, and V s is the shear wave velocity), and classify according to the CPT position and soil unit, construct a cross-classification data structure, count the sample quantity of each classification group (the combination of CPT position and soil unit), form a data summary, and ensure the integrity and accuracy of the data for subsequent analysis and use.

[0108] 2. Construction of the Bayesian regression model:

[0109] 1) Construction of the fully mixed model

[0110] Based on the empirical formula Construct a fully mixed regression model, which is expressed on the natural logarithm scale as

[0111]

[0112] where i = 1,…,N represents the data point serial number, is the regression coefficient to be estimated, ε i is the random error term, and p aAtmospheric pressure (100 kPa) is used for normalization to avoid problems of unit inconsistency.

[0113] E represents the expected value of ln(G max(i) ); q t(i) represents the tip resistance of the i-th data point; β0 represents the intercept of the regression model; represents the regression coefficient of; represents the regression coefficient of.

[0114] A prior distribution is set for the model parameters. A weakly informative prior is used to reflect the lack of specific prior knowledge. Specifically: β0 ~ Normal(0, 5 2 )(2a)

[0115]

[0116] τ ε ~ Half - normal(0, 1 2 )(2d)

[0117] where τ ε represents the standard deviation of the error term in the model; these prior distributions reflect the initial estimate of the possible value range of the parameters in the absence of specific prior knowledge, and at the same time avoid the problems that may be brought by traditional vague / flat priors (such as posterior result bias when the data is scarce or the model is complex).

[0118] 2) Construction of the partially - mixed (cross - classified hierarchical) model:

[0119] Construct a cross - classified hierarchical model, considering the variation between CPT positions and soil elements. Assume an exchangeable intercept β0, and other parameters are the same among different CPT positions and soil elements. The model formula is

[0120]

[0121] where G max(ijk) represents the i - th (i = 0, …, N jk ) observation in the jk - th group cross - classified by location j and unit; β 0(jk) is the intercept corresponding to location j and unit k, and its overall distribution reflecting the exchangeability assumption is specified as a common normal prior distribution.

[0122]

[0123] where ε ijkrepresents the random error term for the \(i\)-th data point (position \(j\), soil element \(k\)); \(j = 1,\ldots,J\) represents the CPT positions and \(k = 1,\ldots,K\) represents the soil elements.

[0124] Set the prior distributions for the hyperparameters:

[0125]

[0126] where has the same prior distribution as the fully mixed model.

[0127] 3. Model fitting and analysis:

[0128] Use the probabilistic programming language Stan combined with the Markov chain Monte Carlo simulation method No-U-Turn Sampler to fit the constructed model. Run multiple Markov chains (at least three) in parallel, and each chain conducts a large number of simulation samplings (such as 5000 times) to obtain the posterior distribution of the model parameters. During the simulation, closely monitor the convergence of the chains, and judge whether the posterior distribution of the parameters converges by checking the statistical indicators within and between the chains (such as the Rhat value). If the Rhat value is close to 1 (usually within 1.05), it is considered that the model converges; otherwise, the simulation parameters need to be adjusted (such as increasing the number of iterations, adjusting the prior distribution, etc.), and the simulation is run again until a stable posterior distribution is obtained.

[0129] Analyze the posterior distribution of the model parameters, calculate statistical indicators such as the posterior mean, standard deviation, 95% confidence intervals (CIs), etc., to evaluate the fitting effect of the model on the data. Calculate the residual standard deviation \(\tau\) ε , and judge the goodness of fit of the model by comparing the \(\tau\) ε values of the fully mixed model and the cross-classified hierarchical model. A smaller \(\tau\) ε value usually indicates a better fit of the model to the data. At the same time, calculate the intraclass correlation coefficient (ICC), and the formula is to measure the degree of similarity between data groups, and further evaluate the rationality of the model assumptions. The closer the ICC value is to 0, the more similar the groups are, and the closer it is to 1, the greater the difference between the groups.

[0130] Use the leave-one-out cross-validation (LOO-CV) method to compare the expected log predictive density (elpd) of different models, evaluate the prediction accuracy of the models. The larger the elpd value, the stronger the out-of-sample prediction ability of the model. According to the elpd value, its difference (\(\Delta\) elpd) and standard error (SE), select the model with the best prediction performance. When the absolute value of \(\Delta\) elpd is greater than several times (such as more than 3 times) its standard error, it can be considered that there are significant differences between the models; otherwise, the differences are not significant.

[0131] 4. G maxProfile Prediction and Update:

[0132] Predict G for the following scenarios based on the CPT location and whether the soil element is in the calibration dataset max Profile:

[0133] Existing location and existing element with observed data: directly use the Group-specific β0 estimated in the calibration data and substitute it into model formula (3a) to predict G max Profile.

[0134] Existing location and existing element but without observed data: also use the estimated β 0(jk) Calculate β0 for prediction because the location and element effects have been estimated in the calibration data.

[0135] Existing location and new element: sample β0 from and then substitute it into the model formula to predict G max Profile, where and are the corresponding hyperparameters and parameter estimates.

[0136] New location and existing element: sample β0 from and then substitute it into the model formula to predict G max Profile, where and are the corresponding hyperparameters and parameter estimates.

[0137] New location and new element: sample β0 from and then substitute it into the model formula to predict G max Profile, where and are the corresponding hyperparameters and parameter estimates.

[0138] Perform leave-one-group (LOGO) analysis. Each time, leave the data of one soil element at one CPT location as the validation set, fit the model with the remaining data and make prior predictions, and then add the validation data for posterior predictions. Compare the changes in the representative profiles (i.e., Profiles A - E) of the prior and posterior predictions of G max to analyze the update of the model after observing new data, so as to evaluate the performance of the model in fusing new data and the improvement effect of the prediction of the representative profile.

[0139] The Bayesian cross-classification stratified model adopts leave-one-group cross-validation analysis to calibrate model parameters in order to provide reliable prior predictions when site data is insufficient.

[0140] Predict G for future unobserved soil layers and CPT locations through the Bayesian modelmax Perform probabilistic prediction on the values and generate a representative profile.

[0141] Through the posterior update of the model, further optimize the prediction accuracy of G after obtaining new CPT or soil layer data. max of.

[0142] In summary, by using the Bayesian cross-classification hierarchical model, the present invention can more effectively utilize CPT data, consider the potential differences of different CPT positions and soil layer units, and provide a high-resolution probabilistic prediction of the small-strain shear modulus G of the soil layer in the offshore wind farm. max Through this method, a more accurate representative profile of the soil layer can be generated, improving the accuracy and reliability in designing the wind turbine foundation.

[0143] Next, taking the publicly available geotechnical investigation dataset of the TNW offshore wind farm as an example, the implementation process of the present invention will be described in detail. It should be noted that the present invention introduces a general method and does not depend on the specific TNW dataset of the wind farm in the example.

[0144] 1. Data preparation stage:

[0145] 1) Data collection:

[0146] Refer to Figure 3 , Figure 3 is an overview map of the TNW wind farm with CPT borehole positions: showing the geographical location of the wind farm, the CPT borehole positions and the SCPT borehole markings, and explaining the spatial distribution of data collection. Obtain the geological, geophysical and geotechnical data of the Dutch North Sea Tennoor denvande Waddeneilanden (TNW) offshore wind farm from publicly available data sources. These data include detailed measurement records at 162 CPT positions, among which 14 sand units and 7 clay units are identified, forming a total of 896 cross-classification groups. At the same time, obtain 407 pairs of SCPT and in-situ G max values as the calibration dataset for the CPT-based empirical model, which covers 125 cross-classification groups, involving 13 sand units and 6 clay units, from 31 CPT positions.

[0147] 2) Data preprocessing:

[0148] Calculate the effective vertical stress σ′ v , and calculate the effective vertical stress corresponding to each data point according to the formula σ′ v = depth below the seabed × 10 kPa / m (based on the average effective stress gradient in the North Sea waters). Convert the in-situ G max value from the in-situ shear wave velocity V s , using the formula (where ρ is the soil density, determined according to soil type and actual measurement), and convert the G max value to MPa units to ensure the consistency and accuracy of the data in subsequent calculations.

[0149] Sort the data according to the CPT location and soil unit, and construct a cross-classification data structure. Count the sample numbers for each CPT location, soil unit, and cross-classification group to form the sample size summary table 1, so as to clearly understand the data distribution and provide a basis for subsequent model construction and analysis.

[0150] Table 1. Sample sizes for different combinations of CPT locations and soil units in the TNW dataset

[0151]

[0152] 2. Model construction and fitting stage

[0153] Refer to Figure 4 , Figure 4 is the directed acyclic graph of the Bayesian regression model, which includes a cross-classification hierarchical model (as shown in (a) of Figure 4 ), a location classification partially mixed model (as shown in (b) of Figure 4 ), and a diagram of the unit classification partially mixed model (as shown in (c) of Figure 4 ), visually presenting the model structure and parameter relationships.

[0154] 1) Construction and programming implementation of the fully mixed model

[0155] Based on the assumptions of the fully mixed model, construct a regression model using empirical formulas. In a probabilistic programming environment (such as Stan), convert the model formula into executable code.

[0156] 2) Construction and programming implementation of the cross-classification hierarchical model

[0157] Construct a cross-classification hierarchical model, consider the variation between groups, assume an exchangeable intercept β0, and other parameters are the same among different CPT locations and soil units, and write the code in Stan according to the partially mixed model formula.

[0158] 3) Model fitting and convergence assessment

[0159] Run the MCMC simulation and use the No-U-Turn Sampler in Stan for parameter estimation. Run three Markov chains in parallel, with each chain containing 5000 simulated posterior samples. During the simulation, closely monitor the convergence of the chains. Determine whether the posterior distribution of the parameters has converged by checking statistical metrics within and between the chains (such as the Rhat value). If the Rhat value is close to 1 (usually considered to have good convergence within 1.05), it indicates that the model has converged; otherwise, adjust the simulation parameters (such as increasing the number of iterations, adjusting the prior distribution, etc.) and run the simulation again until a stable posterior distribution is obtained.

[0160] As Figure 5 shown, Figure 5 is the posterior estimation plot of the CPT location and soil element effects: showing the posterior means and 95% confidence intervals of the CPT location and soil element effects for the sand data and clay data in TNW respectively ( and ) for analyzing the variation and its uncertainty between locations and elements. Among them, Figure 5 (a) shows the sandy soil data, Figure 5 (b) shows the clayey soil data.

[0161] Figure 6 is the comparison plot of β0 estimates: comparing the β0 estimates of the fully mixed model and the partially mixed model for specific sand groups (such as the groups related to 083 - BH - CPT and 072 - BH - CPT), highlighting the deficiency of the fully mixed model in considering variation.

[0162] Figure 7 is the posterior standardized residual plot: showing the posterior standardized residual distributions of the fully mixed and partially mixed models on the TNW dataset for a preliminary check of the model fit and to judge the deviation degree of data points from the model predictions.

[0163] Figure 8 is the LOGO analysis result plot: presenting the measured q t profiles, the representative G max profiles of the prior and posterior predictions for multiple selected leave-one-out populations (such as different location and element combinations), as well as the relative changes of profiles A and B, visually demonstrating the prediction and updating capabilities of the model. Figure 8 (a) in it is the new clay unit GGM31B at the new location 081 - SCPT; Figure 8 (b) in it is the existing clay unit GGM52 at the new location 016 - BH - CPT; Figure 8 (c) in it is the existing sandy soil unit GGM42 at the existing location 072 - BH - CPT; Figure 8 (d) in it is the existing sandy soil unit GGM31A at the existing location 034 - SCPT.

[0164] Figure 9 This is the final prediction example diagram of the entire wind farm: taking the existing location 034 - SCPT and the new location 006 - PCPT as examples, it shows the predicted G of all soil units based on CPT measurement data max profile, covering different prediction scenarios, and illustrating the application effect of the model in the entire wind farm.

[0165] 3. Model evaluation and comparison stage:

[0166] 1) Posterior distribution analysis:

[0167] Calculate posterior means, 95% confidence intervals (CIs), residual standard deviation τ ε , intraclass correlation coefficient (ICC) and other indicators for model parameters. The posterior mean can be used as the best estimate of the parameter, and the 95% confidence interval reflects the uncertainty range of the parameter estimate. By analyzing these indicators, evaluate the fitting effect and interpretability of the model to the data. For example, a smaller residual standard deviation τ ε usually indicates a better fit of the model to the data; the ICC value can be used to measure the similarity degree between data groups and help judge whether the model assumptions (such as the full - mixture model or the partial - mixture model) are reasonable.

[0168] 2) Model comparison and selection

[0169] Use the leave - one - out cross - validation (LOO - CV) method to calculate the expected log - predictive density (elpd) of the full - mixture model, the cross - classified hierarchical model, and other alternative models (such as the location - classified and unit - classified partial - mixture models). The larger the elpd value, the higher the prediction accuracy of the model. Calculate the difference (Δelpd) between the elpd value of each model and the elpd value of the best model (the cross - classified hierarchical model in the present invention) and its standard error (SE), and judge the superiority and inferiority between models according to the magnitudes of the difference and the standard error. If the absolute value of Δelpd is greater than several times (such as more than 3 times) its standard error, it can be considered that there is a significant difference between models; otherwise, the difference is not significant. According to the comparison results, select the model with the highest prediction accuracy as the final model for subsequent G max profile prediction.

[0170] 4.G max Profile prediction and update stage

[0171] 1) Calculation of β0 under different prediction scenarios

[0172] According to whether the CPT location and soil unit are in the calibration dataset, different prediction scenarios are determined. For the case of existing locations and existing units with observed data (prediction scenario 1), directly use the Calculate the group-specific β0. For the existing location and existing element but without observed data (prediction scenario 2), the above formula is also used to calculate β0 because the location and element effects have been estimated in the calibration data. For the existing location and new element (prediction scenario 3), β0 is obtained by sampling from ; for the new location and existing element (prediction scenario 4), β0 is sampled from ; for the new location and new element (prediction scenario 5), β0 is sampled from .

[0173] 2) G max Profile Prediction and Uncertainty Analysis

[0174] Using the calculated β0 value, predict the G max profile (ABCDE profile) for each CPT location and soil element according to the constructed model formula. For each predicted profile, calculate statistical indicators such as its mean, 95% confidence interval, and 90% prediction interval to reflect the uncertainty of the prediction results. Analyze the G max profile change trends of different soil elements and CPT locations, and evaluate the impact of the spatial variability of soil properties on the design of wind turbine foundations. For example, a larger prediction interval indicates greater uncertainty in the G max value at this location or element, and more conservative design parameters need to be considered when designing wind turbine foundations.

[0175] 3) Leave-One-Group (LOGO) Analysis

[0176] Conduct LOGO analysis. Each time, select the data of one soil element at one CPT location from the original dataset as the validation set, and keep it to observe the prediction ability of the model before and after observing new data. Fit the model with the remaining data, conduct prior prediction on the data reserved for validation, and obtain the prior predicted G max profile (including profiles A - E). Then add the validation data to the model for posterior prediction, and calculate the posterior predicted G max profile and its related statistical indicators again. Compare the changes in the prior predicted and posterior predicted profiles, and analyze the update of the model after observing new data, such as the relative changes (δ A and δ B ) of profiles A and B, to reflect the effect of reducing the uncertainty of the model and its adaptability to new data. Through multiple examples of LOGO analysis for different combinations of locations and elements, verify the prediction and update abilities of the model in different situations to ensure that the model has good generalization and reliability.

[0177] 5. Practical Application Stage

[0178] 1) Full Wind Farm G max Profile Prediction

[0179] Apply the finally selected cross - classification hierarchical model to the entire TNW wind farm, and perform G max profile prediction for all CPT positions and soil elements not included in the calibration dataset. According to different prediction scenarios, calculate the β0 value of each group according to the above calculation method, and predict the corresponding G max profile. Generate the G max profile prediction map of the entire wind farm, visually showing the G max distribution at different positions and soil elements, providing comprehensive data support for the design of offshore wind turbine foundations.

[0180] 2) Design optimization and decision - making support

[0181] According to the predicted G max profile, optimize the design of the wind turbine foundation. For example, when determining the size, embedment depth, and stiffness of the wind turbine foundation, fully consider the spatial variability of soil properties, select appropriate design parameters, and ensure the stability and safety of the wind turbine during its entire service life. At the same time, the prediction results can also be used to evaluate the bearing capacity and deformation characteristics of wind turbine foundations at different positions, providing a decision - making basis for the layout planning and operation management of the wind farm. For example, for areas with low G max values or large variability, additional reinforcement measures or adjustment of the wind turbine layout plan can be taken to reduce risks and improve the overall performance of the wind farm.

[0182] 3) Model update and continuous improvement

[0183] During the construction and operation of the wind farm, as new data is continuously obtained (such as new CPT test data or long - term monitoring data), the model can be updated regularly. Incorporate the new data into the model calibration process, re - evaluate the model parameters and prediction results, and further optimize the accuracy of the G max profile prediction. At the same time, according to actual engineering experience and feedback, continuously improve the assumptions and structure of the model, enhance the adaptability of the model to complex marine geological conditions, and ensure that the model can always provide reliable technical support for the design and operation of offshore wind farms.

[0184] Apply the finally selected model to the entire TNW wind farm, perform G max profile prediction for all CPT positions and soil elements not included in the calibration dataset, providing a basis for the design of offshore wind turbine foundations, ensuring that the design takes into account the spatial variability of soil parameters, and improving the reliability and safety of the design. At the same time, according to actual engineering feedback and the acquisition of new data, the model and prediction results can be further optimized.

[0185] Compared with the existing technology, the present invention has the following advantages:

[0186] 1. The Bayesian cross-classification hierarchical model of the present invention can effectively capture the variation between CPT positions and soil units, and can provide a more reasonable estimate of parameter uncertainty and better fit to the data compared with the traditional fully mixed model.

[0187] 2. Through LOO-CV analysis and verification, the model has a significant improvement in the out-of-sample prediction accuracy of G max and can provide a more reliable G max representative profile prediction for the design of wind turbine foundations in offshore wind farms, which helps to more reasonably design wind turbine foundations and improve the safety and economy of offshore wind farms.

[0188] Another aspect of the embodiments of the present invention also provides a Bayesian hierarchical probability prediction system for the small-strain shear modulus of an offshore wind farm based on CPT, including:

[0189] A first module for obtaining the small-strain shear modulus data of soil at different CPT positions in an offshore wind farm and constructing a training sample;

[0190] A second module for constructing a Bayesian regression model according to the training sample, including constructing a fully mixed model and a partially mixed model;

[0191] A third module for performing Markov chain Monte Carlo (MCMC) simulation on the constructed Bayesian regression model to obtain the posterior distribution of model parameters, and then evaluating the fitting effect of the model to obtain a target model that meets the prediction effect requirements;

[0192] A fourth module for dividing multiple scenario prediction G max profiles according to the target model, and then determining whether the CPT position and the soil unit are in the calibration dataset to complete the probability prediction.

[0193] It can be understood that the content in the above method embodiments is applicable to the system embodiments of the present invention. The functions specifically implemented by the system embodiments of the present invention are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0194] The embodiments of the present invention also provide an electronic device. The electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the above-mentioned Bayesian hierarchical probability prediction method for the small-strain shear modulus of an offshore wind farm based on CPT. The electronic device can be any intelligent terminal including a tablet computer, an in-vehicle computer, etc.

[0195] It can be understood that the content in the above method embodiments is applicable to the device embodiments of the present invention. The functions specifically implemented by the device embodiments of the present invention are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0196] An embodiment of the present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned Bayesian hierarchical probability prediction method for small-strain shear modulus of an offshore wind farm based on CPT is implemented.

[0197] It can be understood that the content in the above method embodiments is applicable to this storage medium embodiment. The functions specifically implemented by this storage medium embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0198] As a non-transitory computer-readable storage medium, the memory can be used to store non-transitory software programs and non-transitory computer-executable programs. In addition, the memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0199] It should be noted that in each specific embodiment of the present invention, when it comes to relevant processing based on data related to the user's identity or characteristics, such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first. Moreover, the collection, use, and processing of these data will comply with relevant laws, regulations, and standards. In addition, when the embodiments of the present invention need to obtain the user's sensitive personal information, the user's separate permission or separate consent will be obtained through methods such as pop-up windows or jumping to a confirmation page. After clearly obtaining the user's separate permission or separate consent, the necessary user-related data for the normal operation of the embodiments of the present invention will be obtained.

[0200] The embodiments described in the embodiments of the present invention are for more clearly illustrating the technical solutions of the embodiments of the present invention, and do not constitute a limitation to the technical solutions provided by the embodiments of the present invention. Those skilled in the art know that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of the present invention are also applicable to similar technical problems.

[0201] Those skilled in the art can understand that the technical solutions shown in the figures do not constitute a limitation to the embodiments of the present invention, and may include more or fewer steps than those shown in the figures, or combine some steps, or different steps.

[0202] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0203] Those of ordinary skill in the art can understand that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, and their appropriate combinations.

[0204] As used in the specification of the present invention and the above-mentioned drawings, the terms "first", "second", "third", "fourth", etc. (if any) are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0205] It should be understood that in the present invention, "at least one (item)" means one or more, and "a plurality" means two or more. "And / or" is used to describe the association relationship of associated objects and indicates that three relationships can exist. For example, "A and / or B" can mean: only A exists, only B exists, and both A and B exist at the same time. Among them, A and B can be singular or plural. The character " / " generally indicates that the associated objects before and after are in an "or" relationship. "At least one (one)" or its similar expression below refers to any combination of these items, including any combination of single item (one) or plural items (ones). For example, at least one (one) of a, b, or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0206] In several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the above-mentioned units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of devices or units can be in electrical, mechanical or other forms.

[0207] The units described above as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0208] In addition, each functional unit in various embodiments of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0209] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs that can store programs.

[0210] The preferred embodiments of the embodiments of the present invention have been described above with reference to the accompanying drawings. However, this does not limit the scope of the rights of the embodiments of the present invention. Any modifications, equivalent replacements, and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present invention shall fall within the scope of the rights of the embodiments of the present invention.

Claims

1. A Bayesian hierarchical probability prediction method for small-strain shear modulus of an offshore wind farm based on CPT, characterized in that, Including the following steps: Obtain the small-strain shear modulus data of the soil at different CPT positions in the offshore wind farm and construct a training sample; Construct a Bayesian regression model based on the training sample, including constructing a fully pooled model and a partially pooled model; Perform Markov Chain Monte Carlo (MCMC) simulation on the constructed Bayesian regression model to obtain the posterior distribution of the model parameters, and then evaluate the fitting effect of the model to obtain a target model that meets the prediction effect requirements; Divide multiple scenario predictions G according to the target model max for the cross-section, and then determine whether the CPT position and soil unit are in the calibration dataset to complete the probability prediction; The step of constructing the fully pooled model in the step of constructing the Bayesian regression model according to the training sample includes the following steps: Based on empirical formulas Construct a fully mixed regression model, where q t represents the total cone tip resistance of the static cone penetration test (CPT) corrected for pore pressure; p a represents the atmospheric pressure; σ v ′ represents the effective vertical stress; α0, α1, α2 represent the three unknown parameters of the model; Set a prior distribution for the model parameters and use a weakly informative prior to reflect the lack of specific prior knowledge; Characterize the preliminary estimate of the possible value range of the parameters according to the prior distribution; The step of constructing the partially pooled model in the step of constructing the Bayesian regression model according to the training sample includes the following steps: Consider the variation between the CPT positions and the soil elements and construct a cross-classified hierarchical model based on the set exchangeable intercept; 2. The Bayesian hierarchical probability prediction method for small strain shear modulus of an offshore wind farm based on CPT according to claim 1, characterized in that, The step of obtaining the small-strain shear modulus data of each soil element at different CPT positions in the offshore wind farm and constructing a training sample includes the following steps: Collect geological, geophysical, and geotechnical data of an offshore wind farm from specified data sources, including cone tip resistance, pore water pressure data obtained from piezocone tests at multiple CPT locations, and in-situ G max values measured by seismic piezocone tests SCPT or other methods; Calculate the effective vertical stress according to the site geological conditions and the measurement depth, specifically: use the method of multiplying the depth below the seabed by the average effective stress gradient to calculate the effective vertical stress corresponding to each data point; Convert the G max value from the original measurement unit to MPa, classify it according to the CPT position and soil element, construct a cross-classification data structure, count the number of samples in each classification group, form a data summary, and construct the training sample obtained above.

3. A Bayesian hierarchical probability prediction method for the small strain shear modulus of an offshore wind farm based on CPT according to claim 1, characterized in that, The step of performing Markov Chain Monte Carlo (MCMC) simulation on the constructed Bayesian regression model to obtain the posterior distribution of the model parameters includes the following steps: Use the probabilistic programming language Stan combined with the Markov Chain Monte Carlo simulation method to fit the constructed Bayesian regression model; Run multiple Markov chains in parallel, and each chain performs the corresponding number of simulation samplings to obtain the posterior distribution of the model parameters; Among them, during the simulation process, judge whether the posterior distribution of the parameters converges by checking the statistical indicators within and between the chains. If the posterior distribution does not converge, adjust the simulation parameters and re-run the simulation until a stable posterior distribution is obtained; among them, the simulation parameters include increasing the number of iterations and adjusting the prior distribution.

4. The Bayesian hierarchical probability prediction method for the small strain shear modulus of an offshore wind farm based on CPT according to claim 3, wherein, The step of evaluating the fitting effect of the model to obtain a target model that meets the prediction effect requirements includes the following steps: Analyze the posterior distribution of the model parameters, calculate the statistical indicators of the posterior mean, standard deviation, and 95% confidence interval, and evaluate the fitting effect of the model to the data; Calculate the residual standard deviation τ ε , and judge the goodness of fit of the model by comparing the τ ε values of the completely mixed model and the cross-classified stratified model; Calculate the intraclass correlation coefficient, and the formula is Measure the similarity degree between data groups through the intraclass correlation coefficient; where ICC represents the intraclass correlation coefficient; Represents the variance of the local group and also represents G max The variance between different CPT positions; Represents the variance of the superior group and also represents G max The variance between different soil units; Represents G max The variance of the residuals; Use the leave-one-out cross-validation method to compare the expected log predictive densities of different models and evaluate the prediction accuracy of the models; Select the target model with the best prediction performance according to the values of the predictive densities, their differences, and the standard errors; 5. A Bayesian hierarchical probability prediction method for small strain shear modulus of an offshore wind farm based on CPT according to claim 1, characterized in that, Dividing multiple scenario predictions G according to the target model max The profile is further used to determine whether the CPT position and soil unit are in the calibration dataset, and probability prediction is completed, including the following steps: Predict G in the following scenarios based on the CPT position and whether the soil element is in the calibration dataset max Profile: Existing position and existing unit with observed data: using the estimates in the calibration data Calculate group-specific β0 and then predict G max Profile; where β 0(kk) Represents the intercept at CPT position j and soil unit k; Represents the overall mean of the intercept β0; Represents the local deviation for position j; Represents the local deviation for unit k; β0 represents the overall base intercept; Existing location and existing unit but no observed data: Use β estimated from calibration data 0(jk) Calculate β0 for prediction; Existing position and new unit: Sampling from to obtain β0, and then substituting it into the model formula to predict G max Profile, where and are the corresponding hyperparameters and parameter estimates; New location and existing unit: Sampling β0 from and then substituting it into the model formula to predict G max Profile, where and are the corresponding hyperparameters and parameter estimates; New location and new unit: Sampling β0 from and then substituting it into the model formula to predict G max profile, where and are the corresponding hyperparameters and parameter estimates; Perform a leave-one-LOGO analysis. Each time, leave the data within a soil unit located at a certain CPT position as the validation set, fit the model with the remaining data and make a prior prediction, and then add the validation data for posterior prediction; compare the G of the prior and posterior predictions max changes in the profile, analyze the update of the model after observing new data, and by comparing these changes in the profile, analyze the improvement of the target model after obtaining new data.

6. A system for implementing a Bayesian hierarchical probability prediction method for the small strain shear modulus of an offshore wind farm based on CPT as described in any one of claims 1-5, characterized in that, Including: The first module is used to obtain the small-strain shear modulus data of the soil at different CPT positions in the offshore wind farm and construct a training sample; The second module is used to construct a Bayesian regression model based on the training sample, including constructing a fully pooled model and a partially pooled model; The third module is used to perform Markov Chain Monte Carlo (MCMC) simulation on the constructed Bayesian regression model to obtain the posterior distribution of the model parameters, and then evaluate the fitting effect of the model to obtain a target model that meets the prediction effect requirements; The fourth module is used to divide multiple scenario predictions G according to the target model max profiles, and then determine whether the CPT position and soil units are in the calibration dataset to complete probability prediction.

7. An electronic device, characterized in that, Including a processor and a memory; The memory is used to store a program; The processor executes the program to implement the method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The storage medium stores a program, and the program is executed by the processor to implement the method according to any one of claims 1 to 5.

Images