Multi-source geothermal data fusion and intelligent processing method and device based on uncertainty perception, equipment and storage medium

Abstract

The application provides a kind of based on uncertainty perception multi-source geothermal data fusion and intelligent processing method, device, equipment and storage medium, it is related to geothermal resource exploration technical field.The method comprises: the mixed interpolation of uncertainty driven geothermal data is generated interpolation data and corresponding comprehensive uncertainty;Based on comprehensive uncertainty, adaptively adjust local threshold for anomaly detection;Anomaly detection is carried out on geothermal data using the adjusted local threshold, and abnormal data is identified and marked.The method couples interpolation uncertainty and anomaly detection threshold, and in geothermal data sparse or complex geological area (such as thermal reservoir boundary, fracture zone) can automatically relax anomaly determination standard, significantly reduce false alarm rate, while improving interpolation accuracy.Experiments show that, compared with prior art, interpolation root mean square error is reduced by 10%-20%, anomaly detection F1 value is improved to 0.84, especially in complex geological area, the effect is more prominent.

Description

Technical Field

[0001] This application relates to the field of geothermal resource exploration technology, and in particular to a method, apparatus, equipment and storage medium for multi-source geothermal data fusion and intelligent processing based on uncertainty perception. Background Technology

[0002] Geothermal resource exploration and development rely on multi-source heterogeneous data, including well logging temperature, pressure, core thermal conductivity, and regional geological survey data. Geothermal data has the following unique characteristics: First, data sparsity: geothermal well drilling is costly, and the data points per well are limited and the distance between wells is large; Second, depth correlation: geothermal parameters exhibit strong correlations along the well depth direction, influenced by geothermal gradients, lithological variations, and reservoir boundaries; Third, multi-source heterogeneity: different measurement methods (steady-state temperature measurement, transient temperature measurement, distributed fiber optic temperature measurement) show significant differences in accuracy and spatiotemporal resolution; Fourth, physical constraints: geothermal parameters must satisfy the heat conduction equation and geological structural constraints, and outlier identification requires the integration of geothermal geology knowledge.

[0003] Existing data processing methods have the following shortcomings: First, traditional interpolation methods (such as linear interpolation and K-nearest neighbor) do not consider the depth correlation and physical constraints of geothermal data, resulting in large interpolation errors in areas with abrupt temperature gradient changes, such as the boundaries of geothermal reservoirs.

[0004] Second, single machine learning methods (such as random forests) lack the ability to quantify uncertainty and cannot identify high-risk interpolation points, leading to systematic biases in downstream thermal energy storage evaluation.

[0005] Third, the anomaly detection uses a globally fixed threshold and does not take into account the differences in the statistical characteristics of geothermal data at different depths and in different geological units, resulting in a high false alarm rate in thermal anomaly areas (such as heat flow channels in fault zones).

[0006] Fourth, existing methods treat interpolation and anomaly detection as independent steps, without establishing an information feedback mechanism between the two, thus failing to achieve closed-loop optimization of quality control.

[0007] Therefore, there is a need for a data quality control and intelligent processing solution that is tailored to the characteristics of geothermal data, integrates prior knowledge of geothermal geology, and has the ability to quantify uncertainty and adaptively regulate. Summary of the Invention

[0008] This application provides a method, apparatus, device, and storage medium for multi-source geothermal data fusion and intelligent processing based on uncertainty perception, to solve at least one of the following technical problems: First, how to integrate prior knowledge of geothermal geology (geothermal gradient, lithospheric thermal structure, thermal reservoir boundary constraints) into geothermal data interpolation to overcome the insufficient accuracy of traditional interpolation methods in regions with abrupt temperature gradient changes. Second, how to provide reliable uncertainty quantification for each interpolation point so that downstream thermal storage evaluation can make risk-aware decisions and avoid systematic evaluation bias caused by high uncertainty interpolation values; Third, how to establish an information coupling mechanism between interpolation and anomaly detection so that the threshold for anomaly detection can be adaptively adjusted according to data uncertainty, thereby reducing the false alarm rate in hot anomaly areas; Fourth, how to construct a closed-loop quality control process that can automatically backtrack and adjust when interpolation or testing results do not meet quality requirements, and provide a priority ranking basis for manual review.

[0009] Firstly, this application provides a method for multi-source geothermal data fusion and intelligent processing based on uncertainty perception, including: Uncertainty-driven hybrid interpolation is performed on geothermal data to generate interpolated data and the corresponding comprehensive uncertainty; Based on the comprehensive uncertainty, the local threshold used for anomaly detection is adaptively adjusted; the adjusted local threshold is used to detect anomalies in geothermal data, and the abnormal data is identified and marked.

[0010] In one possible design, uncertainty-driven hybrid interpolation is performed on geothermal data to generate interpolated data and the corresponding comprehensive uncertainty, including: For each data point to be interpolated, multiple estimators are used to generate candidate estimates and their corresponding uncertainties. Based on the candidate estimates and their uncertainties, the final interpolated values ​​and their combined uncertainties are calculated using a linear fusion strategy that employs inverse variance weighting or covariance sensing.

[0011] In one possible design, when using inverse variance weighted fusion, the weights are calculated using the following formula. Final interpolation value and comprehensive uncertainty : in, For the first i One estimate The corresponding uncertainty, To prevent division by zero of small constants; When using covariance-sensing linear fusion, the fusion value is calculated using the following formula. and uncertainty : in, To estimate the covariance matrix of the instrument, It is a unit vector.

[0012] In one possible design, the multiple estimators include a prior model based on geothermal geology, which generates candidate estimates. Determined by the following formula: in, For geothermal gradient, For the thickness of the lithosphere, For burial depth, This represents the regional heat flux value. , , , , and These are coefficients determined based on regional geostatistical regression.

[0013] In one possible design, based on the comprehensive uncertainty, a local threshold for anomaly detection is adaptively adjusted; the adjusted local threshold is used to perform anomaly detection on geothermal data, identifying and labeling anomalous data, including: A comprehensive uncertainty index u is calculated for each sample, and the comprehensive uncertainty index is informationally coupled with the comprehensive uncertainty. Based on the comprehensive uncertainty index u, dynamically adjust the local threshold; The final anomaly score is obtained by integrating the outputs of multiple anomaly detection algorithms, and the final anomaly score is compared with an adjusted local threshold to determine an anomaly.

[0014] In one possible design, the formula for calculating the uncertainty index u is: in, For normalized data variance, To normalize the inconsistency in predictions between models, This is a normalized neighborhood sparsity statistic obtained based on neighborhood density conversion. , , The weighting coefficients are and satisfy the following conditions: + + =1; Based on the comprehensive uncertainty index u, the local threshold is dynamically adjusted using the following formula. : in, The preset baseline threshold, This is a preset adjustable coefficient; The formula for calculating the final anomaly score is as follows: in, These are the normalized isolated forest algorithm isolated scores. The reconstruction error is obtained by the autoencoder algorithm. To integrate weights, This is the normalization function; The specific method for comparing the final anomaly score with the adjusted local threshold to determine anomalies is as follows: if the final anomaly score is greater than or equal to the local threshold, the sample is determined to be an anomaly; otherwise, it is determined to be a normal sample.

[0015] In one possible design, after performing uncertainty-driven hybrid interpolation on geothermal data to generate interpolated data and the corresponding comprehensive uncertainty, an adaptive backtracking step is also included; the adaptive backtracking step includes: When the overall uncertainty exceeds a preset threshold τ var If the deviation between the interpolated value and the prior value obtained based on the geothermal geology prior model exceeds a preset range, backtracking is triggered to adjust the parameters of the interpolation method or initiate manual review. The preset range is determined by the following formula: in, For preset coefficients, To account for the overall uncertainty, The prior uncertainty corresponding to the prior model; Based on the comprehensive uncertainty, the local threshold used for anomaly detection is adaptively adjusted; after using the adjusted local threshold to detect anomalies in geothermal data, identifying and marking anomalous data, a spatial folding management and backtracking step is also included; the spatial folding management and backtracking step includes: Spatial layers are divided according to wells or geographic grids. Anomaly detection is performed on each spatial layer. When the detection error exceeds the limit, the parameters of the corresponding spatial layer are recalibrated or the model is locally retrained.

[0016] Secondly, this application provides a device for multi-source geothermal data fusion and intelligent processing based on uncertainty perception, the device comprising: The hybrid interpolation unit is configured to perform uncertainty-driven hybrid interpolation on geothermal data, generating interpolated data and the corresponding comprehensive uncertainty; An anomaly detection unit is configured to adaptively adjust a local threshold for anomaly detection based on the comprehensive uncertainty; and to use the adjusted local threshold to detect anomalies in geothermal data, identify and mark the anomalous data.

[0017] Thirdly, embodiments of this application provide an electronic device, including: at least one processor and a memory; the memory stores computer execution instructions; the at least one processor executes the computer execution instructions stored in the memory, causing the at least one processor to perform the uncertainty-aware multi-source geothermal data fusion and intelligent processing method described in the first aspect and various possible designs of the first aspect.

[0018] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions. When a processor executes the computer-executable instructions, it implements the multi-source geothermal data fusion and intelligent processing method based on uncertainty perception as described in the first aspect and various possible designs of the first aspect.

[0019] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the uncertainty-aware multi-source geothermal data fusion and intelligent processing method described in the first aspect and various possible designs of the first aspect.

[0020] The method, apparatus, equipment, and storage medium for multi-source geothermal data fusion and intelligent processing based on uncertainty perception provided in this application have at least the following beneficial effects: First, the interpolation accuracy is significantly improved. By integrating prior knowledge of geothermal geology with data-driven methods, the interpolation accuracy in regions with abrupt temperature gradient changes, such as geothermal reservoir boundaries, is significantly better than traditional methods. Compared to single K-nearest neighbor interpolation, the root mean square error is reduced by approximately 10%-20%; compared to single random forest regression, the root mean square error is reduced by approximately 6%-12%. The introduction of geological priors makes the interpolation results more physically reasonable, avoiding the overfitting problem of purely data-driven methods in sparse data regions.

[0021] Second, it exhibits superior anomaly detection performance. An adaptive threshold adjustment mechanism based on uncertainty is designed to address the spatial heterogeneity of geothermal data, significantly reducing the false alarm rate in geologically complex areas such as fault zones and heat flow channels, while maintaining the ability to detect errors in real data. The F1 score for anomaly detection improved from 0.69-0.72 at the baseline to 0.84.

[0022] Third, information coupling between interpolation and anomaly detection. This invention establishes a quantitative correlation between interpolation uncertainty and anomaly detection threshold, enabling the two processes to operate in synergistic optimization rather than in isolation. High-uncertainty interpolation points automatically receive more lenient anomaly thresholds, avoiding the cascading amplification of false alarms.

[0023] Fourth, improved performance in downstream tasks. Using the data processed by this invention to train the prediction model, the coefficient of determination is improved by approximately 8%-15% in downstream tasks such as reservoir thermal conductivity prediction and thermal reserve assessment. The quantified uncertainty information can be transmitted downstream to support risk-aware thermal reserve evaluation decisions.

[0024] Fifth, complete traceability and closed-loop quality control. The metadata recording mechanism ensures that the processing history of each data point is traceable, and the backtracking mechanism automatically triggers parameter adjustments or manual review when quality fails to meet standards, forming a closed-loop quality control process. Attached Figure Description

[0025] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0026] Figure 1 A flowchart illustrating a method for multi-source geothermal data fusion and intelligent processing based on uncertainty perception, provided in an embodiment of this application; Figure 2 A flowchart of uncertainty-driven geothermal data hybrid interpolation provided for embodiments of this application; Figure 3 This is a flowchart of the adaptive anomaly detection process provided in an embodiment of this application; Figure 4 A structural diagram of the multi-source geothermal data fusion and intelligent processing system based on uncertainty perception provided in the embodiments of this application; Figure 5 This is a structural diagram of the multi-source geothermal data fusion and intelligent processing device based on uncertainty perception provided in the embodiments of this application.

[0027] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concepts of this application to those skilled in the art through reference to specific embodiments. Detailed Implementation

[0028] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0029] The collection, storage, use, processing, transmission, provision, and disclosure of financial data or user data involved in the technical solution of this application all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0030] It should be noted that in the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, it does not mean that the applicant has used or necessarily used the solution.

[0031] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0032] This application provides an uncertainty-aware multi-source geothermal data fusion and intelligent processing method to address the problems of geothermal data sparsity, deep correlation, and physical constraints. The core of this method comprises two parts: First, uncertainty-driven hybrid interpolation combines K-nearest neighbors, random forests, and geothermal geological priors (such as geothermal gradient, lithosphere thickness, and heat flow) to generate multiple candidate estimates and calculate the uncertainty for each estimate. Then, inverse variance weighted or covariance-aware linear fusion is used to automatically rely more on geological priors in sparse data areas, and adaptive backtracking and manual verification mechanisms are introduced to ensure the reliability of the results. Second, adaptive anomaly detection dynamically adjusts local thresholds based on sample-level uncertainty, automatically relaxing thresholds in geologically complex areas such as geothermal reservoir boundaries and fault zones to reduce false alarms. Anomaly discrimination is achieved by fusing the outputs of isolated forests and autoencoders, and a confidence level is given for each anomaly. The system consists of modules for data access, parsing and preprocessing, anomaly detection, hybrid interpolation, data fusion, quality assessment, and storage, supporting layered deployment, online monitoring, and rollback. Experimental results show that, compared with traditional methods, the root mean square error of interpolation is reduced by 10%–20% (approximately 33% in the boundary region of thermal reservoirs), the F1 value of anomaly detection is increased to 0.84, and the false alarm rate in fault zone areas is reduced from 35% to 12%, with particularly significant effects in geologically complex areas.

[0033] like Figure 1 As shown, the multi-source geothermal data fusion and intelligent processing method based on uncertainty perception specifically includes the following steps S10-S20.

[0034] S10: Perform uncertainty-driven hybrid interpolation on geothermal data to generate interpolated data and the corresponding comprehensive uncertainty.

[0035] Step S10 is used to implement uncertainty-driven hybrid interpolation of geothermal data. In some embodiments, such as Figure 2 As shown, step S10 can be implemented through the following steps S101-S104.

[0036] S101: Multiple estimation generation.

[0037] For each data point to be interpolated, candidate estimates and their respective uncertainties are generated using K-nearest neighbor interpolation, random forest regression, and a prior model based on geothermal geology.

[0038] Specifically, K-nearest neighbor interpolation searches for neighboring points in the depth-temperature feature space, with the number of neighbors k ranging from 3 to 15. The distance metric comprehensively considers depth difference and horizontal distance, and the uncertainty is calculated based on the weighted variance of the values ​​within the neighborhood. Random forest uses depth, lithology coding, and regional geothermal gradient as input features, with the number of decision trees ranging from 50 to 500, and the uncertainty is calculated based on the variance of the predicted values ​​of each decision tree.

[0039] The geothermal geology prior model is established based on the regional heat flow background and the thermal structure of the lithosphere. The model form is as follows: in, For geothermal gradient, For the thickness of the lithosphere, For burial depth, This represents the regional heat flux value. , , , , and These are coefficients determined based on regional geostatistical regression.

[0040] Prior variance was prior The variance of the regression residuals is synthesized from the propagation of parameter uncertainty. It automatically increases near the boundary of the thermal reservoir due to abrupt changes in the temperature gradient, reflecting geological uncertainty.

[0041] S102: Uncertainty Fusion.

[0042] The candidate estimates are weighted and fused based on their uncertainties, and the fusion strategy is automatically selected based on the correlation between the estimators.

[0043] When the estimators are approximately independent, such as K-nearest neighbors and geological priors which are based on different information sources, inverse variance weighting is used to calculate the weights. : Calculate the final interpolation value The calculation formula is: Overall uncertainty The calculation formula is: in, For the first i One estimate The corresponding uncertainty, To prevent small constants from being divided by zero, this fusion method gives greater weight to high-confidence estimates, automatically favoring geological priors in sparse data regions and data-driven estimates in dense data regions.

[0044] When there is a significant correlation between estimators, such as K-nearest neighbors and random forest using the same neighborhood data, covariance-aware linear fusion (BLUE) is employed, with the fusion formula as follows: in, To estimate the covariance matrix of the instrument, It is a unit vector.

[0045] Estimator covariance matrix Numerical stability is ensured by estimating using historical validation data or by employing the Ledoit-Wolf contraction method. This fusion approach avoids underestimation of uncertainty due to neglecting the correlation of estimators.

[0046] In some embodiments, to ensure the comparability and propagation of uncertainties generated by different estimators, the present invention adopts the following conventions and calculation methods in its specific implementation: (1) Calculation of K-nearest neighbor uncertainty For K-nearest neighbor interpolation, let the set of neighboring points be... J The distance weights corresponding to each neighboring point are: w j For example, Gaussian weights can be used: ;in d j The distance between the point to be interpolated and its neighboring points. σd Let be the scaling parameter. Then the K-nearest neighbor estimate is... Weighted variance was knn Calculate using the following formula: This weighted variance reflects the degree of dispersion of values ​​within the neighborhood and serves as the uncertainty of the K-nearest neighbor estimate.

[0047] (2) Calculation of uncertainty in random forest For random forest regression, the variance between the predictions of each decision tree is used as an estimate of the uncertainty. Let the random forest contain T decision trees, and let the i-th... t The predicted value of each tree Then the random forest estimate x rf variance was rf for: in If necessary, probability calibration or quantile regression can be used to adjust the results. was rf Corrections are made to make it more accurately reflect the true error distribution.

[0048] (3) Calculation of uncertainty of prior model in geothermal geology The uncertainty of the prior model is a synthesis of the variance of the regression residuals and the propagation of parameter uncertainties. One approximation method is a first-order Taylor expansion: in For the prior model x prior Regarding the Jacobian matrix of each parameter, Let be the covariance matrix of the model parameters. Another more robust approach is to use Monte Carlo simulation: sample the parameter distribution of the prior model multiple times, calculate the corresponding sample variance of the prior estimates, and use this as... The estimate.

[0049] (4) Rules for the fusion and propagation of uncertainty When multiple estimators are approximately independent, inverse variance weighted fusion is used. Let the estimated value of the i-th estimator be... x i Its uncertainty is was i The combined uncertainty after fusion was imputed and final interpolation value x imputed for: in, To prevent small constants from being divided by zero.

[0050] When there is a significant correlation between the estimators, it is necessary to construct the covariance matrix of the estimators. Σ Its diagonal elements are the variances of each estimator. was i The off-diagonal elements are the covariance (cov) between the estimates. x i , x j The best linear unbiased estimate (BLUE) is used for fusion, and the fused value is... x comb and its uncertainty was comb .

[0051] covariance matrix Σ Stable estimation is crucial for BLUE. This embodiment suggests using shrinkage covariance estimation and provides degradation rules: The Ledoit-Wolf shrinkage method is used to correct the sample covariance matrix, and its expression is as follows: in, The sample covariance matrix is ​​calculated directly from historical validation data or currently available samples. The target matrix, used to guide the contraction direction, can be a diagonal matrix diag( was i (i.e., only retaining the diagonal matrix of the variances of each estimator) or ( It is the identity matrix. (The mean of the variances of each estimator). The contraction strength, with a value range of [0,1], is used to balance the weights of the target matrix and the sample covariance matrix. Its value can be adaptively selected according to the sample size, for example, determined through cross-validation, or preset with an empirical initial value (such as...). =0.1) Then perform calibration using offline data.

[0052] Using shrinkage covariance matrix Before performing BLUE fusion, its numerical stability needs to be evaluated. Specifically, the calculation... The condition number cond( If any of the following conditions are met, a degenerate strategy is triggered, abandoning the use of the complete covariance matrix and instead employing a diagonal approximation (i.e., using only the variances of each estimator, equivalent to inverse variance weighted fusion): 1) The number of conditions exceeds the preset threshold, i.e., cond( )>cond thresh , where cond thresh It can be set based on experience, for example, cond thresh =10 6 ; 2) Effective sample size n eff Less than the number of estimators p Add 2, that is n eff < p +2. The effective sample size can be defined as the number of independent samples participating in the covariance estimation, depending on the actual situation.

[0053] When degradation occurs, the fusion strategy automatically switches to inverse variance weighting to ensure robustness and real-time performance of the computation.

[0054] Considering computational efficiency and real-time requirements, when the number of estimators... p For systems with large values, such as p>10, or those operating in a real-time lightweight layer and sensitive to latency, a diagonal approximation (inverse variance weighting) can be prioritized for fusion to reduce computational complexity and memory overhead. This strategy can be used as the default option to improve processing speed while ensuring basic performance.

[0055] Through the aforementioned shrinkage estimation and degradation mechanism, the method of this application can stably estimate the covariance relationship between estimators under different data conditions, providing reliable input for BLUE fusion while ensuring the real-time performance and robustness of the system.

[0056] (5) Uncertainty calibration To ensure the reliability of the uncertainty estimate, a reliability plot (relationship between prediction variance and empirical mean square error) can be drawn, and a scaling factor c can be calculated such that: If c deviates significantly from 1, then in subsequent applications... Multiply Perform calibration to make it more consistent with the true error distribution.

[0057] S103: Adaptive backtracking.

[0058] When the overall uncertainty exceeds a preset threshold τ var Or the interpolated value deviates from the prior value by more than [a certain amount]. When a backtracking process is triggered, to avoid frequent backtracking due to transient noise, the triggering condition can be set to be that the backtracking occurs consecutively within the sliding window size N (in example, N = 3 adjacent samples) or exceeds a proportion p (in example, p = 0.5) within the sliding window to be considered a valid trigger; a cooling-off period is set for the same spatial fold after each trigger.T cool (Example) T cool =24 hours or batch processing) to prevent jitter, and a maximum backtracking limit can be set (example 3 times / 24 hours). After triggering, the number of neighbors and distance weights of K-nearest neighbors and the hyperparameters of random forest are adjusted in sequence. If the conditions are still not met, it is marked as high uncertainty imputation and manual review is triggered.

[0059] In some embodiments, to ensure that the interpolation results satisfy the basic physical laws of the geothermal field, the present invention further introduces a physical consistency check after hybrid interpolation, and uses the check result as a backtracking trigger, specifically including the following check mechanism: 1) Gradient test Calculate the local temperature gradient of the data points after interpolation. The finite difference method can be used as an approximation. The calculation results are then compared with the regional geothermal gradient. G By comparison, if the following formula is satisfied, then it is determined that there is a physical anomaly at that point: in, The standard deviation of the regional geothermal gradient reflects its natural fluctuation range; For a preset multiple threshold (e.g.) m =3), which can be adjusted according to the regional geological characteristics.

[0060] 2) Conservation / Residual Test When the thermal conductivity is known or can be approximately estimated k Under these conditions, the consistency of the heat conduction equation is checked on the interpolation results. In the simplified form, the second-order difference residual is calculated. r : in, The second spatial derivative of temperature. For internal heat source items (if there is no internal heat source, then...) q =0). When the second-order difference residual is 0). r When the absolute value exceeds the preset threshold, it is judged as a physical anomaly, triggering a backtracking or manual review.

[0061] 3) Soft constraints and model training To further ensure the physical validity of the interpolation results, physical constraints can be introduced as regularization terms in the loss function during the offline model training phase. An example loss function is as follows: in, This refers to the data fitting loss (such as mean squared error). This is the physical residual term (either the gradient residual or the conserved residual can be selected). These are physical constraint weights used to balance the importance of data fit and physical consistency. The value of can be optimized through cross-validation, so that the model can fit the data as closely as possible to the physical laws.

[0062] 4) Triggering logic Perform the aforementioned physical consistency check on each interpolation point or spatial transition (e.g., a single well or geographic grid). Set a sliding window (e.g., N consecutive sample points) and count the number of physical consistency check failures within the window. If the number of failures exceeds a preset threshold (e.g., more than 50% of the sample points within the window are judged as physically abnormal), mark the point or spatial transition as a high-priority object for manual review and trigger a backtracking adjustment process to recalibrate or optimize the parameters of the interpolation method.

[0063] By using the aforementioned physical consistency verification mechanism, the constraints of geothermal physical laws can be incorporated into the data-driven approach, effectively identifying and correcting interpolation results that violate basic physical principles, thereby further improving the geological rationality and engineering usability of the processing results.

[0064] S104: Meta-information record.

[0065] For each interpolation, record the estimated value, uncertainty, weight, final interpolation value, and method parameters used for traceability and quality assessment.

[0066] To further ensure the geological rationality of the interpolation results, this invention introduces systematic geological constraints during the interpolation and fusion process, mainly including but not limited to the following: (1) Temperature gradient constraint Define a reasonable range for regional temperature gradients: in, and These are the lower and upper limits of the regional geothermal gradient, respectively, for example. =1.0℃ / 100m, =6.0℃ / 100m, which can be adjusted according to the specific geological region. If the interpolation result causes the local gradient to exceed this range, projection correction or soft penalty term can be used for correction.

[0067] (2) Thermal conductivity / heat capacity and lithological constraints Thermal conductivity is set for different lithological types. k and heat capacity c The physical reasonable range: k min ≤ k ≤ kmax , c min ≤ c ≤ c max in, k min , k max , c min , c max These represent the upper and lower bounds for thermal conductivity and heat capacity under the corresponding lithology. The prior model will provide prior estimates based on the above constraints, according to the lithology code of the sample point. x prior and its uncertainty was prior The initial estimate.

[0068] (3) Interlayer continuity and boundary conditions At the formation interface, the continuity of temperature and heat flux should be ensured (or the known boundary conditions should be met). For discontinuities in the interpolation results at the interface, corrections should be made through local equilibrium adjustments or forced constraints to ensure that the temperature and heat flux on both sides of the interface meet the physical continuity requirements.

[0069] (4) Porosity / permeability and pressure constraint When the heat transfer process is significantly affected by fluid convection or porous media, the physical reasonable range of parameters such as pressure P and permeability κ should be jointly verified, and if necessary, they should be used as joint constraints to ensure that the interpolation results are compatible with the basic laws of the seepage field.

[0070] (5) Constraints on regional heat flow uniformity When the regional heat flux value H When the temperature gradient obtained after interpolation is known or can be estimated, the thermal conductivity should satisfy the basic relationship of the heat conduction equation: q ≈ - k ·d T / d z in, q For heat flux density, a certain tolerance range should be allowed to ensure that the interpolation results are consistent with the regional heat flux background.

[0071] In some embodiments, to achieve the above-mentioned geological constraints, the following engineering methods are provided, which may be used alone or in combination: ① Hard-constrained projection After calculating the preliminary interpolation values x imputedThen, feasible region projection is performed for each well or each grid point. This method can be modeled as solving the following quadratic programming problem: in, and The coefficient matrix and boundary vectors are constructed based on geological constraints to ensure that linear inequalities or equality constraints (such as gradient range, thermal conductivity interval, etc.) are strictly satisfied. This method is applicable to real-time post-processing or offline consistency steps.

[0072] ② Soft constraint penalties During offline model training or online backtracking parameter tuning, a physical penalty term is introduced into the objective function, in the following form: in, This refers to the data fitting loss (such as mean squared error). It serves as a quantitative indicator of the degree of violation of physical constraints (such as gradient exceedance magnitude, interface discontinuity, heat flux deviation, etc.). The physical constraint weights are determined through cross-validation or calibration with historical data.

[0073] ③ Prior correction Geological constraints are incorporated as an informative prior into the Bayesian fusion framework. By adjusting the prior mean and prior variance, the fusion results automatically favor regions that conform to geological constraints while satisfying data fitting requirements, thereby influencing the fusion weights and final uncertainty.

[0074] ④ Hierarchical post-processing and local balancing Each well or spatial breakpoint undergoes specialized post-processing steps, including smoothing, interface rebalancing, and gradient correction. All correction operations are recorded in the metadata for traceability and manual review.

[0075] When implementing step S10 above, the lithology-parameter boundary table can be included as an appendix or implementation configuration (e.g., a JSON table) for use by the prior model and projection operator; "physical constraint violation" can be added as a triggering factor to the backtracking triggering rules, and a priority can be assigned to it for manual review of reasonable resource allocation; the weight of soft penalty terms can be adjusted during the offline calibration phase. Sensitivity analysis was performed on gradient thresholds and projection tolerances, and default values ​​and adjustment strategies were recorded. Combining geological constraints with existing uncertainty quantification, covariance degradation strategies, and physical consistency checks can significantly improve the geological rationality and engineering usability of interpolation results.

[0076] S20: Based on the comprehensive uncertainty, the local threshold used for anomaly detection is adaptively adjusted; the adjusted local threshold is used to detect anomalies in geothermal data, and the abnormal data is identified and marked.

[0077] Step S20 is used to implement adaptive anomaly detection. In some embodiments, such as Figure 3 As shown, step S20 can be implemented through the following steps S201 to S204.

[0078] S201: Uncertainty calculation.

[0079] Calculate the comprehensive uncertainty index for each sample : in Normalized data variance (reflecting measurement uncertainty) The normalized inter-model prediction inconsistency (reflecting model uncertainty). This is a normalized neighborhood sparsity statistic (reflecting data sparsity) obtained based on neighborhood density conversion. , , The weighting coefficients are and satisfy the following conditions: + + =1.

[0080] The uncertainty index u is related to the interpolation module. was imputed Forming information coupling: High was imputed The region typically corresponds to a high u value, which allows anomaly detection to automatically use a more lenient threshold in the interpolation uncertainty region, avoiding misjudging interpolation errors as anomalies.

[0081] S202: Parameter self-adaptation.

[0082] The local threshold for anomaly detection is dynamically adjusted based on the uncertainty index u, and the calculation method is as follows: in, The preset baseline threshold, The value ranges from 0.3 to 0.7. This is a preset adjustable coefficient. The value ranges from 0.2 to 0.5.

[0083] This adaptive mechanism is designed specifically for the characteristics of geothermal data: in geologically complex areas such as geothermal reservoir boundaries and fault zones, where data uncertainty is high and u-values ​​are large, the local threshold automatically increases to avoid misjudging geologically caused temperature anomalies as data errors; in geologically stable areas, where u-values ​​are small and the local thresholds are close to the baseline value, maintaining sensitivity to real anomalies.

[0084] S203: Anomaly detection.

[0085] The isolated forest algorithm was used to obtain and normalize the isolated scores. The reconstruction error is obtained using an autoencoder algorithm. According to fusion weight Calculate the final score : The value of λ ranges from 0.4 to 0.7.

[0086] Will and Compare the output anomaly labels and confidence scores; if the final anomaly score is... Greater than or equal to the local threshold If the condition is met, the sample is considered an outlier; otherwise, it is considered a normal sample.

[0087] S204: Spatial Fold Management and Retrospection.

[0088] Spatial folds are divided according to wells or geographic grids. Self-adaptive detection is performed on each fold. When the judgment error exceeds the limit, parameter recalibration or local model retraining is triggered.

[0089] This application also provides a system for multi-source geothermal data fusion and intelligent processing based on uncertainty perception, such as... Figure 4As shown, the system includes a data access module 401, a data parsing module 402, a preprocessing module 403, an anomaly detection module 404, a hybrid interpolation module 405, a data fusion module 406, a quality assessment module 407, and a storage and service module 408. The data access module 401 is used to access multi-source heterogeneous geothermal data, supporting CSV, Excel, LAS, DLIS, JSON, and XML data formats. The data parsing module 402 is used to parse data of different formats, performing timestamp standardization, coordinate system unification, and depth benchmark alignment. The preprocessing module 403 is used to perform noise filtering, unit conversion, and format standardization operations. The anomaly detection module 404 is used to implement the aforementioned adaptive anomaly detection method (step S20), identifying and marking abnormal data. The hybrid interpolation module 405 is used to implement the aforementioned uncertainty-driven hybrid interpolation method (step S10), filling in missing data. The data fusion module 406 is used to fuse multi-source data based on an uncertainty weighting strategy. The quality assessment module 407 is used to generate traceable processing logs and calculate data quality indicators such as completeness, accuracy, consistency, timeliness, and uncertainty scores. The storage and service module 408 is used to store processed data and metadata, supports mixed storage of relational databases and time-series databases, and provides data query and export services.

[0090] Based on the above system, this embodiment uses the missing temperature data at 300m (near the top interface of the thermal reservoir) in a certain exploration well depth sequence as an example to illustrate the processing flow of the hybrid interpolation method. This location exhibits abrupt temperature gradient changes due to the thermal reservoir boundary effect, making it a typical area where traditional interpolation methods have significant errors. To enhance the realism of the embodiment, the following provides actual operating data for a local depth window near the missing point, which can be directly used to draw depth-temperature curves, confidence interval strip plots of the interpolation point, or comparison charts before and after interpolation.

[0091] Table 1. Local depth window operation data near the 300m missing point

[0092] As shown in Table 1, the interpolation result of 46.4℃ at 300m is between the measured value of 46.1℃ at 299m and the measured value of 46.9℃ at 301m, maintaining the heating trend near the top interface of the thermal reservoir without any sudden jumps contrary to the local geothermal gradient. If estimated according to the 95% confidence interval, the temperature at 300m can be expressed as 46.4±1.31℃, corresponding to a range of approximately 45.1℃ to 47.7℃.

[0093] The hybrid interpolation module 405 is configured to perform the following steps S1 to S4.

[0094] Step S1, multiple estimation generation.

[0095] K-nearest neighbor estimation is performed, with the number of neighbors set to k=5. The five nearest known points in the depth-temperature feature space are found, with temperatures of 45.2℃, 47.8℃, 46.5℃, 48.1℃, and 45.9℃ respectively. The distance-weighted average is then calculated. x knn =46.7℃, the neighborhood variance was calculated to obtain was knn = 1.42℃ 2 .

[0096] Random forest estimation was performed using a random forest model containing 100 decision trees. Input features included depth, lithology coding, and regional geothermal gradient. The ensemble prediction values ​​were then calculated. x rf = 46.5℃, the variance between trees was calculated to obtain was rf =0.89℃ 2 .

[0097] Conduct a priori estimation of geothermal geology based on the geological characteristics of the area (geothermal gradient). G = 3.2℃ / 100m, Lithosphere thickness L = 85km, burial depth D = 300m, regional heat flux value H = 70mW / m 2 Substituting these values ​​into the regional regression model yields the prior mean. x prior = 45.8℃. Since this point is located near the top interface of the thermal reservoir, the temperature gradient is uncertain, and the prior variance... was prior = 2.5℃ 2 (1.5℃ higher than the typical value in the stable region) 2 ).

[0098] Step S2, uncertainty fusion.

[0099] Since K-nearest neighbors and random forests use the same neighborhood data and are correlated, while geological priors are based on independent regional statistics, theoretically, linear fusion with covariance perception should be adopted. However, in this embodiment, the effective historical sample size for stable estimation of the covariance matrix is ​​insufficient, triggering a diagonal approximation degradation strategy. Therefore, inverse variance weighting is used for fusion.

[0100] Calculate the weights of each estimate, taking ε = 10. -6 , w knn = 1 / 1.42 = 0.704, w rf = 1 / 0.89 = 1.124, wprior = 1 / 2.5 = 0.400, the sum of weights Σ w = 2.228.

[0101] Computational fusion results x imputed = (0.704×46.7 + 1.124×46.5 + 0.400×45.8) / 2.228 = 46.4℃ was imputed = 1 / 2.228 = 0.449℃ 2 .

[0102] Table 2 Summary of 300m Missing Point Fusion Log

[0103] Furthermore, the local depth window data in Table 1 can be displayed together with the fusion log in Table 2: when plotting the local temperature curve with depth as the x-axis and temperature as the y-axis, the 300m interpolation point falls near the line connecting adjacent measured points; when plotting the weighted bar chart with candidate sources as the x-axis, it can be intuitively shown that the random forest estimation contributed the most in this fusion, while the geological prior mainly played a role in constraint and correction.

[0104] Analysis of fusion results: Due to the high uncertainty of geological priors ( was prior = 2.5), with a relatively small weight (0.400 / 2.228 = 18%), the fusion result is more inclined towards data-driven estimation. In sparse data regions, the uncertainty of K-nearest neighbors and random forest increases, and the weight of geological prior will automatically increase, reflecting the characteristics of adaptive fusion.

[0105] Step S3, adaptive backtracking.

[0106] Set threshold τ var = 3.0℃ 2 (Example value, calibrated via cross-validation or historical residual distribution), γ = 2.0. Test the first condition. was imputed = 0.449< τ var = 3.0, meets the requirement; check the second condition | x imputed - x prior | = 0.64< The result is approximately 3.43, which meets the requirements. Since both conditions are met, there is no need to trigger backtracking, and the interpolation result is valid.

[0107] Step S4: Record metadata.

[0108] The interpolation metadata includes data point identifiers, processing timestamps, estimated values ​​and variances, weight values, and the final interpolated value of 46.4℃ with a variance of 0.449℃. 2 Geological location label (top interface of the thermal reservoir), processing status is normal. Example fields include: timestamp (processing time), estimator_list, estimator_vars, weights, final_value, final_var, hyperparams, location_tag, review_flag.

[0109] This embodiment uses an anomaly detection module 404 to detect anomalies at 100 temperature measurement points of a geothermal well. The well traverses a concealed fault zone, near which a heat flow channel exists, resulting in a temperature distribution significantly different from the surrounding area. This is a typical scenario where traditional fixed threshold methods are prone to false alarms. The anomaly detection module 404 is configured to execute the following steps T1~T4.

[0110] Step T1: Uncertainty calculation.

[0111] For each measurement point, an uncertainty index is calculated, and a weighting coefficient is set. w 1 = 0.4 w 2 = 0.3 w 3 = 0.3, calculate u = 0.4 × was norm + 0.3× inconsistency norm + 0.3× density norm .

[0112] In the region near the fault zone, due to abnormal temperature gradients, the inconsistency between model predictions is high. inconsistency norm ≈ 0.7), while the number of data points in this area is relatively small ( density norm ≈ 0.6), and we get u ≈0.4×0.5 + 0.3×0.7 + 0.3×0.6 = 0.59.

[0113] In geologically stable regions, all indicators are low, and the calculated value is u ≈ 0.25.

[0114] Step T2, parameter self-adaptation.

[0115] Set baseline threshold threshold base = 0.5, adjustable coefficient β = 0.3.

[0116] For the high uncertainty region near the fault zone (u = 0.59), calculate the local threshold. threshold local =0.5 × (1 + 0.3 × 0.59) = 0.59. Raising the threshold ensures that geologically occurring temperature anomalies in the region will not be misjudged as data errors.

[0117] For the geologically stable region (u = 0.25), calculate the local threshold. threshold local = 0.5 × (1 + 0.3 × 0.25) = 0.54. The threshold is close to the baseline value, maintaining sensitivity to errors in the actual data.

[0118] Step T3, anomaly detection.

[0119] When training an isolation forest model, the number of estimators is set to 100, and an isolation score is calculated for each sample. s if An autoencoder network was constructed with an encoder structure of 64-32-16 and a decoder structure of 16-32-64. The reconstruction error for each sample was calculated. e ae Then, normalization was performed. The fusion weight λ = 0.6 was set, and the final score was calculated. score final = 0.6 × s if + 0.4 × e ae ,Will score final and threshold local Compare the output anomaly labels and confidence levels.

[0120] At a certain measurement point near the fault zone score final = 0.55, because threshold local = 0.59, this point is considered normal, avoiding false alarms. If a fixed threshold of 0.5 is used, this point will be falsely judged as abnormal.

[0121] Step T4: Space management and backtracking.

[0122] The data is divided into spatial folds according to hash numbers, and the above detection process is performed on each fold. When manual review finds that the detection error of a fold exceeds 10%, the β coefficient of that fold is backtracked and adjusted. If the requirements are still not met after adjustment, the isolated forest and autoencoder of that fold are locally retrained.

[0123] The system is deployed in a layered architecture. The real-time lightweight layer is responsible for low-latency pre-screening and online interpolation, while the offline heavyweight layer is responsible for model training and accurate uncertainty estimation. The offline results are periodically sent to the real-time layer to update the strategy.

[0124] During fusion computation, the real-time layer prioritizes shrinking covariance or diagonal approximation to ensure numerical stability. When the condition number of the covariance matrix exceeds a preset threshold or the effective sample size falls below a preset threshold, a degradation strategy is triggered, employing prior estimation, a single estimator output, or manual verification. To ensure the numerical stability of the real-time layer, the Ledoit–Wolf shrinkage method (Ledoit & Wolf, 2004) can be used to shrink the covariance. The shrinkage parameter can be adaptively selected according to the sample size, and a default shrinkage strength (example value 0.1) can be set for the real-time layer, calibrated using historical data in the offline layer.

[0125] The quality assessment module generates a processing log for each data record and calculates data quality indicators including completeness, accuracy, consistency, timeliness, and uncertainty scores. The system establishes an online monitoring and rollback mechanism to monitor key indicators and roll back to the most recent stable model when limits are exceeded. High-value samples are selected for manual review based on uncertainty indicators to reduce overall review costs.

[0126] To further verify the effectiveness of the method proposed in this application, a comparative experiment was designed. Scheme A is a traditional method using single K-nearest neighbor interpolation combined with fixed threshold anomaly detection; Scheme B is an improved method using single random forest interpolation combined with isolated forest anomaly detection; and Scheme C is the method proposed in this application.

[0127] The comparative experiments used 5-fold cross-validation; the results are expressed as mean ± standard error. If necessary, t-tests or other statistical tests (e.g., p < 0.05 is considered statistically significant) were used to verify the significance of differences between methods. The specific experimental division and random seed are described in the appendix or examples.

[0128] Interpolation accuracy comparison results: For temperature data, the root mean square error (RMSE) of scheme A is 2.34℃, scheme B is 2.12℃, and scheme C is 1.89℃. Scheme C is 19.2% lower than scheme A and 10.8% lower than scheme B. For pressure data, the RMSE of scheme A is 0.45MPa, scheme B is 0.41MPa, and scheme C is 0.36MPa. For permeability data, the RMSE of scheme A is 15.2mD, scheme B is 13.8mD, and scheme C is 12.1mD.

[0129] Regional interpolation accuracy analysis: In the boundary region of the thermal reservoir (temperature gradient abrupt change zone), the root mean square error of scheme A is 3.21℃, scheme B is 2.87℃, and scheme C is 2.15℃. Scheme C has a more obvious advantage (33% lower than scheme A), indicating that the introduction of geothermal geology priors has a significant effect in geologically complex areas.

[0130] Anomaly detection performance comparison results: Solution A has a precision of 0.68, a recall of 0.71, and an F1 score of 0.69; Solution B has a precision of 0.74, a recall of 0.70, and an F1 score of 0.72; and Solution C has a precision of 0.86, a recall of 0.83, and an F1 score of 0.84.

[0131] Regional anomaly detection analysis: In the fault zone heat flow channel area, the false alarm rate of scheme A is 35%, scheme B is 28%, and scheme C is 12%. The adaptive threshold mechanism significantly reduces the false alarm rate of scheme C in geologically complex areas.

[0132] Downstream task performance comparison results: Taking reservoir thermal conductivity prediction as the downstream task, the coefficient of determination of scheme A on the test set is 0.71, scheme B is 0.74, and scheme C is 0.82.

[0133] Ablation Experiment: To verify the contributions of each technical module, an ablation experiment was conducted. After removing the geological prior module, the temperature interpolation RMSE increased from 1.89℃ to 2.05℃ (+8.5%); after removing the adaptive threshold module, the anomaly detection F1 decreased from 0.84 to 0.76 (-9.5%); and after removing the backtracking mechanism, the interpolation RMSE in high uncertainty regions increased from 2.15℃ to 2.68℃ (+24.7%). The ablation experiment shows that each technical module contributes independently, and the contribution is more significant in geologically complex regions.

[0134] The experimental results above show that the method of this application is significantly better than existing methods in terms of interpolation accuracy, anomaly detection performance and downstream task performance, especially in geologically complex areas such as thermal reservoir boundaries and fault zones.

[0135] This application also provides a device for multi-source geothermal data fusion and intelligent processing based on uncertainty perception, such as... Figure 5 As shown, the multi-source geothermal data fusion and intelligent processing device based on uncertainty perception includes: The hybrid interpolation unit 501 is configured to perform uncertainty-driven hybrid interpolation on geothermal data to generate interpolated data and the corresponding comprehensive uncertainty. Anomaly detection unit 502 is configured to adaptively adjust the local threshold for anomaly detection based on the comprehensive uncertainty; and to use the adjusted local threshold to perform anomaly detection on geothermal data, identify and mark the anomalous data.

[0136] This application provides an electronic device. The electronic device may include a processor and a memory, wherein the processor and the memory can communicate; exemplarily, the processor and the memory communicate via a communication bus.

[0137] The processor executes computer execution instructions stored in memory, causing the processor to perform the scheme in the above embodiments. The processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0138] The communication bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The system bus can be divided into address bus, data bus, control bus, etc. Transceivers are used to enable communication between database access devices and other computers (e.g., clients, read-write libraries, and read-only libraries). Memory may include random access memory (RAM) and may also include non-volatile memory.

[0139] The electronic device provided in this application embodiment can be the terminal device described in the above embodiments.

[0140] This application also provides a computer-readable storage medium storing computer instructions. When the computer instructions are executed on a computer, the computer performs the technical solution of the multi-source geothermal data fusion and intelligent processing method based on uncertainty perception described in the above embodiments.

[0141] This application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium. When the at least one processor executes the computer program, it can implement the technical solution of the multi-source geothermal data fusion and intelligent processing method based on uncertainty perception in the above embodiments.

[0142] In the several embodiments provided in this application, 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 instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0143] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to implement the solution of this embodiment according to actual needs.

[0144] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The unit composed of the above modules can be implemented in hardware or in the form of hardware plus software functional units.

[0145] The integrated modules described above, implemented as software functional modules, can be stored in a computer-readable storage medium. These software functional modules, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods of the various embodiments of this application.

[0146] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly manifested as execution by a hardware processor, or execution by a combination of hardware and software modules within the processor.

[0147] The memory may include high-speed RAM, and may also include non-volatile storage (NVM), such as at least one disk storage device, and may also be a USB flash drive, external hard drive, read-only memory, disk or optical disc, etc.

[0148] Buses can be Industry Standard Architecture (ISA) buses, Peripheral Component Interconnect (PCI) buses, or Extended Industry Standard Architecture (EISA) buses, etc. Buses can be categorized into address buses, data buses, control buses, etc.

[0149] The aforementioned storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0150] An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Alternatively, the storage medium can be an integral part of the processor. The processor and storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and storage medium can exist as discrete components in an electronic control unit or main control device.

[0151] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for multi-source geothermal data fusion and intelligent processing based on uncertainty perception, characterized in that, The method includes: Uncertainty-driven hybrid interpolation is performed on geothermal data to generate interpolated data and the corresponding comprehensive uncertainty; Based on the comprehensive uncertainty, the local threshold used for anomaly detection is adaptively adjusted; the adjusted local threshold is used to detect anomalies in geothermal data, and the abnormal data is identified and marked.

2. The method for multi-source geothermal data fusion and intelligent processing based on uncertainty perception according to claim 1, characterized in that, Uncertainty-driven hybrid interpolation is performed on geothermal data to generate interpolated data and corresponding comprehensive uncertainties, including: For each data point to be interpolated, multiple estimators are used to generate candidate estimates and their corresponding uncertainties. Based on the candidate estimates and their uncertainties, the final interpolated values ​​and their combined uncertainties are calculated using a linear fusion strategy that employs inverse variance weighting or covariance sensing.

3. The method for multi-source geothermal data fusion and intelligent processing based on uncertainty perception according to claim 2, characterized in that, When using inverse variance weighted fusion, the weights are calculated using the following formula. Final interpolation value and comprehensive uncertainty : in, For the first i One estimate The corresponding uncertainty, To prevent division by zero of small constants; When using covariance-sensing linear fusion, the fusion value is calculated using the following formula. and uncertainty : in, To estimate the covariance matrix of the instrument, It is a unit vector.

4. The method for multi-source geothermal data fusion and intelligent processing based on uncertainty perception according to claim 2, characterized in that, The various estimators include a prior model based on geothermal geology, which generates candidate estimates. Determined by the following formula: in, For geothermal gradient, For the thickness of the lithosphere, For burial depth, This represents the regional heat flux value. , , , , and These are coefficients determined based on regional geostatistical regression.

5. The method for multi-source geothermal data fusion and intelligent processing based on uncertainty perception according to claim 1, characterized in that, Based on the comprehensive uncertainty, the local threshold used for anomaly detection is adaptively adjusted; Anomaly detection of geothermal data is performed using the adjusted local threshold, and abnormal data is identified and marked, including: A comprehensive uncertainty index u is calculated for each sample, and the comprehensive uncertainty index is informationally coupled with the comprehensive uncertainty. Based on the comprehensive uncertainty index u, dynamically adjust the local threshold; The final anomaly score is obtained by integrating the outputs of multiple anomaly detection algorithms, and the final anomaly score is compared with an adjusted local threshold to determine an anomaly.

6. The method for multi-source geothermal data fusion and intelligent processing based on uncertainty perception according to claim 5, characterized in that, The formula for calculating the uncertainty index u is: in, For normalized data variance, To normalize the inconsistency in predictions between models, This is a normalized neighborhood sparsity statistic obtained based on neighborhood density conversion. , , The weighting coefficients are satisfied. + + =1; Based on the comprehensive uncertainty index u, the local threshold is dynamically adjusted using the following formula. : in, The preset baseline threshold, This is a preset adjustable coefficient; The formula for calculating the final anomaly score is as follows: in, These are the normalized isolated forest algorithm isolated scores. The reconstruction error is obtained by the autoencoder algorithm. To integrate weights, This is the normalization function; The specific method for comparing the final anomaly score with the adjusted local threshold to determine anomalies is as follows: if the final anomaly score is greater than or equal to the local threshold, the sample is determined to be an anomaly; otherwise, it is determined to be a normal sample.

7. The method for multi-source geothermal data fusion and intelligent processing based on uncertainty perception according to any one of claims 1 to 6, characterized in that, After performing uncertainty-driven hybrid interpolation on geothermal data to generate interpolated data and corresponding comprehensive uncertainty, an adaptive backtracking step is also included. The adaptive backtracking step includes: When the overall uncertainty exceeds a preset threshold τ var If the deviation between the interpolated value and the prior value obtained based on the geothermal geology prior model exceeds a preset range, backtracking is triggered to adjust the parameters of the interpolation method or initiate manual review. The preset range is determined by the following formula: in, For preset coefficients, To account for the overall uncertainty, The prior uncertainty corresponding to the prior model; Based on the comprehensive uncertainty, the local threshold used for anomaly detection is adaptively adjusted; after using the adjusted local threshold to detect anomalies in geothermal data, identifying and marking anomalous data, a spatial folding management and backtracking step is also included; the spatial folding management and backtracking step includes: Spatial layers are divided according to wells or geographic grids. Anomaly detection is performed on each spatial layer. When the detection error exceeds the limit, the parameters of the corresponding spatial layer are recalibrated or the model is locally retrained.

8. A device for multi-source geothermal data fusion and intelligent processing based on uncertainty perception, characterized in that, The device includes: The hybrid interpolation unit is configured to perform uncertainty-driven hybrid interpolation on geothermal data, generating interpolated data and the corresponding comprehensive uncertainty; An anomaly detection unit is configured to adaptively adjust a local threshold for anomaly detection based on the comprehensive uncertainty; and to use the adjusted local threshold to detect anomalies in geothermal data, identify and mark the anomalous data.

9. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes the computer execution instructions stored in the memory to implement the multi-source geothermal data fusion and intelligent processing method based on uncertainty perception as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the multi-source geothermal data fusion and intelligent processing method based on uncertainty perception as described in any one of claims 1-7.

Images