Deformation monitoring method for pumped storage dams based on Beidou positioning

Abstract

The present invention relates to the technical field of real-time correction of dynamic deformation measurement errors based on Beidou satellite navigation, and in particular to a method for monitoring deformation of pumped-storage dams based on Beidou positioning. This method addresses the technical problem of pseudo-fluctuations in monitoring data caused by the frequency domain aliasing of Beidou signals' multipath time-varying errors and the structural real displacement in a high-dynamic environment. Millimeter-level real-time solution is achieved through the construction of a multimodal collaborative monitoring network and dynamic feature analysis. The method includes: deploying a dual-frequency Beidou receiver array and environmental sensors, integrating Beidou observation data, reservoir water pressure gradients, foundation vibration spectra, and three-dimensional geological structure data, constructing a robust solution model constrained by a geological model, outputting confidence intervals for deformation parameters, and suppressing outliers through robust estimation; employing a dynamic time warping algorithm to match a historical operating condition library, combining a closed-loop feedback mechanism to dynamically calibrate warning thresholds and reversely optimize the monitoring network. The present invention effectively suppresses multipath time-varying errors.

Description

Technical Field

[0001] The present invention relates to the technical field of real-time correction of dynamic deformation measurement errors based on Beidou satellite navigation, and in particular to a method for monitoring deformation of a pumped storage dam based on Beidou positioning. Background Art

[0002] Beidou-based pumped-storage dam deformation monitoring technology uses the high-precision Beidou Navigation Satellite System (BDS) to acquire real-time three-dimensional coordinate data of the dam surface and key structural points. Combined with multi-frequency signal processing and carrier phase differential technology, it achieves millimeter-level deformation monitoring accuracy. This system utilizes a network of Beidou base stations and monitoring stations, and employs Kalman filtering and wavelet analysis algorithms to reduce noise from raw observation data, eliminating interference from error sources such as ionospheric delay, tropospheric refraction, and multipath effects. It dynamically calculates the dam's horizontal displacement, vertical settlement, and deflection changes under the influence of reservoir water pressure, temperature loads, and geological tectonic stresses. Monitoring data is compared with finite element simulation results using an adaptive threshold model to quantitatively assess dam structural stability and provide early warning of localized stress concentrations or material creep trends.

[0003] The Beidou-based dam deformation monitoring system faces a conflict between multipath residual errors and real-time data processing delays in highly dynamic environments. During the operation of a pumped-storage power station, water vapor evaporation around the dam and the rapid rise and fall of reservoir water levels trigger non-uniform mutations in the signal propagation medium, causing the Beidou signal multipath error to exhibit time-varying nonlinear characteristics. Existing wavelet threshold denoising algorithms are insufficiently adaptable to transient interference, which can easily lead to lags and oscillations in the deformation solution results. For example, when a sudden drop in reservoir water level causes instantaneous reverse deformation of the dam's water-facing surface, multipath errors and the structure's true displacement alias in the frequency domain. Conventional differential models struggle to effectively separate the two signal components within a sampling period of seconds, resulting in pseudo-fluctuations of 5-8 mm in the horizontal displacement output by the monitoring system, affecting the real-time determination of the structural safety status. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a pumped-storage dam deformation monitoring method based on Beidou positioning. The present invention solves the technical problem that the multi-path time-varying errors in the Beidou deformation monitoring signal and the real displacement components of the structure are aliased in the frequency domain in a high-dynamic environment, resulting in pseudo-fluctuations in the monitoring data, making it difficult to achieve millimeter-level real-time solution accuracy.

[0005] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows:

[0006] The method for monitoring deformation of pumped storage dams based on Beidou positioning includes:

[0007] Step 1: Receive BeiDou observation data of the dam surface and surrounding bedrock, and simultaneously collect environmental parameters and geological structure data to generate a dam monitoring dataset;

[0008] Step 2: Input the geological structure data and historical deformation information in the dam monitoring data set into a genetic optimization algorithm to construct a monitoring network topology that is resistant to dynamic interference;

[0009] Step 3: Perform dynamic feature analysis on the BeiDou observation data received in step 1 to extract the coupling feature map of multipath error and structural deformation;

[0010] Step 4: Input the coupling characteristic spectrum into a preset mixed-domain signal decomposition model, adjust the decomposition parameters through frequency domain energy distribution, and separate the low-frequency structural displacement main mode and the high-frequency multipath interference submode;

[0011] Step 5: construct a geological model based on the geological structure data in step 1, embed a robust solution model with physical constraints in the geological model, and output confidence intervals of deformation parameters;

[0012] Step 6: According to the matching degree between the deformation parameter confidence interval output in step 5 and the historical operating condition library, the parameter threshold of the early warning model is dynamically adjusted.

[0013] Furthermore, in the method for monitoring deformation of a pumped-storage dam based on Beidou positioning according to the present invention, the dam monitoring dataset in step 1 includes:

[0014] A dual-frequency Beidou receiver array is deployed on the dam surface to obtain carrier phase observations and satellite ephemeris in real time to generate Beidou observation data.

[0015] Reservoir water level pressure sensors and three-dimensional vibration sensors are deployed along the dam axis to synchronously record water pressure gradients and foundation micro-vibration spectra to generate environmental parameter data;

[0016] The geological radar detection data and the mechanical parameters of the drill core are aligned with the 3D model of the dam to generate geological structure data including the distribution of fault zones;

[0017] The Beidou observation data, environmental parameter data and geological structure data are integrated according to a unified time and space reference to obtain a dam monitoring data set.

[0018] Furthermore, in the method for monitoring deformation of a pumped storage dam based on Beidou positioning according to the present invention, the step 2 of constructing a monitoring network topology structure includes:

[0019] Based on the geological structure data and historical deformation information generated in step 1, the coordinates of the fault zone deformation sensitive area are extracted;

[0020] Generating initial site layout data based on the coordinates of the fault zone deformation sensitive area and satellite visibility prediction results;

[0021] The multipath suppression rate and baseline solution accuracy are used as fitness functions, and the optimal site set is iteratively selected through a genetic algorithm.

[0022] The foundation micro-vibration spectrum and temperature and humidity data in the dam monitoring data set are associated with the monitoring station data stream of the optimal station set, a spatiotemporal index table is established, and the monitoring network topology is updated.

[0023] Furthermore, in the method for monitoring deformation of a pumped storage dam based on Beidou positioning according to the present invention, the dynamic characteristic analysis in step 3 includes:

[0024] Implement a cycle slip detection algorithm on the carrier phase of the BeiDou observation data received in step 1, mark abnormal observation points, and generate a corrected phase sequence;

[0025] The reservoir water level change rate in the dam monitoring dataset, the phase sequence corrected in step 3, and the monitoring network topology generated in step 2 are input into the spatiotemporal convolutional network, and a dynamic interference feature map is generated through data fusion.

[0026] Furthermore, in the method for monitoring deformation of a pumped storage dam based on Beidou positioning according to the present invention, the mixed domain signal decomposition model in step 4 includes:

[0027] Based on the frequency domain energy distribution of the dynamic interference characteristic spectrum generated in step 3, the penalty factor of the variational mode decomposition is dynamically adjusted;

[0028] The corrected phase sequence in step 3 is input into the decomposition module to separate the low-frequency structural displacement main mode and the high-frequency multipath interference submode. The high-frequency submode carries the time-varying error label marked in step 3 and is output to the anti-error solution model in step 5.

[0029] Furthermore, in the method for monitoring deformation of a pumped storage dam based on Beidou positioning according to the present invention, the robustness calculation model in step 5 includes:

[0030] Based on the bedrock elastic parameters of the geological model constructed in step 5, a state transfer equation is generated and a temperature strain compensation term is embedded;

[0031] Input the low-frequency structural displacement main mode separated in step 4 into the robustness calculation model, and output the confidence interval of the deformation parameter;

[0032] When the degree of deviation between the confidence interval of the deformation parameter and the finite element simulation result exceeds a preset threshold, the robust estimator is activated to suppress outliers and the confidence interval is updated.

[0033] Furthermore, in the method for monitoring deformation of a pumped storage dam based on Beidou positioning according to the present invention, the dynamic adjustment of the early warning model in step 6 includes:

[0034] Input the confidence interval of the deformation parameters output in step 5 into the improved dynamic time warping algorithm to match the current deformation mode with similar working conditions in the historical working condition library;

[0035] According to the credibility score output by the robust solution model in step 5, a dynamic regularization constraint is applied to the weight matrix of the early warning model;

[0036] The adjusted warning parameter threshold is transmitted back to the genetic optimization algorithm in step 2 to update the fitness function of the monitoring network topology structure.

[0037] Beneficial effects of the present invention:

[0038] The present invention deploys a dual-frequency Beidou receiver array and environmental sensors, integrates Beidou observation data, reservoir water pressure gradient, foundation vibration spectrum and three-dimensional geological structure data, constructs an anti-error solution model constrained by the geological model, and combines the mixed domain signal decomposition technology to separate the low-frequency structural displacement main mode and the high-frequency multipath interference submode, effectively suppressing the pseudo-fluctuation of monitoring data caused by frequency domain aliasing; uses genetic algorithms to optimize the monitoring network topology, combines satellite visibility prediction and multipath suppression rate to dynamically screen the site locations, and reduces the dynamic interference of the signal propagation path; embeds temperature strain compensation terms through the state transition equation of physical constraints to correct the influence of environmental factors on deformation solution, combines the robust estimator to suppress outliers and output the confidence interval of deformation parameters; adopts dynamic time warping algorithm to match the historical working condition library, dynamically calibrates the warning threshold and reversely optimizes the fitness function of the monitoring network to form a closed-loop feedback mechanism, thereby improving the real-time performance and solution accuracy of millimeter-level deformation monitoring in high dynamic environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on the drawings without paying any creative labor.

[0040] Figure 1 This is a flowchart of a method for monitoring deformation of a pumped-storage dam based on Beidou positioning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. The technical solutions provided by each embodiment of the present invention are described in detail below in conjunction with the drawings. In order to better understand the purpose of the present invention, the present invention is further described in detail below.

[0042] See also Figure 1 The method for monitoring deformation of pumped storage dams based on Beidou positioning includes:

[0043] Step 1: Receive BeiDou observation data of the dam surface and surrounding bedrock, and simultaneously collect environmental parameters and geological structure data to generate a dam monitoring dataset;

[0044] In step 1, a dual-frequency Beidou receiver array is deployed at key points on the dam surface and surrounding bedrock to collect carrier phase observations and satellite ephemeris data in real time. The receiver has a built-in anti-multipath choke antenna to suppress signal reflection interference; a reservoir water level pressure sensor array is evenly arranged along the axis of the dam body, and a three-dimensional vibration sensor is used to synchronously record the water pressure gradient and foundation micro-vibration spectrum to generate environmental parameter data; a geological radar is used to detect the spatial distribution of the fault zone, and the bedrock elastic modulus and shear strength parameters are obtained through core drilling tests. The geological exploration data are aligned with the three-dimensional model of the dam to form geological structure data including the mechanical properties of the fault zone; the spatial and temporal references of the Beidou observation data, environmental parameter data and geological structure data are unified by the GNSS disciplined atomic clock, and the interpolation algorithm is used to fill the data missing areas to generate a spatiotemporally continuous dam monitoring data set, providing standardized input for subsequent deformation analysis. In each sub-step, Beidou observation data builds a deformation monitoring foundation through carrier phase difference technology, environmental parameter data captures the load effect of dynamic changes in reservoir water level on the dam body, geological structure data characterizes the spatial mechanical characteristics of fault zone risk areas, and the data fusion module eliminates data heterogeneity through spatiotemporal alignment and coordinate registration, forming a multi-dimensional collaborative monitoring data base.

[0045] Step 2: Input the geological structure data and historical deformation information in the dam monitoring data set into a genetic optimization algorithm to construct a monitoring network topology that is resistant to dynamic interference;

[0046] In step 2, based on the geological structure data in the dam monitoring dataset, the coordinate point set of the fault zone deformation sensitive area is extracted by principal component analysis, and the high-risk monitoring points are screened by combining the displacement gradient distribution under different working conditions in the historical deformation database; according to the fault zone coordinates and Beidou satellite visibility prediction results, the signal propagation path is simulated by a three-dimensional ray tracing algorithm, the areas with high incidence of multipath effects are eliminated, and the initial monitoring station layout plan is generated; the multipath suppression rate, baseline solution accuracy and signal redundancy are used as the fitness function of the genetic algorithm, the fault zone avoidance constraint is introduced by the crossover operator to avoid the layout of stations in mechanically weak areas, the mutation operator is used to dynamically adjust the station spacing to adapt to the change of satellite elevation angle, and the optimal layout set resistant to dynamic interference is output after iterative optimization; the foundation micro-vibration spectrum and temperature and humidity data are linked to the monitoring station data stream through timestamps, and a spatiotemporal index table is established to record the spectrum amplitude, temperature and humidity curve and spatiotemporal correlation of the Beidou observation sequence of each station to drive the dynamic update of the monitoring network topology. In each sub-step, the screening of deformation-sensitive areas provides a risk positioning basis for network optimization; satellite visibility prediction and path simulation synergistically suppress signal propagation interference; the genetic algorithm balances monitoring accuracy and anti-interference capability through multi-objective optimization; and the spatiotemporal index table realizes the logical coupling of environmental dynamic factors and the monitoring network through data association, forming a topological closed-loop optimization link that is resistant to dynamic interference.

[0047] Step 3: Perform dynamic feature analysis on the BeiDou observation data received in step 1 to extract the coupling feature map of multipath error and structural deformation;

[0048] In step 3, the TurboEdit cycle slip detection algorithm is applied to the carrier phase sequence in the BeiDou observation data received in step 1. Anomalous observation points are marked by combining carrier phase and pseudorange observations, eliminating cycle slips and gross errors, and generating a smoothed and corrected phase sequence. The corrected phase sequence uses double-difference processing to eliminate receiver clock errors and satellite orbit errors, forming a basic data source for high-precision deformation monitoring. The reservoir water level change rate data collected in step 1, the corrected phase sequence, and the monitoring network topology generated in step 2 are input into a spatiotemporal convolutional network. The first branch performs a temporal convolution operation on the reservoir water level change rate to extract the dynamic characteristics of the hydraulic load. The second branch performs a spatial convolution operation on the corrected phase sequence to capture the spatial correlation of deformation. The third branch constructs a graph convolution layer based on the topological relationship of the monitoring stations to explore the coupling pattern of multipath interference and structural deformation. Multi-branch features are weighted and fused through an attention mechanism to generate a dynamic interference feature map containing three-dimensional coupling features in time, space, and frequency. This map quantitatively represents the superposition relationship between the multipath error distribution and the structural deformation trend. In each sub-step, cycle slip detection and double-difference processing improve the credibility of Beidou observation data. The spatiotemporal convolutional network realizes the heterogeneous feature fusion of reservoir water pressure, Beidou deformation and site topology through multi-branch collaborative analysis. The dynamic interference feature map provides the basis for separating the frequency domain aliasing features for the subsequent signal decomposition module, supporting the precise decoupling of high-frequency interference and low-frequency deformation components.

[0049] Step 4: Input the coupling characteristic spectrum into a preset mixed-domain signal decomposition model, adjust the decomposition parameters through frequency domain energy distribution, and separate the low-frequency structural displacement main mode and the high-frequency multipath interference submode;

[0050] In step 4, based on the dynamic interference characteristic map generated in step 3, the frequency domain energy distribution parameters are extracted, the penalty factor of the variational modal decomposition is dynamically optimized by the genetic algorithm, the parameters are adjusted according to the bandwidth and energy concentration of the multipath interference submode, and the spectrum aliasing of the low-frequency deformation signal by high-frequency noise is suppressed; the phase sequence corrected in step 3 is input into the variational modal decomposition module, the initial number of modes is set to the expected frequency band range of the main mode of structural displacement and the submode of multipath interference, the optimal modal component is iteratively solved by the alternating direction multiplier method, and the low-frequency main mode of structural displacement and the high-frequency multipath interference submode are separated; the high-frequency submode is labeled with a time-varying error label through time-frequency analysis, and the label contains interference intensity and timestamp information, and is synchronously transmitted to the anti-error solution model in step 5 for dynamic correction of residual errors. In each sub-step, frequency domain energy distribution analysis provides a physical basis for decomposition parameter optimization. Variational modal decomposition achieves accurate separation of primary and secondary modes of the signal through adaptive frequency band segmentation. The generation and transmission of error labels constructs a closed-loop correction link for multipath interference, providing time-varying characteristic information of high-frequency errors for anti-error solution, thereby supporting the improvement of the credibility of the main component of low-frequency structural displacement.

[0051] Step 5: construct a geological model based on the geological structure data in step 1, embed a robust solution model with physical constraints in the geological model, and output confidence intervals of deformation parameters;

[0052] In step 5, a three-dimensional geological model is constructed based on the bedrock elastic modulus and fault zone shear strength parameters extracted from the geological structure data generated in step 1. The state transfer equation is generated in combination with the rock mass constitutive relationship, and the temperature strain compensation term calculated by the temperature and humidity data collected in step 1 is embedded to correct the deformation deviation caused by the thermal expansion effect of the material; the main mode of the low-frequency structural displacement separated in step 4 is input into the robust solution model, and the displacement observation sequence and the state transfer equation are fused using the Kalman filter algorithm. The observation noise and model error are jointly estimated through the covariance matrix, and the confidence intervals of the deformation parameters in the horizontal, vertical and deflection directions are output; when the Hausdorff distance between the confidence interval and the finite element simulation result constructed based on the geological structure data and the monitoring network topology exceeds the preset threshold, the robust estimator based on M estimation is activated, the abnormal observation value is weighted and suppressed, the confidence interval of the deformation parameter is recalculated and updated to the early warning model. In each sub-step, the geological model provides physical constraints for the solution through elastic parameters and temperature compensation. The Kalman filter algorithm achieves a dynamic balance between observed data and model predictions. The robust estimator forms a closed-loop calibration of the solution results through outlier suppression and feedback mechanisms. The quantitative output of the confidence interval provides a multidimensional credibility criterion for deformation safety assessment.

[0053] Step 6: According to the matching degree between the deformation parameter confidence interval output in step 5 and the historical operating condition library, the parameter threshold of the early warning model is dynamically adjusted.

[0054] In step 6, the deformation parameter confidence interval output in step 5 is input into the improved dynamic time warping algorithm, and the geological structure weight factor is introduced to perform multi-dimensional time series alignment of horizontal displacement, vertical settlement and deflection change, match the current deformation mode with similar working conditions in the historical working condition library, and generate a similarity index and warning reference threshold based on dynamic bending path calculation; according to the credibility score output by the anti-error solution model in step 5, a dynamic regularization constraint is applied to the weight matrix of the warning model. The credibility score is jointly calculated by the covariance matrix trace of the deformation parameter confidence interval and the finite element simulation residual. The regularization coefficient is dynamically adjusted with the score to suppress the interference of low-credibility historical working conditions on the threshold; the adjusted warning parameter threshold is transmitted back to the genetic optimization algorithm in step 2 through the feedback link, the fitness function of the monitoring network topology is updated, and the multi-path error suppression term and warning matching factor are added to drive the iterative optimization of the site layout plan. In each sub-step, the dynamic time warping algorithm achieves scenario adaptation of the warning threshold through historical pattern matching. The regularization constraint driven by the credibility score enhances the sensitivity of the warning model to transient deformation. The feedback mechanism dynamically associates the warning parameters with the monitoring network, forming a collaborative iteration of the solution results and network optimization. The closed-loop link improves the timeliness of deformation monitoring and the reliability of warning in highly dynamic environments.

[0055] The present invention relates to deformation monitoring of pumped-storage dams based on Beidou positioning, which specifically includes the following steps: deploying a dual-frequency Beidou receiver array on the dam surface and surrounding bedrock to collect carrier phase observation values ​​and satellite ephemeris data in real time; synchronously laying out reservoir water level pressure sensors and three-dimensional vibration sensors along the dam axis to record water pressure gradients and foundation micro-vibration spectra; combining geological radar detection data with drill core mechanical parameters, and registering them to a three-dimensional dam model through coordinate transformation to generate geological structure data including fault zone distribution; and fusing Beidou observation data, environmental parameter data, and geological structure data according to a unified time and space reference to form a heterogeneous dam monitoring data set.

[0056] Based on the geological structure data and historical deformation information in the dam monitoring dataset, the coordinates of the fault zone deformation-sensitive areas are extracted, and the initial monitoring station layout plan is generated according to the satellite visibility prediction results. The multipath suppression rate and baseline solution accuracy are used as the fitness function. The optimal site set is iteratively selected through a genetic algorithm, and the foundation micro-vibration spectrum and temperature and humidity data are associated with the monitoring station data stream. A spatiotemporal index table is established and the monitoring network topology is updated to achieve monitoring network optimization that is resistant to dynamic interference.

[0057] A cycle slip detection algorithm is implemented on the carrier phase in the Beidou observation data to mark abnormal observation points and generate a corrected phase sequence. The reservoir water level change rate, the corrected phase sequence and the monitoring network topology are integrated and input into the spatiotemporal convolutional network to extract the coupling characteristic spectrum of multipath error and structural deformation. The penalty factor of the variational mode decomposition is dynamically adjusted through the frequency domain energy distribution to decompose the Beidou observation sequence into a low-frequency structural displacement main mode and a high-frequency multipath interference submode carrying a time-varying error label.

[0058] A state transfer equation is constructed based on the bedrock elastic parameters of the geological model, and a temperature strain compensation term is embedded to form a physically constrained robust solution model. The main modes of low-frequency structural displacement are input into the model, and the confidence interval of the deformation parameter is output. When the deviation between the confidence interval and the finite element simulation results exceeds a preset threshold, the robust estimator is activated to suppress outliers and update the solution results.

[0059] The confidence interval of the deformation parameters is input into the improved dynamic time warping algorithm to match the current deformation pattern with similar working conditions in the historical working condition library. Dynamic regularization constraints are applied to the weight matrix of the early warning model based on the credibility score of the solution result. The adjusted parameter threshold is transmitted back to the monitoring network optimization module to update the fitness function of the genetic algorithm, forming a closed-loop feedback link from data acquisition to parameter optimization.

[0060] In the above steps, heterogeneous data fusion provides a unified input for subsequent network optimization, dynamic feature analysis and signal decomposition synergistically suppress multipath time-varying errors, the physical constraint solution model combines geological characteristics to enhance the credibility of deformation parameters, and the closed-loop feedback mechanism achieves dynamic calibration of warning thresholds through historical working condition matching, ultimately improving the real-time performance and accuracy of millimeter-level deformation monitoring in highly dynamic environments.

[0061] Specifically, in the method for monitoring deformation of a pumped storage dam based on Beidou positioning according to the present invention, the dam monitoring dataset in step 1 includes:

[0062] A dual-frequency Beidou receiver array is deployed on the dam surface to obtain carrier phase observations and satellite ephemeris in real time to generate Beidou observation data.

[0063] Reservoir water level pressure sensors and three-dimensional vibration sensors are deployed along the dam axis to synchronously record water pressure gradients and foundation micro-vibration spectra to generate environmental parameter data;

[0064] The geological radar detection data and the mechanical parameters of the drill core are aligned with the 3D model of the dam to generate geological structure data including the distribution of fault zones;

[0065] The Beidou observation data, environmental parameter data and geological structure data are integrated according to a unified time and space reference to obtain a dam monitoring data set.

[0066] In the Beidou positioning-based pumped-storage dam deformation monitoring method described in the present invention, step 1 of generating a dam monitoring data set is specifically implemented as follows: a dual-frequency Beidou receiver array is deployed on the dam surface and key points of the surrounding bedrock. The receiver has a built-in anti-multipath choke antenna, and the carrier phase observation values ​​and satellite ephemeris data of the L1 / L2 frequency points are acquired in real time at a sampling frequency of not less than 10 Hz to generate Beidou observation data; the Beidou observation data is embedded with a precise timestamp when transmitted through the optical fiber network, and the Kalman smoothing algorithm is used to eliminate transmission delay jitter to ensure the synchronization of the time-frequency reference.

[0067] An array of reservoir water level pressure sensors is evenly distributed along the axis of the dam body, and a thermometer and hygrometer and a three-dimensional vibration sensor are installed on the top of the monitoring station to record the reservoir water pressure gradient, atmospheric refractive index and foundation micro-vibration spectrum in real time to generate environmental parameter data; the environmental parameter data is time-correlated indexed with the Beidou observation data through a synchronous trigger module to achieve time alignment of the data.

[0068] The spatial distribution information of the fault zone is obtained through geological radar detection. Combined with the mechanical parameter test results of the drill core, the bedrock elastic modulus and the shear strength characteristics of the fault zone are extracted. The geological data are aligned to the dam BIM coordinate system through a coordinate conversion algorithm to generate geological structure data including the three-dimensional spatial distribution and mechanical properties of the fault zone. During the alignment process, the least squares adjustment algorithm is used to optimize the spatial matching accuracy and eliminate the coordinate system conversion residual.

[0069] Beidou observation data, environmental parameter data and geological structure data are input into the data fusion module, and a unified time and space reference is used to align heterogeneous data. The time and space reference is provided by the GNSS disciplined atomic clock, and the time synchronization error is controlled within ±0.5ms. The fusion module fills the data missing areas through the interpolation algorithm to generate a spatiotemporally continuous dam monitoring data set, providing standardized input for subsequent deformation analysis.

[0070] In the above steps, Beidou observation data provides the foundation for millimeter-level deformation monitoring, environmental parameter data captures the dynamic impact of reservoir water level and foundation vibration on deformation, geological structure data characterizes the dam's mechanical properties and fault zone risk areas, and the data fusion module eliminates data inconsistencies through spatiotemporal alignment and interpolation processing, forming a high-precision, multi-dimensional monitoring data base. Each sub-step is logically connected through timestamp synchronization, coordinate registration, and interpolation algorithms, jointly supporting the subsequent optimization of the monitoring network and deformation resolution.

[0071] Specifically, in the method for monitoring deformation of a pumped storage dam based on Beidou positioning according to the present invention, constructing a monitoring network topology structure in step 2 includes:

[0072] Based on the geological structure data and historical deformation information generated in step 1, the coordinates of the fault zone deformation sensitive area are extracted;

[0073] Generating initial site layout data based on the coordinates of the fault zone deformation sensitive area and satellite visibility prediction results;

[0074] The multipath suppression rate and baseline solution accuracy are used as fitness functions, and the optimal site set is iteratively selected through a genetic algorithm.

[0075] The foundation micro-vibration spectrum and temperature and humidity data in the dam monitoring data set are associated with the monitoring station data stream of the optimal station set, a spatiotemporal index table is established, and the monitoring network topology is updated.

[0076] In the Beidou positioning-based pumped storage dam deformation monitoring method of the present invention, step 2 of constructing the monitoring network topology structure is specifically implemented as follows:

[0077] Based on the three-dimensional geological structure data generated in step 1, a machine learning algorithm is used to extract the coordinate point set of the deformation-sensitive area around the fault zone. Combined with the displacement changes under different working conditions in the historical deformation database, the deformation gradient distribution of the sensitive area is calculated, and high-risk deformation monitoring points are screened out. The deformation gradient distribution is reduced in dimension using the principal component analysis method, and the key eigenvectors are retained for network topology optimization.

[0078] Based on the coordinates of the fault zone deformation-sensitive areas and the Beidou satellite visibility prediction results, a three-dimensional ray tracing algorithm was used to simulate the propagation path of satellite signals in the dam terrain, eliminate areas with high incidence of multipath effects, and generate an initial monitoring station layout plan; each station in the initial plan includes the deployment coordinates of the Beidou receiver, temperature and humidity sensor, and vibration sensor, and marks the satellite signal obstruction probability and baseline solution accuracy prediction value.

[0079] The multipath suppression rate, baseline solution accuracy and signal redundancy are used as fitness functions, and iterative optimization is performed through a genetic algorithm. The crossover operator introduces fault zone avoidance constraints to prevent sites from being deployed in mechanically weak areas. The mutation operator dynamically adjusts the site spacing according to the change of satellite elevation angle, and finally outputs the optimal site set that meets the requirements of anti-dynamic interference. The fitness function balances different optimization objectives through weighted summation, and the weight coefficient is determined by regression analysis of historical deformation data.

[0080] The foundation micro-vibration spectrum and temperature and humidity data in the dam monitoring dataset are associated with the monitoring data stream of the optimal site set through timestamp matching, and a spatiotemporal index table is established; the spatiotemporal index table records the foundation micro-vibration spectrum amplitude, temperature and humidity change curve and corresponding Beidou observation time series of each site. The data association status is updated in real time through a sliding window mechanism to drive the dynamic adjustment of the monitoring network topology structure.

[0081] In these steps, coordinate extraction of deformation-sensitive areas provides a risk-based basis for network deployment. Satellite visibility prediction and genetic optimization collaborate to mitigate multipath interference. Environmental data association achieves information fusion through spatiotemporal indexing. The dynamically updated topology balances monitoring accuracy and anti-interference capabilities. Each sub-step forms a progressive technical chain through feature extraction, path simulation, iterative optimization, and data association, supporting the stable operation of the monitoring network in highly dynamic environments.

[0082] Specifically, in the method for monitoring deformation of a pumped storage dam based on Beidou positioning according to the present invention, the dynamic characteristic analysis in step 3 includes:

[0083] Implement a cycle slip detection algorithm on the carrier phase of the BeiDou observation data received in step 1, mark abnormal observation points, and generate a corrected phase sequence;

[0084] The reservoir water level change rate in the dam monitoring dataset, the phase sequence corrected in step 3, and the monitoring network topology generated in step 2 are input into the spatiotemporal convolutional network, and a dynamic interference feature map is generated through data fusion.

[0085] In the Beidou positioning-based pumped-storage dam deformation monitoring method described in the present invention, the dynamic feature analysis of step 3 is specifically implemented as follows: the TurboEdit cycle slip detection algorithm is implemented on the carrier phase sequence of the Beidou observation data received in step 1, abnormal observation points are marked by a joint criterion of carrier phase and pseudorange, cycle slips and gross error data are eliminated, and a smoothed and corrected phase sequence is generated; the corrected phase sequence is subjected to double difference processing to eliminate the receiver clock error and satellite orbit error, forming a basic data source for high-precision deformation monitoring.

[0086] The reservoir water level change rate data, the corrected phase sequence and the monitoring network topology relationship generated in step 2 in the dam monitoring dataset are input into the spatiotemporal convolutional network. The first branch of the network performs time domain convolution on the reservoir water level change rate to extract the dynamic characteristics of the hydraulic load. The second branch performs spatial domain convolution on the corrected phase sequence to capture the spatial correlation of deformation. The third branch constructs a graph convolution layer based on the topological relationship of the monitoring sites to explore the coupling pattern of multipath interference and structural deformation. The multi-branch features are weightedly fused through the attention mechanism to generate a dynamic interference feature map, which includes the multipath error distribution and structural deformation trend of the time-space-frequency three-dimensional coupling feature map.

[0087] In the above steps, the cycle slip detection algorithm provides highly reliable BeiDou observation data for subsequent analysis. The spatiotemporal convolutional network integrates reservoir water pressure, BeiDou deformation, and site topology through multi-branch collaborative processing. The dynamic interference signature map quantifies the coupling strength between multipath error and actual deformation, providing input for the signal decomposition module. Each sub-step forms a progressive analysis chain through data correction, feature extraction, and fusion mechanisms, supporting the precise separation of error and deformation in highly dynamic environments.

[0088] Specifically, in the method for monitoring deformation of a pumped storage dam based on Beidou positioning according to the present invention, the mixed domain signal decomposition model in step 4 includes:

[0089] Based on the frequency domain energy distribution of the dynamic interference characteristic spectrum generated in step 3, the penalty factor of the variational mode decomposition is dynamically adjusted;

[0090] The corrected phase sequence in step 3 is input into the decomposition module to separate the low-frequency structural displacement main mode and the high-frequency multipath interference submode. The high-frequency submode carries the time-varying error label marked in step 3 and is output to the anti-error solution model in step 5.

[0091] In the Beidou positioning-based pumped-storage dam deformation monitoring method described in the present invention, the mixed domain signal decomposition model of step 4 is specifically implemented as follows: based on the dynamic interference characteristic spectrum generated in step 3, the frequency domain energy distribution parameters are extracted, and the penalty factor of the variational modal decomposition is dynamically optimized through a genetic algorithm; the penalty factor is adaptively adjusted according to the bandwidth and energy concentration of the multipath interference submode to suppress the spectral aliasing of high-frequency noise on the low-frequency deformation signal.

[0092] The corrected phase sequence in step 3 is input into the variational modal decomposition module. The initial number of modes is set to the expected frequency band range of the main mode of structural displacement and the submode of multipath interference. The optimal modal component is iteratively solved by the alternating direction multiplier method. After decomposition, the low-frequency main mode of structural displacement is extracted as the true deformation representation of the dam, and the high-frequency multipath interference submode is annotated with a time-varying error label through time-frequency analysis. The label includes interference intensity and timestamp information.

[0093] The decomposition module outputs the low-frequency structural displacement main mode to the robustness solution model in step 5, and the high-frequency multipath interference submode and its time-varying error label are synchronously transmitted to the error correction unit in step 5; the error label is used to dynamically correct the interference of multipath residual error on deformation parameters during the solution process, thereby improving the robustness of the robustness solution.

[0094] In the above steps, the frequency-domain energy distribution of the dynamic interference signature provides a physical basis for penalty factor optimization. Variational mode decomposition achieves precise segmentation of the signal frequency band through adaptive parameter adjustment. The generation and transmission of error labels form a closed-loop correction link for multipath interference. Through the synergistic effect of frequency-domain analysis, modal decomposition, and error labeling, each sub-step ensures the effective separation of the principal deformation component and the interference component, providing highly reliable input for the subsequent physical constraint solution.

[0095] Specifically, in the method for monitoring deformation of a pumped storage dam based on Beidou positioning according to the present invention, the robustness calculation model in step 5 includes:

[0096] Based on the bedrock elastic parameters of the geological model constructed in step 5, a state transfer equation is generated and a temperature strain compensation term is embedded;

[0097] Input the low-frequency structural displacement main mode separated in step 4 into the robustness calculation model, and output the confidence interval of the deformation parameter;

[0098] When the degree of deviation between the confidence interval of the deformation parameter and the finite element simulation result exceeds a preset threshold, the robust estimator is activated to suppress outliers and the confidence interval is updated.

[0099] In the Beidou positioning-based pumped storage dam deformation monitoring method described in the present invention, the anti-error solution model of step 5 is specifically implemented as follows: based on the bedrock elastic modulus and fault zone shear strength parameters in the geological structure data generated in step 1, a three-dimensional geological model is constructed, and a state transfer equation is generated in combination with the rock mass constitutive relationship in the geological model. The equation corrects the material thermal expansion effect caused by temperature load by introducing the temperature strain compensation term calculated by the temperature and humidity data collected in step 1; the initial parameters of the state transfer equation are randomly sampled and optimized by the Monte Carlo method to improve the model's tolerance to geological parameter uncertainties.

[0100] The main modes of low-frequency structural displacement separated in step 4 are input into the robustness solution model, and the Kalman filter algorithm is used to fuse the displacement observation sequence and the state transfer equation to output the confidence interval of the deformation parameter; the confidence interval describes the uncertainty range of the displacement in the horizontal, vertical and deflection directions through the covariance matrix, and the width of the confidence interval is determined by the joint estimation of observation noise and model error.

[0101] The Hausdorff distance calculation is performed between the confidence interval and the finite element simulation results. The finite element simulation is constructed based on the geological structure data of step 1 and the monitoring network topology relationship of step 2. When the deviation exceeds the preset threshold, the robust estimator based on M estimation is activated, the abnormal observation value is weightedly suppressed, the confidence interval of the deformation parameter is recalculated and updated to the early warning model, and the dynamic adjustment instruction of the monitoring network topology structure in step 2 is triggered at the same time.

[0102] In the above steps, the geological model provides physical constraints for the state transition equations, the temperature-strain compensation term eliminates the influence of environmental factors on material deformation, the Kalman filter algorithm fuses observed data with model predictions to generate credible, quantified deformation parameters, and the robust estimator achieves adaptive calibration of the solution results through outlier suppression and feedback mechanisms. The synergistic effect of physical model construction, data fusion, and closed-loop correction in these substeps ensures the stability and accuracy of deformation parameter calculations in highly dynamic environments.

[0103] Specifically, in the method for monitoring deformation of a pumped storage dam based on Beidou positioning according to the present invention, the dynamic adjustment of the early warning model in step 6 includes:

[0104] Input the confidence interval of the deformation parameters output in step 5 into the improved dynamic time warping algorithm to match the current deformation mode with similar working conditions in the historical working condition library;

[0105] According to the credibility score output by the robust solution model in step 5, dynamic regularization constraints are applied to the weight matrix of the early warning model;

[0106] The adjusted warning parameter threshold is transmitted back to the genetic optimization algorithm in step 2 to update the fitness function of the monitoring network topology structure.

[0107] In the Beidou positioning-based pumped storage dam deformation monitoring method of the present invention, the dynamic adjustment of the early warning model in step 6 is specifically implemented as follows:

[0108] The confidence interval of the deformation parameters output in step 5 is input into the improved dynamic time warping algorithm. The algorithm introduces a geological structure weight factor and performs multi-dimensional time series alignment on the horizontal displacement, vertical settlement, and deflection change to match the current deformation pattern with similar conditions in the historical condition library. The similarity measurement is calculated through dynamic curved path to screen out historical deformation patterns that match the current reservoir water level, temperature load, and geological parameters, and generate a similarity index and warning reference threshold.

[0109] Based on the credibility score output by the robustness solution model in step 5, a dynamic regularization constraint is imposed on the weight matrix of the early warning model; the credibility score is jointly calculated by the covariance matrix trace of the deformation parameter confidence interval and the finite element simulation residual, and the regularization coefficient of the weight matrix is ​​dynamically adjusted according to the score to suppress the interference of low-credibility historical conditions on the early warning threshold and enhance the model's sensitivity to transient abnormal deformation.

[0110] The adjusted warning parameter threshold is transmitted back to the genetic optimization algorithm in step 2 through a feedback link to update the fitness function of the monitoring network topology structure; a multipath error suppression term and a warning matching factor are added to the fitness function to drive the genetic algorithm to iteratively optimize the spatial distribution and sensor configuration of monitoring sites, forming a closed-loop feedback mechanism from deformation solution to network optimization.

[0111] In the above steps, the dynamic time warping algorithm adapts the warning threshold to the scenario by matching historical operating conditions. A regularization constraint driven by credibility scores improves the warning model's noise immunity. A feedback mechanism dynamically links warning parameters with the monitoring network, enabling the coordinated iteration of deformation monitoring and network optimization. Each sub-step, through a logical progression of pattern matching, weight optimization, and closed-loop feedback, improves the timeliness and reliability of deformation warnings in highly dynamic environments.

[0112] The specific implementation method of the present invention is based on the Beidou positioning pumped storage dam deformation monitoring scenario. It addresses the problem of multi-path time-varying errors and frequency domain aliasing of the real displacement of the structure in a high dynamic environment, and realizes millimeter-level real-time monitoring through data collaborative collection, dynamic network optimization and closed-loop feedback mechanism.

[0113] A dual-frequency Beidou receiver array is deployed on the dam surface and at key points in the surrounding bedrock, with a sampling frequency of no less than 10 Hz. Real-time L1 / L2 frequency carrier phase observations and satellite ephemeris data are acquired. Reservoir water level pressure sensors and three-dimensional vibration sensors are synchronously deployed along the dam axis to record water pressure gradients and foundation micro-vibration spectra. Geological radar is used to detect the spatial distribution of fault zones and the mechanical parameters of drill cores. These data are then aligned to the dam's BIM three-dimensional model through coordinate conversion to generate geological structure data including the mechanical properties of the fault zones. Beidou observation data, environmental parameter data, and geological structure data are fused using a unified spatiotemporal reference using a GNSS disciplined atomic clock. The time synchronization error is controlled within ±0.5 ms, forming a spatiotemporally continuous heterogeneous monitoring dataset that provides standardized input for subsequent analysis.

[0114] Based on geological structure data and historical deformation information, principal component analysis is used to extract the coordinates of fault zone deformation-sensitive areas. The three-dimensional ray tracing algorithm is used to simulate the satellite signal propagation path, eliminate areas with high incidence of multipath effects, and generate an initial monitoring station layout plan. The multipath suppression rate, baseline solution accuracy, and signal redundancy are used as fitness functions. The genetic algorithm is used to iteratively optimize the spatial distribution of stations. The crossover operator introduces fault zone avoidance constraints, and the mutation operator dynamically adjusts the station spacing to output the optimal station layout set that is resistant to dynamic interference. The foundation micro-vibration spectrum and temperature and humidity data are associated with the monitoring station data stream, and a spatiotemporal index table is established to drive the dynamic update of the network topology, thereby improving the adaptability of the monitoring network to transient conditions such as sudden drops in reservoir water levels.

[0115] The TurboEdit cycle slip detection algorithm is implemented on the Beidou carrier phase data. Abnormal observation points are marked to generate a corrected phase sequence. The reservoir water level change rate, the corrected phase sequence and the monitoring network topology are integrated and input into the spatiotemporal convolutional network to extract the three-dimensional coupling feature map. Based on the frequency domain energy distribution of the feature map, the variational mode decomposition penalty factor is dynamically adjusted to decompose the Beidou observation sequence into the low-frequency structural displacement main mode and the high-frequency multipath interference submode. The high-frequency submode is transmitted to the robust solution model with the time-varying error label. The state transfer equation is constructed based on the bedrock elastic parameters of the geological model, and the temperature strain compensation term calculated by temperature and humidity data is embedded. The displacement observation sequence is fused through Kalman filtering to output the confidence interval of the deformation parameter. When the confidence interval exceeds the Hausdorff distance of the finite element simulation, the M-estimator robustor is activated to suppress the outlier and update the solution result.

[0116] The confidence interval of the deformation parameters is input into the improved dynamic time warping algorithm to match the current deformation mode with similar working conditions in the historical working condition library, and generate a similarity index and warning reference threshold; based on the credibility score output by the robust solution model, a dynamic regularization constraint is imposed on the warning model weight matrix to suppress the interference of low-credibility historical data; the adjusted warning parameter threshold is transmitted back to the genetic optimization module to update the multipath error suppression term and warning matching factor in the fitness function of the monitoring network, forming a closed-loop feedback link from data acquisition, error separation to network optimization, and ultimately achieving real-time calibration and stability improvement of millimeter-level deformation monitoring accuracy.

[0117] The present invention solves the frequency domain aliasing problem of multipath time-varying errors and real displacements of structures in high dynamic environments through the construction of a multimodal collaborative monitoring network and dynamic feature analysis. First, based on geological structure data and satellite visibility prediction, a genetic algorithm is used to optimize the monitoring site layout plan, and the optimal site set is selected with the multipath suppression rate and baseline solution accuracy as the fitness function. The three-dimensional ray tracing algorithm is used to eliminate high-incidence areas of multipath and reduce the dynamic interference of the signal propagation path. Secondly, cycle slip detection and spatiotemporal convolutional network fusion analysis are performed on Beidou observation data to extract the coupling characteristic map of multipath error and deformation, and the variational modal decomposition parameters are dynamically adjusted in combination with the frequency domain energy distribution to separate the low-frequency structural displacement main mode and the high-frequency multipath interference sub-mode, thereby eliminating the pseudo-fluctuations caused by frequency domain aliasing.

[0118] The reliability of deformation parameters is enhanced through a physically constrained robust solution model. A state transition equation is constructed based on the bedrock elastic parameters of the geological model, with temperature-strain compensation terms embedded to correct for environmental influences. The separated low-frequency displacement main modes are input into the model, and confidence intervals for the deformation parameters are output. When the confidence interval deviates beyond the limit set by the finite element simulation, a robust estimator is activated to suppress outliers. Alert thresholds are dynamically adjusted based on matching with a historical operating condition database, achieving closed-loop calibration with millimeter-level solution accuracy.

[0119] A closed-loop feedback mechanism further optimizes the monitoring network and solution model. An improved dynamic time warping algorithm matches current deformation patterns with historical operating conditions. Dynamic regularization constraints are applied to the early warning model based on credibility scores. Adjusted parameter thresholds are then fed back to the genetic optimization module to update the monitoring network's fitness function. Data fusion, signal decomposition, and physical model iterations are combined to achieve real-time suppression of multipath errors and improved deformation solution accuracy in highly dynamic environments.

Claims

1. A method for monitoring deformation of a pumped storage dam based on Beidou positioning, characterized in that: include: Step 1: Receive BeiDou observation data of the dam surface and surrounding bedrock, and simultaneously collect environmental parameters and geological structure data to generate a dam monitoring dataset; Step 2: Input the geological structure data and historical deformation information in the dam monitoring data set into the genetic optimization algorithm to construct the monitoring network topology structure; Step 3: Perform dynamic feature analysis on the BeiDou observation data received in step 1 to extract the coupling feature map of multipath error and structural deformation; Step 4: Input the coupling characteristic spectrum into a preset mixed-domain signal decomposition model, adjust the decomposition parameters through frequency domain energy distribution, and separate the low-frequency structural displacement main mode and the high-frequency multipath interference submode; Step 5: construct a geological model based on the geological structure data in step 1, embed a robust solution model with physical constraints in the geological model, and output confidence intervals of deformation parameters; Step 6: According to the matching degree between the deformation parameter confidence interval output in step 5 and the historical operating condition library, the parameter threshold of the early warning model is dynamically adjusted.

2. The method for monitoring deformation of a pumped storage dam based on Beidou positioning according to claim 1 is characterized in that: The dam monitoring dataset in step 1 includes: A dual-frequency Beidou receiver array is deployed on the dam surface to obtain carrier phase observations and satellite ephemeris in real time to generate Beidou observation data. Reservoir water level pressure sensors and three-dimensional vibration sensors are deployed along the dam axis to synchronously record water pressure gradients and foundation micro-vibration spectra to generate environmental parameter data; The geological radar detection data and the mechanical parameters of the drill core are aligned with the 3D model of the dam to generate geological structure data including the distribution of fault zones; The Beidou observation data, environmental parameter data and geological structure data are integrated according to a unified time and space reference to obtain a dam monitoring data set.

3. The method for monitoring deformation of a pumped storage dam based on Beidou positioning according to claim 2 is characterized in that: The step 2 of constructing the monitoring network topology structure includes: Based on the geological structure data and historical deformation information generated in step 1, the coordinates of the fault zone deformation sensitive area are extracted; Generating initial site layout data based on the coordinates of the fault zone deformation sensitive area and satellite visibility prediction results; The multipath suppression rate and baseline solution accuracy are used as fitness functions, and the optimal site set is iteratively selected through a genetic algorithm. The foundation micro-vibration spectrum and temperature and humidity data in the dam monitoring data set are associated with the monitoring station data stream of the optimal station set, a spatiotemporal index table is established, and the monitoring network topology is updated.

4. The method for monitoring deformation of a pumped storage dam based on Beidou positioning according to claim 3 is characterized in that: The dynamic feature analysis in step 3 includes: Implement a cycle slip detection algorithm on the carrier phase of the BeiDou observation data received in step 1, mark abnormal observation points, and generate a corrected phase sequence; The reservoir water level change rate in the dam monitoring dataset, the phase sequence corrected in step 3, and the monitoring network topology generated in step 2 are input into the spatiotemporal convolutional network, and a dynamic interference feature map is generated through data fusion.

5. The method for monitoring deformation of a pumped storage dam based on Beidou positioning according to claim 4 is characterized in that: The mixed domain signal decomposition model in step 4 includes: Based on the frequency domain energy distribution of the dynamic interference characteristic spectrum generated in step 3, the penalty factor of the variational mode decomposition is dynamically adjusted; The corrected phase sequence in step 3 is input into the decomposition module to separate the low-frequency structural displacement main mode and the high-frequency multipath interference submode. The high-frequency submode carries the time-varying error label marked in step 3 and is output to the anti-error solution model in step 5.

6. The method for monitoring deformation of a pumped storage dam based on Beidou positioning according to claim 1, characterized in that: The robustness calculation model in step 5 includes: Based on the bedrock elastic parameters of the geological model constructed in step 5, a state transfer equation is generated and a temperature strain compensation term is embedded; Input the low-frequency structural displacement main mode separated in step 4 into the robustness calculation model, and output the confidence interval of the deformation parameter; When the degree of deviation between the confidence interval of the deformation parameter and the finite element simulation result exceeds a preset threshold, the robust estimator is activated to suppress outliers and the confidence interval is updated.

7. The method for monitoring deformation of a pumped storage dam based on Beidou positioning according to claim 1, characterized in that: The dynamic adjustment of the early warning model in step 6 includes: Input the confidence interval of the deformation parameters output in step 5 into the improved dynamic time warping algorithm to match the current deformation mode with similar working conditions in the historical working condition library; According to the credibility score output by the robust solution model in step 5, a dynamic regularization constraint is applied to the weight matrix of the early warning model; The adjusted warning parameter threshold is transmitted back to the genetic optimization algorithm in step 2 to update the fitness function of the monitoring network topology structure.

Images