A method, device, equipment, and program product for identifying abnormal data in water conservancy projects.

Abstract

This application discloses a method, apparatus, equipment, and program product for identifying abnormal data in water conservancy projects. The method includes: collecting environmental and effect data from water conservancy projects to obtain monitoring data; performing multidimensional environmental condition-to-structural effect mapping learning on the monitoring data based on the physical laws of water conservancy projects to construct a Gaussian process regression model; optimizing the Gaussian process regression model using a Bayesian optimization algorithm to obtain an effect prediction model; inputting current environmental data into the effect prediction model for prediction processing, and comparing the predicted values ​​based on dynamic confidence intervals to determine initial abnormal data; and performing random factor analysis processing on the initial abnormal data based on spatiotemporal correlation rules to determine target abnormal data. The embodiments of this application can improve the identification rate of abnormal data in water conservancy projects and can be widely applied in the field of data processing technology.

Description

Technical Field

[0001] This application relates to the field of water conservancy project safety monitoring and data alarm technology, and in particular to a method, device, equipment and program product for identifying abnormal data in water conservancy projects. Background Technology

[0002] In related technologies, there are methods for analyzing hydraulic engineering data using mathematical statistical models, such as threshold setting methods based on the statistical characteristics of historical data. However, these methods are overly sensitive to the distribution of historical data. For complex patterns such as common step transitions and periodic oscillations, the assumptions of traditional methods do not match the data characteristics, and they cannot consider the influence mechanism of effect size, which easily leads to misjudgment and omission, affecting the efficiency of data identification.

[0003] In summary, the technical problems existing in the relevant technologies need to be improved. Summary of the Invention

[0004] The main objective of this application is to propose a method, apparatus, equipment, and program product for identifying abnormal data in water conservancy projects, which can improve the efficiency of identifying water conservancy project data.

[0005] To achieve the above objectives, one aspect of this application proposes a method for identifying abnormal data in water conservancy projects, the method comprising:

[0006] Monitoring data is obtained by collecting environmental and effect data in water conservancy projects; Based on the physical laws of hydraulic engineering, the monitoring data is mapped from multidimensional environmental conditions to structural effects to construct a Gaussian process regression model. The Gaussian process regression model was optimized using a Bayesian optimization algorithm to obtain an effect prediction model. The current environmental data is input into the effect prediction model for prediction processing, and the predicted values ​​are compared based on the dynamic confidence interval to determine the initial abnormal data. Based on spatiotemporal correlation rules, the initial abnormal data is analyzed for accidental factors to determine the target abnormal data.

[0007] In some embodiments, the step of performing multidimensional environmental condition-to-structural effect mapping learning on the monitoring data based on the physical laws of hydraulic engineering to construct a Gaussian process regression model includes the following steps: The mean function and kernel function are configured based on the environmental data in the monitoring data; The parameter search space of the kernel function is pre-defined based on the physical laws of hydraulic engineering; The Gaussian process regression model is constructed based on the mean function, the kernel function, and the parameter search space.

[0008] In some embodiments, optimizing the Gaussian process regression model using a Bayesian optimization algorithm to obtain an effect prediction model includes the following steps: Define a comprehensive evaluation index and use the comprehensive evaluation index as the objective function; Based on the objective function, the kernel function hyperparameters of the Gaussian process regression model are optimized using a Bayesian optimization algorithm to obtain the effect prediction model.

[0009] In some embodiments, the step of inputting current environmental data into the effect prediction model for prediction processing, and comparing the predicted values ​​based on dynamic confidence intervals to determine initial abnormal data includes the following steps: Based on the current environmental data, the corresponding historical environmental data is matched, and the historical environmental data is input into the effect prediction model to output the dynamic confidence interval. The current environmental data is input into the effect prediction model for prediction processing to obtain the predicted value; The predicted value is compared based on the dynamic confidence interval. When the predicted value is not within the range of the dynamic confidence interval, the predicted value is used as the initial abnormal data.

[0010] In some embodiments, the step of performing random factor analysis on the initial abnormal data based on spatiotemporal correlation rules to determine the target abnormal data includes the following steps: Based on spatiotemporal correlation rules, damage mode anomaly analysis is performed on the initial abnormal data to obtain structural anomaly results; The structural anomaly results are analyzed for rationality and the equipment status is correlated to determine the target anomaly data.

[0011] In some embodiments, the step of performing damage mode anomaly analysis on the initial abnormal data based on spatiotemporal correlation rules to obtain structural anomaly results includes the following steps: The initial abnormal data is analyzed for its degree of isolation in the entire spatiotemporal dataset of the monitoring network based on the anomaly detection algorithm, and the spatiotemporal isolated anomaly analysis results are obtained. Based on the spatiotemporal isolated anomaly analysis results, the spatiotemporal evolution feature vector of the initial anomaly data is extracted, and the structural anomaly results are obtained by matching it with historical damage records using a similarity measurement algorithm.

[0012] In some embodiments, the step of performing rationality analysis and equipment status correlation analysis on the structural anomaly results to determine the target anomaly data includes the following steps: The deviation of the structural anomaly data is compared with the preset extreme values ​​of effect data to obtain the physical rationality analysis results; Based on the physical rationality analysis results, the device status indicators are detected, and the device status indicators are analyzed by association rules to determine the target abnormal data.

[0013] To achieve the above objectives, another aspect of this application proposes a device for identifying abnormal data in water conservancy projects, the device comprising: The data acquisition module is used to collect environmental and effect data in water conservancy projects to obtain monitoring data; The model building module is used to perform multidimensional environmental condition to structural effect mapping learning on the monitoring data based on the physical laws of hydraulic engineering, and to build a Gaussian process regression model. The model optimization module is used to optimize the Gaussian process regression model using a Bayesian optimization algorithm to obtain an effect prediction model. The effect prediction module is used to input the current environmental data into the effect prediction model for prediction processing, and compare the predicted values ​​based on the dynamic confidence interval to determine the initial abnormal data. The anomaly identification module is used to perform random factor analysis on the initial anomaly data based on spatiotemporal correlation rules to determine the target anomaly data.

[0014] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method described above.

[0015] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods described above.

[0016] The embodiments of this application include at least the following beneficial effects: This application provides a method, device, equipment, and program product for identifying abnormal data in water conservancy projects. This solution constructs a Gaussian process regression model by mapping monitoring data from multidimensional environmental conditions to structural effects based on the physical laws of water conservancy projects. Furthermore, it performs random factor analysis on the initial abnormal data based on spatiotemporal correlation rules to determine the target abnormal data. This enables the construction of a probability prediction model for monitoring effect quantities based on Gaussian process regression specifically for water conservancy projects. It provides rapid and dynamic alarms for abnormal events that do not conform to the historical change patterns of monitoring data for water conservancy project effect quantities, thereby improving the efficiency of identifying abnormal data in water conservancy projects. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of an implementation environment provided in the embodiments of this application; Figure 2 This is a flowchart of a method for identifying abnormal data in water conservancy projects provided in an embodiment of this application; Figure 3 This is a schematic diagram of a Bayesian optimized threshold generation provided in an embodiment of this application; Figure 4 This is a schematic diagram of the residual of a Bayesian optimized threshold generation model provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of a water conservancy project abnormal data identification device provided in an embodiment of this application; Figure 6 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0019] It is understood that the terms “first,” “second,” etc., used in this application may be used herein to describe various concepts, but unless otherwise stated, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the words “if,” “when,” or “in response to a determination” as used herein may be interpreted as “when…” or “when…” or “in response to a determination.”

[0020] As used in this application, the terms "at least one", "multiple", "each", "any", etc., "at least one" includes one, two or more, "multiple" includes two or more, "each" refers to each of the corresponding multiples, and "any" refers to any one of the multiples.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0022] The safety of water conservancy projects (such as dams, sluices, and dikes) is crucial to many aspects, including the safety of people's lives and property, stable economic and social development, and water resource utilization efficiency. Their safe operation is of paramount importance. Among these, threshold determination, anomaly data identification, and situational alerting are key components in enabling the monitoring and sensing system to function effectively, serving as the vanguard for early warning of safety situations in digital twin water conservancy projects. Currently, the technical challenges of threshold determination, anomaly data identification, and situational alerting urgently need to be addressed to meet the requirements of digital twin water conservancy projects.

[0023] Among related technologies, there are techniques for identifying abnormal data using mathematical statistical models. However, these techniques have limited capabilities for handling complex data. For example, threshold setting methods based on historical data statistical characteristics (such as the low-probability method, Laida's criterion / 3σ principle) are overly sensitive to historical data distribution. For common complex patterns such as step-like transitions and periodic oscillations, the model assumptions in these methods do not match the data characteristics and cannot consider the mechanism of effect size influence, easily leading to misjudgments and omissions. Mechanism models based on mechanical principles and the finite element method (such as dam structure calculations and seepage field simulations) start from physical principles, outputting theoretically meaningful response values ​​through forward analysis or parametric inversion, and using these as threshold benchmarks. While this method has clear physical meaning, it has inherent limitations: the strength of dam materials changes with the service life of the project; the model complexity and computational cost are high; and the adaptive capability is lacking. Different projects require manual modeling and calculation. Therefore, the effect size response of hydraulic engineering is affected by the coupling of multiple environmental quantities, and its threshold determination is essentially a "black-box optimization problem under physical constraints." Relying solely on mathematical models lacks engineering interpretability, while mechanistic models and existing hybrid models struggle to address cost, efficiency, and adaptability.

[0024] In view of this, this application provides a method, device, equipment and program product for identifying abnormal data in water conservancy projects. This solution combines Gaussian processes with Bayesian methods and improves modeling accuracy by optimizing model parameters and data weights. It exhibits strong adaptability in engineering parameter optimization, effectively combines mathematical models and mechanistic models, and finally generates thresholds that are reliable (including confidence intervals), dynamic (adjusted with environmental quantities), and engineering interpretable. It can quickly and dynamically alert abnormal events that do not conform to the historical change patterns of water conservancy project effect monitoring data.

[0025] This application provides a method for identifying abnormal data in water conservancy projects, relating to the field of data processing technology. This method can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, or vehicle-mounted terminal, but is not limited to these. The server can be configured as an independent physical server, a server cluster consisting of multiple physical servers, or a distributed device. It can also be configured as a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network. The software can be an application implementing the method for identifying abnormal data in water conservancy projects, but is not limited to the above forms.

[0026] This application can be used in a wide variety of general-purpose or special-purpose computer device environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor devices, microprocessor-based devices, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above devices, etc. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0027] Figure 1 This is a schematic diagram illustrating the implementation environment of a method provided in an embodiment of this application. (Refer to...) Figure 1 The main hardware and software components of this implementation environment include a terminal 101 and a server 102, which are communicatively connected. The method can be executed based on the interaction between the terminal 101 and the server 102. Furthermore, the terminal 101 and the server 102 can be nodes in a blockchain; this embodiment does not specifically limit this.

[0028] Figure 2 This is an optional flowchart of a method for identifying abnormal data in water conservancy projects provided in an embodiment of this application. Figure 2 The method may include, but is not limited to, steps S201 to S205.

[0029] Step S201: Collect environmental and effect data in water conservancy projects to obtain monitoring data; Step S202: Based on the physical laws of hydraulic engineering, perform multidimensional environmental condition to structural effect mapping learning on the monitoring data to construct a Gaussian process regression model; Step S203: The Gaussian process regression model is optimized using the Bayesian optimization algorithm to obtain the effect prediction model; Step S204: Input the current environmental data into the effect prediction model for prediction processing, and compare the predicted values ​​based on the dynamic confidence interval to determine the initial abnormal data; Step S205: Based on spatiotemporal correlation rules, perform random factor analysis on the initial abnormal data to determine the target abnormal data.

[0030] Steps S201 to S205 of this embodiment involve collecting environmental and effect data from water conservancy projects to obtain monitoring data. Environmental data may include water level and temperature, while effect data includes displacement and seepage pressure. This embodiment addresses the complex dynamic coupling relationship between environmental and effect quantities in water conservancy project safety monitoring. Based on the physical laws of water conservancy projects, it constructs a Gaussian process regression model by mapping the monitoring data from multidimensional environmental conditions to structural effects. Dynamic thresholds are determined by injecting prior knowledge from a surrogate model, using a preset parameter search space, and iteratively searching for the optimal parameter combination. Finally, it analyzes the causes of abnormal data by combining spatiotemporal correlation rules to identify target abnormal data.

[0031] One of the above technical solutions has the following advantages or beneficial effects: This application embodiment combines Gaussian processes with Bayesian methods, improves modeling accuracy by optimizing model parameters and data weights, exhibits strong adaptability in engineering parameter optimization, effectively combines mathematical models and mechanistic models, and finally generates thresholds that are reliable, dynamic, and engineering interpretable, enabling rapid and dynamic alarms for abnormal events that do not conform to the historical change patterns of water conservancy project effect monitoring data.

[0032] In step S201 of some embodiments, environmental data and effect data in the water conservancy project are collected to obtain monitoring data.

[0033] In this embodiment, structural monitoring data, such as displacement, seepage flow, and seepage pressure, can be acquired through a sensor network. Environmental monitoring data, such as reservoir water level, dam internal temperature, regional rainfall, and temperature, can also be acquired through the sensor network. The sampling frequency can be set to once per hour to once per day, and the data undergoes cleaning, noise reduction, and standardization.

[0034] In some embodiments, the step of performing multidimensional environmental condition-to-structural effect mapping learning on the monitoring data based on the physical laws of hydraulic engineering to construct a Gaussian process regression model includes the following steps: The mean function and kernel function are configured based on the environmental data in the monitoring data; The parameter search space of the kernel function is pre-defined based on the physical laws of hydraulic engineering; The Gaussian process regression model is constructed based on the mean function, the kernel function, and the parameter search space.

[0035] In this embodiment, based on the data characteristics of "environmental quantity driven" and "effect quantity response" in the safety monitoring of water conservancy projects, a mean function and kernel function considering environmental quantities are introduced; a parameter search space (such as length scale) is preset based on the structural characteristics of the water conservancy project and the physical meaning of the monitoring data. Signal variance Time decay factor Matern kernel smoothness parameters This application constructs a dedicated Gaussian process regression model by configuring the mean function and kernel function that consider environmental quantities. This allows for the preset parameter search space and the injection of prior knowledge into the surrogate model. The input to the Gaussian process regression model is explicitly defined as a vector of environmental quantities (such as water level and temperature) containing time-series information, and the output is the corresponding effect quantity (such as displacement and seepage pressure). After the prior knowledge is injected, the model learns the mapping from multidimensional environmental conditions to structural responses, laying the foundation for constructing dynamic confidence intervals.

[0036] In some embodiments, optimizing the Gaussian process regression model using a Bayesian optimization algorithm to obtain an effect prediction model includes the following steps: Define a comprehensive evaluation index and use the comprehensive evaluation index as the objective function; Based on the objective function, the kernel function hyperparameters of the Gaussian process regression model are optimized using a Bayesian optimization algorithm to obtain the effect prediction model.

[0037] In this embodiment, a comprehensive evaluation index for anomaly detection performance is defined, with the iterative optimization and search objectives of "maximizing anomaly detection accuracy and minimizing false alarm rate," while also considering historical indicators under similar operating conditions, false alarm rate, false negative rate, and early warning timeliness. Then, based on a dedicated Gaussian process regression model infused with prior physical knowledge, Bayesian optimization is performed to automatically find the model hyperparameters within a preset physical parameter search space. In each iteration, a Gaussian process surrogate model is trained based on existing samples. The most informative parameter combination is selected for evaluation using a collection function (e.g., to improve EI). Within a limited number of iterations, a configuration close to the global optimum is found. Finally, the parameter combination with the highest objective function score is output as the globally optimal hyperparameter configuration, providing model support for subsequent dynamic threshold generation.

[0038] In some embodiments, the step of inputting current environmental data into the effect prediction model for prediction processing, and comparing the predicted values ​​based on dynamic confidence intervals to determine initial abnormal data includes the following steps: Based on the current environmental data, the corresponding historical environmental data is matched, and the historical environmental data is input into the effect prediction model to output the dynamic confidence interval. The current environmental data is input into the effect prediction model for prediction processing to obtain the predicted value; The predicted value is compared based on the dynamic confidence interval. When the predicted value is not within the range of the dynamic confidence interval, the predicted value is used as the initial abnormal data.

[0039] In this embodiment, the confidence interval of monitoring data under the corresponding environmental quantity is calculated or predicted based on the optimized Gaussian process regression model and Bayesian optimization iteration. If the real-time effect quantity exceeds this interval, it is determined to be initial abnormal data. This embodiment aims to achieve the lowest false alarm rate and the lowest false alarm rate, and generates a dynamic threshold by combining the actual characteristics of water conservancy projects, namely "environmental quantity changes + engineering safety level (Level I-IV)". Using historical monitoring data at the same water level and temperature, the optimized model is called, and the theoretical distribution and confidence boundary of the effect quantity under the current environmental quantity are combined to output a dynamic safety threshold with clear statistical confidence significance. The abstract goal of "balancing false alarms and false alarms" is placed in the model parameters optimized by Bayesian to maximize the comprehensive evaluation index, and the prediction uncertainty under different working conditions is automatically quantified by the standard deviation of the posterior distribution, ultimately realizing the output of a dynamic threshold for subsequent anomaly identification. Then, by comparing the real-time monitoring value with the dynamic threshold, if the measured data is not within the range of the dynamic threshold, it is marked as initial abnormal data.

[0040] In some embodiments, the step of performing random factor analysis on the initial abnormal data based on spatiotemporal correlation rules to determine the target abnormal data includes the following steps: Based on spatiotemporal correlation rules, damage mode anomaly analysis is performed on the initial abnormal data to obtain structural anomaly results; The structural anomaly results are analyzed for rationality and the equipment status is correlated to determine the target anomaly data.

[0041] In this embodiment of the application, the causes of anomalies in the initial abnormal data are determined based on spatiotemporal correlation rules and multi-point data. Specifically, damage mode anomaly analysis is performed on the initial abnormal data and historical data through spatiotemporal correlation rules, and rationality analysis and equipment status correlation analysis are performed to determine the target abnormal data.

[0042] In some embodiments, the step of performing damage mode anomaly analysis on the initial abnormal data based on spatiotemporal correlation rules to obtain structural anomaly results includes the following steps: The initial abnormal data is analyzed for its degree of isolation in the entire spatiotemporal dataset of the monitoring network based on the anomaly detection algorithm, and the spatiotemporal isolated anomaly analysis results are obtained. Based on the spatiotemporal isolated anomaly analysis results, the spatiotemporal evolution feature vector of the initial anomaly data is extracted, and the structural anomaly results are obtained by matching it with historical damage records using a similarity measurement algorithm.

[0043] In this embodiment, an anomaly detection algorithm is used to analyze the isolation level of the current anomaly within the entire spatiotemporal dataset of the monitoring network. If an anomaly is determined to be isolated and there are no other anomalies within its preset spatiotemporal neighborhood, it is considered an isolated point with no spatiotemporal correlation. Spatial correlation anomalies are defined as two or more monitoring points exhibiting anomalies within the same monitoring unit (e.g., the same monitoring section of a dam) at the same time period, with consistent anomaly trends. Temporal correlation anomalies are defined as multiple consecutive anomalies at the same monitoring point within a continuous time period. When an anomaly occurs, the spatiotemporal evolution feature vector of the current anomaly event is extracted, and a similarity measurement algorithm is used to match it with records in historical damage data. If the similarity exceeds a preset confidence threshold, the current anomaly is determined to be highly similar to a known structural damage pattern, resulting in a structural anomaly result.

[0044] In some embodiments, the step of performing rationality analysis and equipment status correlation analysis on the structural anomaly results to determine the target anomaly data includes the following steps: The deviation of the structural anomaly data is compared with the preset extreme values ​​of effect data to obtain the physical rationality analysis results; Based on the physical rationality analysis results, the device status indicators are detected, and the device status indicators are analyzed by association rules to determine the target abnormal data.

[0045] In this embodiment, theoretical extreme values ​​for various effect quantities are preset. The deviation of abnormal data, such as sudden displacement changes, is compared with these extreme value ranges; if the deviation exceeds these ranges, the data is considered physically infeasible. Simultaneously, the logical consistency between multiple physical quantities is checked. For example, a sudden increase in seepage pressure without a corresponding displacement response is considered a device malfunction, i.e., a random factor. Furthermore, real-time access or correlation analysis of sensor health status data streams, such as signal noise levels, power supply voltage, and communication packet loss rate, is performed. When an abnormal effect quantity occurs, device status indicators are simultaneously monitored. If status indicators also show abnormalities, such as a sudden increase in signal noise or communication interruption, a preliminary determination of their correlation with device malfunction is made through association rule analysis or causal inference.

[0046] The solutions of this application embodiment will be described in detail and explained below with reference to specific application examples: This application's embodiments construct a dedicated Gaussian process regression model that incorporates the physical characteristics of "environmental quantity driving - effect quantity response" in water conservancy projects. .

[0047] In the formula, Effect size; : Environmental quantity vector (such as water level H, temperature T), m is the number of core environmental quantities; Time series variables (reflecting the time characteristics of the monitoring data); : Environmental quantity-driven weight vector; Mean function; : Kernel functions for hydraulic engineering that incorporate temporal correlations and physical constraints Preset parameter search space ( For environmental quantities, the length scale is used. As a time length scale, The time decay factor, For the Matérn kernel smoothness parameter, For signal variance, (This refers to the noise variance).

[0048] The covariance function (kernel function) of a Gaussian process (GP) can be the RBF kernel function that considers environmental quantities of hydraulic engineering, the Matérn kernel function with time decay factor, or a combination of kernels.

[0049] Standard RBF kernel function: ; In the formula , The hyperparameters of the RBF kernel define the feature length scale for learning the similarity between samples.

[0050] Standard Matérn kernel function: ; In the formula , These are the hyperparameters of the kernel function. To correct the Bessel function, Common choices are 1.5 and 2.5.

[0051] The expression for the RBF kernel function considering environmental quantities in water conservancy projects is as follows: ; The expression for the Matérn kernel function considering environmental quantities in hydraulic engineering is as follows: ; Considering the combination of environmental quantities in water conservancy projects: ; Once the dataset is observed, the posterior distribution of the objective function remains a Gaussian process, and its mean and covariance can be obtained analytically. The posterior formula considering environmental quantities of water conservancy projects is as follows.

[0052] Posterior mean formula: ; Posterior covariance formula: ; in: , : New input to be predicted (such as candidate combinations of hyperparameters); : The covariance vector between the new input and the observed input; The covariance matrix, where ; Noise variance (captures noise in observed data, such as sensor error); : Observed effect size vector (n-dimensional).

[0053] The identity matrix; Bayesian optimization selects the "most valuable" next evaluation point from the posterior distribution using a sampling function. This involves balancing "exploration (of unknown regions)" and "utilization (of known superior regions)." Common approaches include expected improvement and upper confidence bounds.

[0054] Expected Improvement (EI): ; in , This is the current best observation value. and These are the cumulative distribution function and probability density function of the standard normal distribution, respectively.

[0055] Upper Confidence Bound (UCB): ; in It is a trade-off coefficient The larger the area, the more inclined it is to "explore" areas of high uncertainty.

[0056] Based on spatial correlation anomalies, temporal correlation anomalies, and damage pattern similarity anomalies, the system is judged as "structural anomaly" and a safety alarm is triggered; based on spatiotemporal isolation anomalies, physical rationality exceeding limits, and equipment status correlation anomalies, the system is judged as "equipment failure" and an inspection and maintenance alarm is triggered.

[0057] In one feasible embodiment, a medium-sized reservoir is a homogeneous earth dam, consisting of a main dam and a secondary dam. The main dam is 350m long, 7.00m wide at the crest, and has a maximum height of 40.14m. Surface deformation monitoring is performed using GNSS monitoring, with a total of 4 cross-sections and 11 monitoring points.

[0058] Taking a medium-sized homogeneous earth dam as the application example, the dam consists of a main dam and a secondary dam. The main dam is 350m long, 7.00m wide at the crest, and has a maximum height of 40.14m. To achieve real-time monitoring of the dam surface displacement, 11 GNSS displacement monitoring points were set up along the dam body at 4 monitoring sections. The monitoring period covered 2 complete hydrological years. The monitoring data included environmental quantities (reservoir water level, temperature around the dam body) and effect quantities (vertical and horizontal displacement of the GNSS measuring points).

[0059] Preprocessing methods include initial outlier filtering, which removes isolated data points that significantly deviate from their physical meaning. Feature standardization involves using Z-score standardization to process data of different dimensions, such as water level and displacement, into a distribution with a mean of 0 and a standard deviation of 1. This yields a standardized monitoring dataset. The Gaussian process regression model is then constructed, with initial parameters set to a constant value for the mean function. =0. Covariance function: A combination of Matérn kernel and RBF kernel is used. The fitting function is automatically selected through Bayesian optimization. The search space for Bayesian optimization is set to: length scale. [0.1, 10.0], signal variance [0.01, 100.0], dynamic confidence interval parameter. For example... Figure 3 and Figure 4 As shown, Bayesian optimization training, iteration, and validation are performed. The core of Bayesian optimization is balancing "exploration (of unknown hyperparameter regions)" and "utilization (of known optimal regions)." `length_scale_time` represents the time-related length scale, `length_scale_water` represents the water level-related length scale, and `water_scale` represents the water level influence. In the initial iterations (1-8), the target function value fluctuates significantly (-317.0867 to -314.8141), which is the stochastic exploration phase of Bayesian optimization (`init_points=8` in the code). The goal is to quickly cover the hyperparameter search space and find potential optimal regions. In subsequent iterations (9-38), the target function value gradually stabilizes between -315.1114 and -315.4792, with only occasional poor values ​​of -317.0867. This is the utilization phase—based on previous results, focusing on fine-tuning within the optimal regions. The hyperparameters gradually approximate the maximum objective function value (optimal fit). The target value of -314.8141 in the 8th iteration is the global optimum, and the corresponding hyperparameter combination is: length_scale_time=0.784, length_scale_water=9.102, water_scale=0.525, interaction_scale=0.666, noise_level=0.031, nu=2.06, which means that the model has the highest fitting accuracy for the "water level-displacement" data under this combination.

[0060] The specific physical laws governing hydraulic engineering are as follows: Water level-displacement response law: As shown in the example data, when the reservoir water level rises, the GNSS measuring points of the dam body generally produce an increase in displacement in the downstream direction, and the displacement change has continuity and consistent trend, which is consistent with the physical mechanism of dam body deformation under water pressure.

[0061] Temperature-displacement time-varying law: The displacement of the dam body is significantly affected by temperature changes, exhibiting periodic characteristics (such as annual cycle and monthly cycle), reflecting the memory and decay characteristics of thermal expansion and contraction effect.

[0062] Nonlinear Laws Coupled with Multiple Environmental Quantities: Displacement is not a linear function of a single environmental quantity, but rather a nonlinear response to the combined effects of multiple factors such as water level and temperature. In the example, Bayesian optimization automatically learned the water level influence parameter (water_scale) and the interaction parameter (interaction_scale), indicating that the model captured the interactive coupling effect between water level and factors such as time.

[0063] Before optimization, physical correlation analysis (e.g., water level-displacement) was performed on historical monitoring data to determine the degree and direction of influence of each environmental quantity on the effector quantity, and to pre-screen the search range of key hyperparameters based on physical rationality. For example, water_scale (water level influence): before optimization, based on engineering common sense and preliminary data analysis, the search range of key hyperparameters was pre-screened based on physical rationality, setting the initial range to [0.01~2.5]. The water_scale values ​​(0.139~1.466) in the initial iterations (1-8 times) all fell within this range, and none were below 0.1 or above 2.5, proving that the search space was anchored to a physically reasonable range from the outset, laying the foundation for subsequent focus on the 0.5~2.0 range. In the initial exploration, the value ranged from 0.01 (water level has almost no influence) to 2.0 (water level has a strong influence), and subsequently stabilized between 0.5 and 2.0, consistent with the engineering common sense that "water level is a key environmental quantity affecting dam displacement." nu (nonlinear adaptation): The initial range was set to [0.5~3.5]. All iterations of nu values ​​(0.5~3.409) did not exceed this range, and the early explorations (1-8 times) avoided extreme values ​​below 0.5, reflecting the constraint effect of the physical boundary. Most iteration values ​​were concentrated between 1.5 and 2.5, indicating that the "water level-displacement" relationship is moderately nonlinear (not extremely nonlinear), consistent with the gradual characteristics of dam deformation affected by water level in reality; noise_level (noise tolerance): Most iteration values ​​were concentrated between 0.01 and 0.09, which avoided both "the model over-tolerating noise and missing abnormal displacements" caused by noise_level=0.1 and "over-sensitivity and false reporting of normal fluctuations" caused by noise_level=1e-6.

[0064] The final hyperparameters of Bayesian optimization are used to train the Gaussian process model, thereby generating dynamic confidence intervals (anomaly detection thresholds). The optimal noise_level=0.031 determines the model's ability to filter noise, avoiding "mistakenly classifying sensor random errors as structural anomalies" or "missing the report of true minor deformation anomalies" due to improper noise threshold settings.

[0065] Based on the optimized Gaussian process model, a dynamic confidence interval (normal displacement fluctuation range) is generated. Combined with the typical characteristics of displacement monitoring anomalies in water conservancy projects, the abnormal data is qualitatively identified and its causes are determined, categorized into structurally related anomalies (non-accidental factors) and non-structurally related anomalies (accidental factors). The model's effectiveness is qualitatively verified from an engineering practicality perspective through simulation identification of these two typical anomaly scenarios. Structural anomaly capture capability: For simulated gradual and abrupt structural anomalies, the model can accurately identify them based on the dynamic confidence interval, and the anomaly cause determination results are consistent with the engineering physical mechanism (e.g., classifying synchronous displacement anomalies at multiple measuring points caused by a sudden rise in water level as structural anomalies). Non-structural anomaly exclusion capability: For simulated anomalies caused by equipment or transmission, the model can effectively exclude them through spatiotemporal correlation rules, avoiding misjudging isolated, instantaneous non-structural anomalies as structural safety issues. False alarms are concentrated in non-critical operating conditions (e.g., instantaneous fluctuations at a single measuring point during low water levels), minimizing interference with engineering decisions. The dynamic confidence interval output by the model can adaptively adjust with changes in environmental quantities, meeting the monitoring needs of water conservancy projects under different operating conditions.

[0066] Please see Figure 5 This application also provides a device for identifying abnormal data in water conservancy projects, which can implement the above-mentioned method for identifying abnormal data in water conservancy projects. The device includes: The data acquisition module 501 is used to collect environmental and effect data in water conservancy projects to obtain monitoring data; The model building module 502 is used to perform multidimensional environmental condition to structural effect mapping learning on the monitoring data based on the physical laws of hydraulic engineering, and to build a Gaussian process regression model. The model optimization module 503 is used to optimize the Gaussian process regression model using a Bayesian optimization algorithm to obtain an effect prediction model. The effect prediction module 504 is used to input the current environmental data into the effect prediction model for prediction processing, and compare the predicted values ​​based on the dynamic confidence interval to determine the initial abnormal data. The anomaly identification module 505 is used to perform random factor analysis on the initial abnormal data based on spatiotemporal correlation rules to determine the target abnormal data.

[0067] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0068] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0069] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0070] Please see Figure 6 , Figure 6 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 601 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 602 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 602 can store operating devices and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 602 and is called and executed by the processor 601. The input / output interface 603 is used to implement information input and output; The communication interface 604 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 605 transmits information between various components of the device (e.g., processor 601, memory 602, input / output interface 603, and communication interface 604); The processor 601, memory 602, input / output interface 603, and communication interface 604 are connected to each other within the device via bus 605.

[0071] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0072] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0073] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0074] This application provides a method, apparatus, equipment, and program product for identifying abnormal data in water conservancy projects. It relies on Bayesian optimization to intelligently optimize and iterate the hyperparameters (such as length scale and signal variance) of the kernel function of a Gaussian process regression model. This allows for the generation of dynamic thresholds with probability confidence intervals for each monitoring point (such as GNSS displacement measurement points or seepage pressure measurement points on a dam) under different environmental conditions. Compared to traditional mathematical statistics methods such as the 3σ principle and the low-probability method, this method significantly narrows the threshold width, avoiding missed detections caused by the inability of traditional static thresholds to adapt to nonlinear fluctuations caused by sudden rises in water levels during the flood season. Compared to mechanistic models, this method maintains a strong correlation with physical quantities without relying on precise constitutive models and high computational costs. It uses data-driven adaptive learning to establish the relationship between environmental and effect quantities. It provides a unified, adaptive optimization model that automatically adapts to different engineering and monitoring data types through Bayesian optimization, overcoming the inherent limitations of rigid hybrid models and weak generalization capabilities. Furthermore, by combining spatiotemporal correlation rules with multi-point data linkage analysis, this method can automatically identify the causes of anomalies, classifying them as "structural deformation anomalies." If only a single measuring point exhibits an isolated anomaly, it is classified as a "non-structurally related anomaly (accidental factors, sensor malfunctions, etc.)," ​​further improving the efficiency and reliability of monitoring decisions. It also enables rapid and dynamic alerts for anomalies that do not conform to the historical variation patterns of water conservancy project effect quantity monitoring data.

[0075] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of this application, and do not constitute a limitation on the technical solutions provided in this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in this application are also applicable to similar technical problems.

[0076] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0077] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; 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 achieve the purpose of this embodiment according to actual needs.

[0078] Those skilled in the art will understand that all or some of the steps, apparatuses, or functional modules / units in the methods disclosed above can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0079] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0080] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0081] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed between the devices or units may be through some interfaces, and the indirect coupling or communication connection may be electrical, mechanical, or other forms.

[0082] The units described above as separate components may or may not be physically separate. The components shown as units 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 units can be selected to achieve the purpose of this embodiment according to actual needs.

[0083] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0084] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0085] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A method for identifying abnormal data in water conservancy projects, characterized in that, The method includes the following steps: Monitoring data is obtained by collecting environmental and effect data in water conservancy projects; Based on the physical laws of hydraulic engineering, the monitoring data is mapped from multidimensional environmental conditions to structural effects to construct a Gaussian process regression model. The Gaussian process regression model was optimized using a Bayesian optimization algorithm to obtain an effect prediction model. The current environmental data is input into the effect prediction model for prediction processing, and the predicted values ​​are compared based on the dynamic confidence interval to determine the initial abnormal data. Based on spatiotemporal correlation rules, the initial abnormal data is analyzed for accidental factors to determine the target abnormal data.

2. The method according to claim 1, characterized in that, The method of mapping the monitoring data from multidimensional environmental conditions to structural effects based on the physical laws of hydraulic engineering to construct a Gaussian process regression model includes the following steps: The mean function and kernel function are configured based on the environmental data in the monitoring data; The parameter search space of the kernel function is pre-defined based on the physical laws of hydraulic engineering; The Gaussian process regression model is constructed based on the mean function, the kernel function, and the parameter search space.

3. The method according to claim 1, characterized in that, The optimization of the Gaussian process regression model using the Bayesian optimization algorithm to obtain the effect prediction model includes the following steps: Define a comprehensive evaluation index and use the comprehensive evaluation index as the objective function; Based on the objective function, the kernel function hyperparameters of the Gaussian process regression model are optimized using a Bayesian optimization algorithm to obtain the effect prediction model.

4. The method according to claim 1, characterized in that, The process of inputting current environmental data into the effect prediction model for prediction processing, and comparing the predicted values ​​based on dynamic confidence intervals to determine initial abnormal data includes the following steps: Based on the current environmental data, the corresponding historical environmental data is matched, and the historical environmental data is input into the effect prediction model to output the dynamic confidence interval. The current environmental data is input into the effect prediction model for prediction processing to obtain the predicted value; The predicted value is compared based on the dynamic confidence interval. When the predicted value is not within the range of the dynamic confidence interval, the predicted value is used as the initial abnormal data.

5. The method according to claim 1, characterized in that, The step of performing random factor analysis on the initial abnormal data based on spatiotemporal correlation rules to determine the target abnormal data includes the following steps: Based on spatiotemporal correlation rules, damage mode anomaly analysis is performed on the initial abnormal data to obtain structural anomaly results; The structural anomaly results are analyzed for rationality and the equipment status is correlated to determine the target anomaly data.

6. The method according to claim 5, characterized in that, The step of performing damage mode anomaly analysis on the initial abnormal data based on spatiotemporal correlation rules to obtain structural anomaly results includes the following steps: The initial abnormal data is analyzed for its degree of isolation in the entire spatiotemporal dataset of the monitoring network based on the anomaly detection algorithm, and the spatiotemporal isolated anomaly analysis results are obtained. Based on the spatiotemporal isolated anomaly analysis results, the spatiotemporal evolution feature vector of the initial anomaly data is extracted, and the structural anomaly results are obtained by matching it with historical damage records using a similarity measurement algorithm.

7. The method according to claim 5, characterized in that, The process of performing rationality analysis and equipment status correlation analysis on the structural anomaly results to determine the target anomaly data includes the following steps: The deviation of the structural anomaly data is compared with the preset extreme values ​​of effect data to obtain the results of physical rationality analysis. Based on the physical rationality analysis results, the device status indicators are detected, and the device status indicators are analyzed by association rules to determine the target abnormal data.

8. A device for identifying abnormal data in water conservancy projects, characterized in that, The device includes: The data acquisition module is used to collect environmental and effect data in water conservancy projects to obtain monitoring data; The model building module is used to perform multidimensional environmental condition to structural effect mapping learning on the monitoring data based on the physical laws of hydraulic engineering, and to build a Gaussian process regression model. The model optimization module is used to optimize the Gaussian process regression model using a Bayesian optimization algorithm to obtain an effect prediction model. The effect prediction module is used to input the current environmental data into the effect prediction model for prediction processing, and compare the predicted values ​​based on the dynamic confidence interval to determine the initial abnormal data. The anomaly identification module is used to perform random factor analysis on the initial anomaly data based on spatiotemporal correlation rules to determine the target anomaly data.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method according to any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images