A method and device for precision monitoring of water conservancy engineering elements and a monitoring management platform

Abstract

This application discloses a method, device, and monitoring management platform for precise monitoring of water conservancy engineering elements. This application belongs to the field of water conservancy engineering monitoring technology. The method includes: acquiring multi-source monitoring raw data of water conservancy engineering to obtain a standard monitoring dataset; constructing a differentiated full-element monitoring indicator system for the water conservancy monitoring object; receiving expert weighting information to determine the weight value of each monitoring indicator in the indicator system; using a physical cause prediction model combined with the confidence interval method to determine the safety monitoring threshold of each monitoring indicator; inputting a preset prediction model to obtain the predicted monitoring value of each monitoring indicator; and, for the water conservancy monitoring object, comparing the real-time monitoring data and predicted monitoring value of each monitoring indicator with the safety monitoring threshold of each monitoring indicator to identify whether there is an abnormal state of the indicator and to perform abnormal pre-monitoring. This technical solution can achieve full-element, differentiated, and intelligent precise monitoring and proactive control for different monitoring objects.

Description

Technical Field

[0001] This application belongs to the field of water conservancy engineering monitoring technology, specifically relating to a precision monitoring method, device and monitoring management platform for water conservancy engineering elements. Background Technology

[0002] Safety monitoring of water conservancy projects is a crucial link in ensuring the stable operation of hydraulic structures such as sluices and pumping stations. Current water conservancy project monitoring largely relies on manual observation, single-item data collection, and simple comparisons. It depends primarily on manual data recording and single-indicator judgment of safety status, making it difficult to achieve multi-source data fusion and unified processing. Furthermore, existing methods lack differentiated indicator systems for sluices and pumping stations, generally using generic monitoring items that fail to align with the structural characteristics and disease mechanisms of different projects. Moreover, current technologies often rely on experts to subjectively determine monitoring indicators and their weights, leading to highly subjective and unreliable evaluation results. In other scenarios, safety monitoring thresholds are often determined using empirical values ​​or simple statistical methods, resulting in insufficient accuracy in early warning and a high risk of false alarms and missed alarms. In addition, existing systems can only perform real-time data comparisons and cannot conduct predictive analysis based on historical time-series data, making it difficult to detect potential safety hazards in advance. Overall, existing water conservancy project monitoring methods suffer from low data standardization, insufficient objectivity, and delayed early warning methods, failing to meet current safety monitoring needs. Summary of the Invention

[0003] This application provides a method, device, and monitoring management platform for precise monitoring of water conservancy engineering elements. The purpose is to solve the technical problems of non-standard water conservancy monitoring data, unreasonable division of indicator system, and inability to carry out effective early warning. It can realize precise monitoring and early control of all elements, differentiated and intelligent features for different monitoring objects.

[0004] In a first aspect, embodiments of this application provide a method for precise monitoring of water conservancy engineering elements, the method comprising: Obtain raw data from multi-source monitoring of water conservancy projects, and standardize the raw data to obtain a standard monitoring dataset; wherein the raw data from multi-source monitoring includes at least one of sensing data collected by sensing devices, environmental quantity data, and manually supplemented data; Based on the standard monitoring dataset, a differentiated full-element monitoring indicator system is constructed for at least two types of water conservancy monitoring objects; Receive expert weighting information and use a consistency test algorithm to determine the weight value of each monitoring indicator in the indicator system; Based on the aforementioned indicator system and the weight values ​​of each monitoring indicator, the safety monitoring threshold of each monitoring indicator is determined by using a physical cause prediction model combined with the confidence interval method. Collect real-time monitoring data for each monitoring indicator, input the real-time monitoring data and the historical monitoring data of the corresponding indicator into a preset prediction model, and obtain the prediction results of the monitoring values ​​of each monitoring indicator. For the water conservancy monitoring object, based on the real-time monitoring data of each monitoring indicator and the safety monitoring threshold of each monitoring indicator, it is possible to identify whether there is an abnormal state of the indicator; and based on the monitoring value prediction results of each monitoring indicator and the safety monitoring threshold of each monitoring indicator, it is possible to conduct full-element indicator anomaly pre-monitoring of the water conservancy monitoring object.

[0005] Secondly, embodiments of this application provide a precision monitoring device for water conservancy engineering elements, the device being configured as follows: A standard dataset construction module is used to acquire raw data from multi-source monitoring of water conservancy projects, and to standardize the raw data to obtain a standard monitoring dataset; wherein, the raw data from multi-source monitoring includes at least one of sensing data collected by sensing devices, environmental quantity data, and manually supplemented data; The indicator system construction module is used to construct a differentiated full-element monitoring indicator system for at least two types of water conservancy monitoring objects based on the standard monitoring dataset. The weight value determination module is used to receive expert weighting information and use a consistency test algorithm to determine the weight value of each monitoring indicator in the indicator system. The safety monitoring threshold determination module is used to determine the safety monitoring threshold of each monitoring indicator based on the indicator system and the weight value of each monitoring indicator, using a physical cause prediction model combined with the confidence interval method. The prediction result determination module is used to collect real-time monitoring data of each monitoring indicator, input the real-time monitoring data and the historical monitoring data of the corresponding indicator into a preset prediction model, and obtain the prediction result of the monitoring value of each monitoring indicator. The anomaly identification and prediction module is used to identify whether there is an abnormal state of the water conservancy monitoring object based on the real-time monitoring data of each monitoring indicator and the safety monitoring threshold of each monitoring indicator; and to perform full-element indicator anomaly pre-monitoring of the water conservancy monitoring object based on the monitoring value prediction results of each monitoring indicator and the safety monitoring threshold of each monitoring indicator.

[0006] Thirdly, embodiments of this application provide a monitoring and management platform, including a data acquisition terminal, a data processing terminal, and an anomaly response terminal; wherein, The data acquisition terminal is used for: Acquire raw data from multi-source monitoring of water conservancy projects. The raw data from multi-source monitoring includes sensing data collected by sensing devices, environmental quantity data, and manually supplemented data. Send the raw data from multi-source monitoring to the data processing terminal. The data processing terminal is used for: The multi-source monitoring raw data is standardized to obtain a standard monitoring dataset. Based on the standard monitoring dataset, a differentiated full-element monitoring indicator system is constructed for at least two types of water conservancy monitoring objects. Expert weighting information is received, and a consistency check algorithm is used to determine the weight value of each monitoring indicator in the indicator system. Based on the indicator system and the weight values ​​of each monitoring indicator, a physical cause prediction model combined with the confidence interval method is used to determine the safety monitoring threshold of each monitoring indicator. Real-time monitoring data of each monitoring indicator is collected, and the real-time monitoring data and the historical monitoring data of the corresponding indicator are input into a preset prediction model to obtain the monitoring value prediction results of each monitoring indicator. Based on the real-time monitoring data of each monitoring indicator and the safety monitoring threshold, it is identified whether there is an abnormal state of the indicator. Based on the monitoring value prediction results of each monitoring indicator and the safety monitoring threshold, full-element indicator anomaly pre-monitoring is performed on the water conservancy monitoring object. The exception response terminal is used for: When an abnormal state of an indicator is identified or an abnormal trend is detected, abnormal alarm information and pre-monitoring warning information are output, and the corresponding abnormal handling and feedback process is executed.

[0007] Fourthly, embodiments of this application provide an electronic device including a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the method described in the first aspect.

[0008] Fifthly, embodiments of this application provide a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect.

[0009] In a sixth aspect, embodiments of this application provide a chip, the chip including a processor and a communication interface, the communication interface being coupled to the processor, the processor being used to run programs or instructions to implement the method as described in the first aspect.

[0010] The technical solution provided in this application constructs a differentiated index system adapted to the characteristics of water conservancy equipment such as sluice gates and pumping stations. It enhances the objectivity of weights through expert weighting and consistency checks, improves threshold accuracy using physical causal models and confidence interval methods, and achieves future state prediction through a pre-set prediction model. Ultimately, it completes real-time anomaly identification and proactive monitoring. This technical solution realizes precision and proactive monitoring throughout the entire process of water conservancy project monitoring, effectively reducing false alarms and missed alarms, and comprehensively improving the project's safety management capabilities and operational reliability. Attached Figure Description

[0011] Figure 1This is a flowchart illustrating the precise monitoring method for water conservancy engineering elements provided in Embodiment 1 of this application; Figure 2 This is a schematic diagram of the workflow of the EEMD-IPSO-BiLSTM osmosis prediction model provided in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of a precision monitoring device for water conservancy engineering elements provided in Embodiment 2 of this application; Figure 4 This is a schematic diagram of the structure of a monitoring and management platform provided in Embodiment 3 of this application. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of this application clearer, specific embodiments of this application will be described in further detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely for explaining this application and not for limiting it. It should also be noted that, for ease of description, only the parts relevant to this application are shown in the drawings, not all of them. Before discussing exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe operations (or steps) as sequential processes, many of these operations can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but may also have additional steps not included in the drawings. The process can correspond to a method, function, procedure, subroutine, subroutine, etc.

[0013] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0014] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0015] The following, in conjunction with the accompanying drawings, provides a detailed description of the precision monitoring method, device, and monitoring management platform for water conservancy engineering elements provided in this application, through specific embodiments and application scenarios.

[0016] Example 1 Figure 1 This is a flowchart illustrating the precision monitoring method for water conservancy engineering elements provided in Embodiment 1 of this application. Figure 1 As shown, the specific steps include the following: S11, Obtain the original data of multi-source monitoring of water conservancy projects, and perform standardization processing on the original data of multi-source monitoring to obtain a standard monitoring dataset.

[0017] Among them, multi-source monitoring raw data refers to water conservancy project monitoring data collected from multiple sources, which can be at least one of sensing data collected by sensing devices, environmental quantity data, and artificially supplemented data.

[0018] Sensing devices refer to monitoring instruments that automatically collect engineering status data. Examples include GNSS receivers, hydrostatic levels, piezometers, water level gauges, flow meters, as well as crack gauges and opening / closing force sensors.

[0019] Sensing data refers to the structural physical quantity data collected by the equipment, which may include displacement, uplift pressure, structural joint opening data, as well as lateral seepage and riverbed scouring and deposition data.

[0020] Environmental data refers to external data that affects the safety of a project. Specifically, it can include upstream and downstream water levels, air temperature, precipitation, river scouring and silting data, as well as wind speed, water temperature, and river flow data.

[0021] Manually supplemented data refers to data that is manually observed and entered, such as manually measured settlement, cross-section, scour and siltation, manual inspection records, and equipment status notes.

[0022] Standardized processing can refer to standardized data processing procedures such as data cleaning, deduplication, spatiotemporal alignment, unit unification, and outlier removal.

[0023] S12, Based on the standard monitoring dataset, construct a differentiated full-element monitoring indicator system for at least two types of water conservancy monitoring objects.

[0024] Among them, the water conservancy monitoring objects refer to the water conservancy engineering structures being monitored, specifically including hydraulic structures such as sluice gates, pumping stations, control gates, pumping stations, ship locks, and rubber dams.

[0025] A differentiated full-element monitoring indicator system refers to a multi-level set of monitoring indicators constructed for different types of projects. Specifically, it can be a set of indicators for sluice gates or a set of indicators for pumping stations. Different sets of indicators differ in the indicators they contain.

[0026] S13, Receive expert weighting information and use a consistency check algorithm to determine the weight value of each monitoring indicator in the indicator system.

[0027] Expert-weighted information refers to the scoring information assigned by experts to the importance of indicators. It can include the scoring information of water conservancy engineering experts on water level, deformation, and seepage indicators, as well as the weighting information of safety monitoring experts on structural joints, uplift pressure, and opening and closing force indicators.

[0028] Consistency test algorithms are statistical algorithms used to test whether the opinions of multiple experts are similar. For example, they can be Kendall's coefficient of consistency test, Pearson's coefficient of consistency test, or Spearman's correlation coefficient test.

[0029] S14. Based on the indicator system and the weight values ​​of each monitoring indicator, the safety monitoring threshold of each monitoring indicator is determined by using a physical cause prediction model combined with the confidence interval method.

[0030] Physical cause prediction models refer to prediction models built based on structural stress, deformation, and seepage mechanisms. Specifically, they can be deformation statistical regression models, seepage mechanics models, or multi-component combination models such as water pressure-temperature-aging.

[0031] The confidence interval method refers to a calculation method for determining the safe interval based on the statistical distribution of residuals. Specifically, it can be a 2-standard-deviation warning method, a 3-standard-deviation alarm method, or a 95% confidence interval or a 99% confidence interval method.

[0032] Safety monitoring thresholds refer to the critical values ​​for judging whether data is abnormal. Specifically, they include warning thresholds and alarm thresholds. Warning thresholds can be the predicted value ± 2 standard deviations, and alarm thresholds can be the predicted value ± 3 standard deviations.

[0033] S15, collect real-time monitoring data of each monitoring indicator, input the real-time monitoring data and the historical monitoring data of the corresponding indicator into the preset prediction model, and obtain the prediction results of the monitoring values ​​of each monitoring indicator.

[0034] Real-time monitoring data refers to the measured data of indicators collected at the current moment. For example, it can be the current water level, real-time displacement, instantaneous uplift pressure data, or real-time structural joint opening and instantaneous lateral seepage data.

[0035] Historical monitoring data refers to time-series data of indicators from the past period, such as monitoring data from the past year, month, or week, or segmented historical data from the flood season, dry season, or maintenance period.

[0036] Pre-set prediction models refer to pre-trained time series prediction machine learning models, which can be EEMD-IPSO-BiLSTM models, LSTM models, BiLSTM models, or other regression prediction models.

[0037] S16, for the water conservancy monitoring object, based on the real-time monitoring data of each monitoring indicator and the safety monitoring threshold of each monitoring indicator, identify whether there is an abnormal state of the indicator; and, based on the monitoring value prediction results of each monitoring indicator and the safety monitoring threshold of each monitoring indicator, conduct full-element indicator anomaly pre-monitoring of the water conservancy monitoring object.

[0038] Abnormal indicator status refers to the state in which the data exceeds the warning or alarm threshold. For example, it can be excessive displacement, excessive uplift pressure, abnormal opening of structural joints, excessive scouring and silting, overload of opening and closing force, and excessive vibration and sway.

[0039] Anomaly pre-monitoring refers to judging whether there will be a trend of exceeding the standard in the future based on the prediction results. For example, it can be a trend prediction for the next 3 days, 7 days, or 15 days, or an advance risk prediction for the flood season or the high water level operation period.

[0040] The technical solution provided in this embodiment constructs a complete precision safety monitoring technology process for water conservancy projects through key technologies such as standardized processing of multi-source monitoring data, construction of a differentiated safety evaluation index system, a weight determination method combining expert weighting and consistency testing, threshold setting using the physical causal model coupled with confidence interval method, and time-series trend prediction, real-time anomaly identification, and early warning. This solution effectively unifies data standards, improves the relevance of indicators, enhances the objectivity of weight assignment, and improves the rationality of threshold setting. Ultimately, it achieves full-element coverage, automated operation, and early warning safety monitoring of water conservancy projects, significantly improving the safety management level and risk prevention capabilities during the operation of water conservancy projects.

[0041] In one embodiment, optionally, a differentiated full-element monitoring indicator system is constructed for at least two types of water conservancy monitoring objects, including: Primary indicators are extracted from the standard monitoring dataset, including at least one of the following: water level, deformation, seepage, riverbed scouring and silting, metal structure, and water turbine equipment. Based on the correlation between the primary and secondary indicators and the water conservancy monitoring objects, a full-element monitoring indicator system specifically for sluice gates and a full-element monitoring indicator system specifically for pumping stations are constructed respectively.

[0042] The primary indicators refer to the top-level monitoring indicators that reflect the major categories of engineering safety. Specifically, these can be indicators related to water level, deformation, seepage, scouring and silting, metal structures, and water turbine equipment.

[0043] Secondary indicators refer to specific monitoring indicators that can be directly collected after refinement. These can include absolute settlement value, displacement rate, uplift pressure difference, structural joint change rate, crater depth, siltation thickness, and the percentage of rated force of the hoist.

[0044] The correlation refers to the degree to which an indicator has a close impact on the safety of an engineering project. Specifically, it could mean that a sluice gate focuses more on monitoring cratering, while a pumping station focuses more on monitoring siltation and vibration. Alternatively, it could mean that water level difference is more critical to a sluice gate, while uplift pressure is more critical to a pumping station.

[0045] This technical solution extracts primary and secondary indicators hierarchically and constructs an indicator system according to the different structural characteristics and disease mechanisms of sluice gates and pumping stations. This makes the monitoring content more in line with the actual project, covers all key safety elements, avoids redundancy or omission of indicators, and improves the comprehensiveness, pertinence and scientific nature of monitoring.

[0046] In one embodiment, optionally, receiving expert weighting information and using a consistency check algorithm to determine the weight values ​​of each monitoring indicator in the indicator system includes: Obtain the score sets of multiple experts for the primary and secondary indicators, and calculate the initial weights; Calculate the Kendall synergy coefficient based on the initial weights to verify the degree of consistency of expert opinions; The degree of consensus among expert opinions is determined based on the Kendall consensus coefficient. When the Kendall consensus coefficient meets the preset consensus condition, the initial weight is determined as the final weight value of each monitoring indicator.

[0047] The scoring set refers to the set of scores formed by multiple experts rating the importance of each level of indicator. Specifically, it can be the scoring data of 5, 7, or 9 water conservancy safety experts, or it can be the scoring data of experts in design, operation and maintenance, and monitoring.

[0048] The Kendall coefficient of consistency is a statistical coefficient that measures the consistency of multiple evaluation results. Specifically, a value of 0.6 or higher indicates good consistency, 0.7 or higher indicates good consistency, and 0.8 or higher indicates high consistency.

[0049] The pre-set consistency condition refers to the threshold condition for judging the credibility of expert opinions. Specifically, it can be a Kendall coefficient greater than or equal to 0.7, or greater than or equal to 0.75 or 0.8.

[0050] This technical solution effectively reduces the subjective bias of a single expert by introducing scores from multiple experts and using Kendall's coefficient of consistency for verification. This ensures that the indicator weights are more objective and credible, making subsequent evaluation and early warning results more consistent with the actual safety status of the project and improving the overall reliability of the diagnosis.

[0051] In one embodiment, optionally, the method further includes: Based on the measured data and final weight values ​​of each monitoring indicator, the digital characteristics of the integrated cloud are calculated, including expectation Ex, entropy En, and hyperentropy He. The similarity between the comprehensive cloud and the four-level standard clouds (normal, basically normal, abnormal, and anomalous) is calculated. The four-level standard clouds are constructed based on the water conservancy project safety evaluation specifications and project operation experience, corresponding to the numerical ranges and cloud digital features of different health levels. Based on the principle of maximum similarity, the overall engineering health level of the water conservancy monitoring objects is determined.

[0052] Comprehensive cloud digital characteristics refer to parameters that characterize the overall distribution and fuzzy characteristics of indicator data, such as expectation Ex representing data center value, entropy En representing the degree of dispersion, and hyperentropy He representing the degree of fluctuation.

[0053] Level 4 standard cloud refers to a pre-constructed health level benchmark cloud, which can specifically be a cloud model with scores of 90-100 for normal, 75-90 for basically normal, 30-75 for abnormal, and 0-30 for anomalous.

[0054] The maximum similarity principle refers to matching the comprehensive cloud with the closest standard cloud to determine its level. Specifically, the cloud with the largest reciprocal of the weighted Euclidean distance can be used to determine the corresponding health level.

[0055] This technical solution transforms fuzzy health assessment into quantitative calculation through a cloud model. It uses Ex, En, and He to analyze data distribution and compares it with a four-level standard cloud to determine the overall health status. This achieves a combination of qualitative and quantitative methods, solving the problems of fuzziness and randomness in the safety assessment of water conservancy projects and improving diagnostic accuracy.

[0056] In one embodiment, optionally, the physical cause prediction model includes a statistical regression model constructed for deformation monitoring indicators; The statistical regression model is obtained by superimposing and combining water pressure, temperature, and time-related components, and is used to output the predicted value of the deformation index.

[0057] Statistical regression models refer to predictive models based on the deformation patterns of fitting data influenced by multiple factors. Specifically, they can be stepwise multiple regression models, multiple linear regression models, etc.

[0058] The water pressure component refers to the deformation component caused by upstream and downstream water levels, water level difference, uplift pressure, etc. Specifically, it can be the reservoir water pressure component, the dam foundation uplift pressure component, and the dam body uplift pressure component.

[0059] Temperature component refers to the deformation component caused by changes in air temperature and structural temperature. Specifically, it can be the daily air temperature component, the average air temperature component over the past 3 days, the cumulative temperature component, etc.

[0060] The aging component refers to the deformation component caused by concrete creep and foundation compression. Specifically, it can be a logarithmic aging component, a linear aging component, a combined aging component, etc.

[0061] This technical solution constructs a deformation physical cause model by superimposing three components: water pressure, temperature, and aging. This model fully reflects the real mechanism of deformation in hydraulic structures, making the predicted values ​​more consistent with actual stress and environmental changes, significantly improving the accuracy of deformation prediction, and providing a reliable basis for threshold determination.

[0062] In one embodiment, optionally, the step of determining the safety monitoring thresholds for each monitoring indicator using a physical cause prediction model combined with the confidence interval method includes: Based on the physical cause prediction model, the predicted values ​​of the monitoring indicators are obtained, and the residual sequence between the predicted values ​​and the measured values ​​is calculated. Calculate the mean and standard deviation from the residual sequence.

[0063] The safety monitoring thresholds for the monitoring indicators are formed by using ±2 times the standard deviation of the predicted value as the early warning threshold and ±3 times the standard deviation of the predicted value as the alarm threshold.

[0064] The residual sequence refers to the sequence of differences between predicted and measured values. Specifically, it can be one or more of the following: displacement residual sequence, uplift pressure residual sequence, and structural joint opening residual sequence.

[0065] The mean refers to the arithmetic mean of the residual sequence, specifically the mean of displacement residuals or the mean of uplift pressure residuals.

[0066] Standard deviation refers to the statistical value of the dispersion of the residual sequence, specifically the standard deviation of deformation residuals and the standard deviation of osmotic pressure residuals.

[0067] The warning threshold refers to the critical value that prompts attention. Specifically, it can be one or more of the following: vertical displacement warning, horizontal displacement warning, and uplift pressure warning threshold.

[0068] Alarm thresholds refer to the critical values ​​for determining danger, and can specifically include structural joint alarms, sludge and siltation alarms, and opening and closing force alarm thresholds.

[0069] This technical solution combines physical mechanism prediction with statistical confidence intervals, and uses residual distribution to scientifically define early warning and alarm thresholds. It takes into account both mechanical laws and statistical characteristics, making the thresholds more consistent with the long-term operation of the project, reducing false alarms and missed alarms, and improving monitoring reliability.

[0070] In one embodiment, optionally, the preset prediction model is the EEMD-IPSO-BiLSTM model; The prediction results are obtained by inputting real-time monitoring data and historical monitoring data into the model, including: The original time series of osmotic pressure was decomposed by EEMD to obtain several IMF components and residual trend terms.

[0071] An improved particle swarm optimization algorithm (IPSO) is used to optimize the learning rate and number of neurons in a BiLSTM network.

[0072] Each IMF component is input into the optimized BiLSTM network, and the predicted seepage pressure monitoring value is output.

[0073] Among them, EEMD refers to Ensemble Empirical Mode Decomposition, which is used to stabilize non-stationary time series data.

[0074] IMF components refer to the intrinsic mode function components obtained from decomposition, which can be high-frequency components, mid-frequency components, and low-frequency components.

[0075] IPSO stands for Improved Particle Swarm Optimization, used to optimize network hyperparameters.

[0076] BiLSTM stands for Bi-directional Long Short-Term Memory, which is used to capture temporal dependencies.

[0077] This technical solution uses EEMD to decompose non-stationary osmotic pressure sequences, IPSO to optimize hyperparameters, and BiLSTM to capture time-series features in both directions to form a high-precision prediction model. This significantly improves the prediction effect of non-stationary and nonlinear osmotic pressure data and provides accurate data support for advanced monitoring.

[0078] In one embodiment, optionally, the identification of abnormal states based on real-time data and thresholds includes: The real-time monitoring data is compared with the baseline value, early warning threshold, and alarm threshold simultaneously.

[0079] Based on the range of real-time monitoring data, the indicator status is marked as normal, warning, or danger.

[0080] Generate early warning information including early warning number, hydraulic structure name, measuring point number, measured value, and early warning time.

[0081] The benchmark value refers to the reference value of the initial or stable state of the project, which may be the initial elevation, initial displacement, or initial uplift pressure value.

[0082] Warning information refers to abnormal prompts used to notify administrators, which can be platform pop-up messages, SMS notifications, or system log records.

[0083] This technical solution achieves accurate anomaly identification through four-line comparison of benchmark value, measured value, early warning value, and alarm value, and automatically generates structured early warning information, enabling rapid anomaly location, automatic record keeping, and timely push notifications, thereby improving emergency response efficiency and realizing closed-loop management of hydraulic engineering safety.

[0084] To enable those skilled in the art to better understand this solution, this application also provides a preferred embodiment.

[0085] Based on the preceding analysis of defects in sluice gates and pumping stations, the main factors affecting project safety were summarized. Following relevant standards, a more reasonable safety evaluation index system for project monitoring was developed, as shown in the table below:

[0086] Methods for determining the weights of evaluation indicators; 1. Expert empowerment; Suppose there are p experts who score m primary indicators, and n secondary indicators are assigned weights based on the primary indicators. What is the weight of each secondary indicator? The following formula can be used for calculation:

[0087] In the formula: Let k be the score given by the k-th expert to the i-th primary indicator. Let be the weight of the j-th secondary indicator relative to its corresponding primary indicator.

[0088] The weights calculated based on the opinions of various experts will be averaged to obtain the final weights of each indicator.

[0089] Furthermore, to verify whether there are significant differences in expert opinions, which could lead to the calculated weights failing to accurately reflect the operational characteristics of the sluice gate, the Kendall coefficient was used to conduct a consistency evaluation of the calculation results. This coefficient can be calculated using the following formula:

[0090] In the formula: It is the average ranking of the j-th secondary indicator. W represents the average rank. When W approaches 0, it indicates poor consensus; when W approaches 1, it indicates a high degree of consensus among experts.

[0091] 2. Cloud model; The cloud model has significant advantages in handling uncertainty and fuzziness. It is primarily used to characterize the relationship between qualitative concepts and quantitative data. It describes the connotation and extension of a concept through three numerical features (expectation, entropy, and hyperentropy), exhibiting good flexibility and uncertainty handling capabilities. Specifically, the expected value (Ex) reflects the central position of the concept on the number line and is a numerical characteristic value describing the qualitative concept; entropy (En) reflects the uncertainty of the concept, describing the randomness and fuzziness of the qualitative concept in numerical entropy; and hyperentropy (He) reflects the randomness of entropy, describing the degree of uncertainty of the qualitative concept.

[0092] (1) Standard cloud digital feature calculation; Regarding the boundaries of qualitative concepts [I] min ,I max The digital characteristics of a standard cloud are calculated using the following formula:

[0093] In the formula: The k-value is used to measure the uncertainty of cloud droplet granularity, reflecting the uncertainty and randomness of the evaluation process. The k-value was determined to be 0.01 based on the ambiguity of the comments and experience.

[0094] (2) Calculation of integrated cloud digital features; For comprehensive cloud digital features, the following formula can be used for calculation:

[0095] In the formula: The number of samples; This is sample data.

[0096] (3) The cloud generator generates cloud droplets; By introducing randomness, the generated cloud droplets are ensured to be rationally distributed around the concept center, constructing a cloud map that comprehensively describes the characteristics of the evaluation object, thus enhancing the reliability and scientific rigor of the evaluation results. The generation process conforms to the following formula:

[0097] In the formula: Let be a random variable that follows a standard normal distribution.

[0098] (4) Comparison between standard cloud and integrated cloud; After determining the digital characteristics of the standard cloud and the integrated cloud, the weighted Euclidean distance between the two is calculated and its reciprocal is taken. This value is defined as the similarity between the two, and the evaluation level of the integrated cloud is determined based on the maximum similarity.

[0099] Let the comprehensive cloud parameters be... The i-th standard cloud parameter is Weighted Euclidean distance It can be calculated using the following formula:

[0100] In the formula: , , The weights for expectation, entropy, and hyperentropy are respectively set to 0.5, 0.3, and 0.2 based on the influence of the three feature values ​​on similarity.

[0101] Safety monitoring indicator settings; Currently, the main methods for formulating engineering safety monitoring indicators include the typical low-probability method, the confidence interval method, the limit state method, and the mechanical calculation method. The typical low-probability method combines the monitoring effect quantities generated by load combinations detrimental to strength and stability, and formulates monitoring indicators for structures such as dams based on existing observation data. The limit state method, depending on the method used to calculate the total effect S and resistance R of the critical load combination, is divided into the safety factor method, the first-moment extreme value state method, and the second-moment extreme value state method; the calculated monitoring indicators are mainly the extreme values ​​of this effect quantity. The mechanical calculation method uses an elastoplastic theoretical model to simulate the actual stress state of the project, analyzes the distribution of stress and strain at different stages, and thus formulates deformation monitoring indicators for each stage. The confidence interval method uses statistical theory or finite element calculation methods, based on existing engineering safety monitoring data, to establish a mathematical model between the monitoring effect quantity and the load, and uses the model to calculate the monitoring indicators for various loads. This is a common method for establishing early warning indicators based on monitoring data. The determination of safety monitoring indicators for this project mainly adopts the confidence interval method. Some monitoring items with smaller weights or fewer monitoring frequencies in the early stage are determined in combination with relevant design and management requirements.

[0102] 1. Determining the confidence interval; The confidence interval method, based on mathematical statistics, is a method for formulating monitoring indicators solely from a mathematical perspective. Its applicability presupposes that the sample data conforms to a certain probability distribution. Then, based on the small probability criterion, it finds the non-small probability intervals within this probability distribution. Its basic principle is as follows: Let the sample sequence y be a stationary sequence that satisfies a certain probability distribution, as follows:

[0103] Calculate the mean of the sample data y and standard deviation σy ; ; Then, by determining the significance level α (usually 1% to 5%), a confidence band with a probability of 100(1-α)% can be proposed. The monitoring indicator Em determined in this way is: ; In the formula: w – a multiple of σ determined based on the significance level α, usually taken as 3 or 2.

[0104] If the sample value yi falls within the range of the monitoring indicator Em, it is considered a safe value; otherwise, it is considered that a low-probability event has occurred, and its value is considered a dangerous value.

[0105] 2. Methods for formulating monitoring indicators; Monitoring data from hydraulic engineering structures typically exhibit periodicity and monotonic trends, failing to conform to a fixed probability distribution. Therefore, the confidence interval method cannot be directly applied to them. Consequently, a prediction model is usually required to determine the confidence interval. The prediction model describes the interaction between environmental quantities and effect quantities. If the error between the predicted model's estimate and the measured effect quantity does not conform to a normal distribution, it indicates that there is still some usable information in the error sequence, and the model cannot accurately describe the variation of the effect quantity. If the error sequence conforms to a normal distribution, it indicates that the prediction model performs well.

[0106] Therefore, the error sequence between the estimated values ​​of the prediction model and the measured values ​​usually follows a normal distribution. The confidence interval rule uses this error sequence as sample data to establish confidence intervals. The method for establishing engineering monitoring indicators in this way is as follows: Let the prediction model estimate be... The error value between the measured value yi (i=1, 2, ..., n) and the actual value yi. The sample sequence for: ; Calculate sample data The mean and standard deviation are respectively and ; In engineering monitoring, a confidence interval of 2 to 3 times the standard deviation is typically used. When it is 2 times... At that time, the measured value y i Falling The probability of the interval is 95.5%, which is used to estimate the population distribution along the regression line. Both sides The number of individuals within the range accounts for approximately 95.5%; this is approximately 3 times the number of individuals within the range. At that time, the measured value yi Falling The probability of the interval is 99.7%.

[0107] Therefore, the proposed early warning value Emy and alarm value Emb for the monitoring indicators are as follows: ; ; In the formula: —Predicted values ​​of the prediction model (i=n+1, n+2, ..., n+m); Since the prediction model can be used to describe the deformation pattern of the dam body, the predicted value is used to represent the deformation that the project should produce. This is used to judge whether the measured value deviates far from the predicted value. If the measured value falls within the range of the monitoring index Em, it means that the measured value does not deviate far from the predicted value, which is a normal phenomenon and the value is judged to be a safe value. Conversely, it means that the measured value deviates far from the predicted value, which is considered to be a low-probability event and the value is a dangerous value.

[0108] Predictive model building; Based on the effect quantity detection data and in conjunction with the environmental impact monitoring data, a mathematical model is established to quantitatively express and reasonably explain the state and variation law of the effect quantity, and to evaluate the safety status of water conservancy engineering structures.

[0109] The mathematical model for predictive analysis of monitoring quantities mainly reveals the changing patterns of monitoring effect quantities and the influence and extent of environmental quantities on them. Based on this, it predicts the future range of changes in effect quantities. It is generally a model that reflects the causal relationship between environmental quantities and effect quantities. The modeling process involves analyzing various environmental factors that affect the corresponding effect quantities, constructing the structural forms of each environmental impact component, and then determining the parameters in the expressions of each environmental impact component in the model using appropriate physical and mathematical methods based on the measured data of effect quantities and environmental quantities.

[0110] 1. Commonly used monitoring quantity prediction and analysis models; Based on the different methods for determining the undetermined parameters in the model, the predictive analysis models for safety monitoring quantities can be divided into three categories: statistical regression models, deterministic models, and hybrid models.

[0111] (1) Statistical model; A statistical model is a mathematical equation that quantitatively describes the changing patterns of monitored values, established through mathematical statistical analysis to determine the statistical relationship between the monitored effect quantity and the environmental impact component. In the mathematical model of the monitored effect quantity, the coefficients of each environmental component are mainly determined based on mathematical statistical analysis methods. Examples include statistical models for concrete dam deformation monitoring, arch dam deformation, stress-strain models of structures, statistical models for piezometric level at the phreatic line of earth-rock dams, and statistical models for seepage flow.

[0112] (2) Deterministic model; A deterministic monitoring model is a mathematical model describing the variation of monitored values, established by determining the relationship between the monitored effect quantity and environmental impact components through physical calculations. During modeling, a deterministic relationship between environmental impact variables and the monitored effect quantity is first constructed using physical theoretical calculations. Then, the assumptions and calculation parameters used in the physical calculations are reasonably adjusted based on measured values. In the mathematical model of the monitored effect quantity, the coefficients of each environmental component are mainly determined based on the physical calculation results. An example is the deterministic model of seepage pressure in earth-rock dams.

[0113] (3) Hybrid model; The monitoring hybrid model is a mathematical model describing the variation law of monitored values ​​by combining physical calculation results with mathematical statistical analysis methods to determine the relationship between the monitored effect size and the environmental impact components. During modeling, for environmental impact factors with a relatively clear relationship with the effect size, the corresponding physical theoretical calculation results are used to determine the parameters of the environmental impact component expression. For environmental impact factors with an unclear relationship with the effect size, or whose relationship is difficult to determine using physical theoretical calculation results, mathematical statistical methods are used to determine the parameters of the environmental impact component expression.

[0114] 2. Vertical and horizontal displacement prediction models; By comparing the structural and stress characteristics of several dam types, including concrete gravity dams, concrete arch dams, and earth-rock dams, with those of sluice gates and pumping station projects, it was found that sluice gate projects are quite similar to concrete arch dams, and pumping station projects are quite similar to concrete gravity dams. This study aims to construct stepwise multiple regression statistical models for pumping station and sluice gate projects, respectively, referencing the statistical models for concrete gravity dams and concrete arch dams and considering the characteristics of pumping station and sluice gate projects.

[0115] According to the deformation mechanism, the vertical displacement component can be divided into three parts: water pressure component. Temperature components and time-sensitive components ,Right now ; ①Water pressure component; Water pressure components can be divided into reservoir water pressure, dam foundation uplift pressure, and dam body uplift pressure. The mathematical expressions of water pressure components for different dam types are shown in the table below:

[0116] Based on the characteristics of sluice gate and pumping station projects, the water level difference between upstream and downstream is smaller compared to dams, so the influence of upstream and downstream water depths must be considered simultaneously. The water pressure component of this sluice gate and pumping station project adopts a combination of reservoir water pressure and dam uplift pressure, which can be expressed as: Pumping station project: ; Sluice gate project: ; In the formula: , , These are the fitting coefficients. , , These are the upstream water depth, downstream water depth, and reservoir water level (average upstream and downstream water level) at the time of monitoring, respectively, and the average reservoir water level (average upstream and downstream water level) of j days prior to monitoring.

[0117] ② Temperature component; Currently, there is basically no continuous temperature monitoring data for concrete structures in provincial water conservancy projects. Therefore, the model uses air temperature data to reflect the temperature component, which can be specifically represented as: Sluice gate and pumping station projects: ; In the formula: , , is the fitting coefficient, t is the cumulative number of days from the monitoring time to the initial measurement time, and Ti is the average temperature of the i days prior to the monitoring date.

[0118] ③Time-sensitive quantity; The causes of the time-dependent displacement in engineering projects are complex, comprehensively reflecting the creep, plastic deformation of the structural concrete and foundation, as well as the compressive deformation of geological structures. In generally normally operating projects, the time-dependent displacement exhibits a rapid initial change followed by gradual stabilization. To fully reflect the influence of time-dependent factors during initial water impoundment and years of operation, a combination of logarithmic and linear functions is proposed, specifically expressed as: Sluice gate and pumping station projects: ; In the formula: , These are the fitting coefficients. It is t / 100.

[0119] In summary, the statistical regression model for the vertical displacement of the sluice gate can ultimately be expressed as:

[0120] The statistical regression model for vertical displacement of the pumping station can ultimately be expressed as:

[0121] In the formula: is the fitting constant.

[0122] 3. Prediction model for uplift pressure and lateral seepage; Because seepage at dam foundations is influenced by various factors such as water level and rainfall, its seepage pressure values ​​generally exhibit non-stationary time series characteristics. Substituting these non-stationary time series into a machine learning model for training can lead to slow convergence and low accuracy during iteration. To address this, Empirical Mode Decomposition (EMD) can be used to decompose the non-stationary sequence, train each component using machine learning for prediction, and then linearly sum them. In machine learning, the selection of hyperparameters often significantly impacts prediction results. Manual parameter tuning is often time-consuming and laborious, failing to achieve ideal results. Therefore, introducing intelligent algorithms for optimization is recommended. Intelligent algorithms possess strong search capabilities, operate stably and efficiently, and can effectively improve parameter tuning efficiency. Based on the above analysis, this paper proposes an IPSO-BiLSTM seepage pressure prediction model based on Empirical Mode Decomposition (EMD). The seepage pressure sequence is decomposed into stationary components using EEMD, and then substituted into an IPSO-optimized BiLSTM network for prediction.

[0123] (1) Algorithm principle; ① Empirical Mode Decomposition (EMD); EEMD is an information extraction method that incorporates Gaussian white noise. This method utilizes the statistical characteristic of Gaussian white noise having a uniform frequency distribution, which can effectively suppress mode aliasing. Its processing steps are as follows: (a) Add white noise with a standard normal distribution to the original input feature sequence:

[0124] In the formula: This represents the white noise sequence added for the i-th time; This represents the additional noise signal in the i-th test.

[0125] (b) The obtained noisy signal Perform EMD decomposition separately to obtain the form of their respective IMF sums:

[0126] In the formula: The j-th IMF component is obtained after the i-th addition of white noise; is the residual function, representing the average trend of the signal; J is the number of IMF components.

[0127] (c) Repeat steps (a) and (b) M times, adding white noise signals of different amplitudes each time to obtain the set of IMF components.

[0128] (d) Utilizing the principle that the statistical mean of uncorrelated sequences is 0, the corresponding IMF components are subjected to ensemble averaging to obtain the final IMF components after EEMD decomposition. ,Right now:

[0129] ② An improved particle swarm optimization algorithm; Particle Swarm Optimization (PSO) is based on stochastic optimization of a swarm, where each particle has its own position vector. and velocity vector And the fitness determined by an objective function. All particles travel at a certain speed in the search space, following the currently found optimal value. To find the global optimum Traditional particle swarm optimization (PSO) algorithms are prone to getting stuck in local optima and have slow convergence speeds. This paper introduces a new adaptive particle velocity and position update strategy to improve the PSO algorithm.

[0130] In the formula: Inertial weight; , The learning factor is typically 2. , A random number that follows a uniform distribution in the range [0,1]. This represents the ratio of the current particle's fitness to the average fitness of all particles in the population.

[0131] when At this stage, when the algorithm is in its early stages or the particles are relatively dispersed, the global development capability of the algorithm is enhanced. When the algorithm is in its later stages or the particles are relatively concentrated, measures are taken to ensure the algorithm's local exploration capabilities.

[0132] ③ Bidirectional Long Short-Term Memory Neural Network; Long Short-Term Memory (LSTM) neural networks effectively handle long-term dependencies, avoiding the vanishing and exploding gradient problems of traditional RNNs. Their structure mainly consists of a forget gate, an input gate, and an output gate. Bidirectional LSTM networks capture richer information from sequences by combining two LSTMs, one forward and one backward. Their specific computational process can be represented as follows:

[0133] In the formula: Use the Sigmoid activation function; Here is the weight matrix for the forget gate; This is an offset for the forget gate. , These are the weighting coefficients for the output gate and the memory unit, respectively. , These are the biases for the input gate and the memory unit, respectively; This is a memory unit from the previous moment; This is the memory unit for the current moment. These are the weighting coefficients for the output gate; This is the offset value of the output gate; This represents the unit output at time t. , These are the outputs of the forward and reverse LSTM layers at time step t, respectively. This is the output of the BiLSTM at time step t.

[0134] (2) EEMD-IPSO-BiLSTM osmotic pressure prediction model; Based on the above principles and methods, an EEMD-IPSO-BiLSTM prediction model for osmotic pressure is proposed, and the calculation process is as follows: Figure 2 As shown. The specific implementation steps are as follows: ① The ensemble empirical mode EEMD was used to decompose the seepage pressure time series of each measuring point of the gate foundation, and each IMF component and trend term were divided into training set and output set in a 7:3 ratio; ② Determine the structure of the Bidirectional Long Short-Term Memory (BiLSTM) neural network model, and use the improved particle swarm optimization (IPSO) algorithm to optimize the learning rate and number of neurons in the BiLSTM model. ③ Use the optimized parameters to determine the final BiLSTM model, input each feature variable into the model for training and prediction, and output the prediction results; ④ The results of each component are superimposed, and the root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) are used to evaluate the model.

[0135] Figure 2 This is a schematic diagram of the workflow of the EEMD-IPSO-BiLSTM osmosis prediction model provided in the embodiments of this application, as shown below. Figure 2 As shown, the left side is the EEMD module (Data Decomposition and Denoising). The function of EEMD (Ensemble Empirical Mode Decomposition) is to decompose complex, non-stationary original time series data into a series of stationary IMF components and a residual term, thus solving the mode aliasing problem when directly modeling.

[0136] Process steps: Step 1, Input Data Time Series: Input raw monitoring data (such as seepage pressure time series).

[0137] Step 2, Input white noise and determine the amplitude and averaging order: Add Gaussian white noise to the data, and suppress mode aliasing by decomposing it multiple times and then averaging it.

[0138] Step 3, perform EMD decomposition: Perform empirical mode decomposition on the noisy data.

[0139] Step 4, calculate the average value of the IMF: After multiple decompositions, average the obtained intrinsic mode function (IMF) components.

[0140] Step 5, Update Residuals: Calculate the residual sequence and determine whether further decomposition is possible.

[0141] Step 6: Loop through and check if "can it be further decomposed": If it can be decomposed, return to step 2 and continue; if not, proceed to the next step.

[0142] Step 7, Output each IMF component and residual: Obtain the final stationary component and send it into the main process.

[0143] The intermediate main process (BiLSTM prediction model training and prediction) is the core prediction process of the model. It combines the components after EEMD decomposition and the parameters after IPSO optimization to complete the model training and prediction.

[0144] Process steps: Step 1, Input each IMF component + each feature variable: Use the IMF components output by EEMD, along with other feature data, as the model input.

[0145] Step 2, Divide the data into training and testing sets: Divide the data into training and testing sets according to the proportions for model training and validation.

[0146] Step 3, Construct BiLSTM structure: Build a bidirectional long short-term memory network to capture the dependencies between time-series data.

[0147] Step 4, Assign the optimal parameters to BiLSTM: Assign the optimal hyperparameters (learning rate, number of neurons, etc.) output by the IPSO module to the model.

[0148] Step 5, train the model network: train the BiLSTM model using the training set data.

[0149] Step 6, Model prediction output: Predict the test set data to obtain the prediction results.

[0150] Step 7, Model Evaluation: Use evaluation metrics (such as MAE, RMSE) to assess the prediction performance. The process ends here.

[0151] On the right is the IPSO module (BiLSTM hyperparameter optimization). The role of IPSO (improved particle swarm optimization) is to automatically find the optimal hyperparameters of the BiLSTM network, avoid the blindness of manual parameter tuning, and improve the prediction accuracy of the model.

[0152] Process steps: Step 1, Initialize particle velocity and position: Initialize the velocity and position of the particle swarm, with each particle representing a set of hyperparameters.

[0153] Step 2, update particle state and calculate fitness: train the model based on the hyperparameters corresponding to the particle position and calculate fitness (i.e. prediction error).

[0154] Step 3, obtain the individual optimal value of each particle: record the best position of each particle in history.

[0155] Step 4, obtain the global optimum of the swarm: record the optimal position throughout the history of the entire particle swarm.

[0156] Step 5, optimize particle velocity and position: update particle velocity and position based on individual optimality and global optimality.

[0157] Step 6, determine if the termination condition is met: if the number of iterations or the required accuracy are reached, output the optimal individual; otherwise, return to step 2 to continue iterating.

[0158] Step 7, Output the optimal individual: Output the optimal hyperparameter combination and feed it into the BiLSTM model of the main process.

[0159] The overall working logic is as follows: EEMD decomposes non-stationary data into stationary components, solving the problems of noise and mode mixing. IPSO finds the optimal hyperparameters for BiLSTM, improving model performance. The main BiLSTM process uses the decomposed data and optimized parameters to train the model and output prediction results. The combination of these three elements forms a complete solution for data decomposition, parameter optimization, and bidirectional time series prediction, which is very suitable for early warning scenarios involving complex time series data such as seepage pressure in water conservancy projects.

[0160] Safety monitoring indicators have been determined. In this study, safety monitoring indicators for all engineering safety evaluation and monitoring items, except for vertical displacement, horizontal displacement, base uplift pressure, and lateral seepage, were determined with reference to relevant design results, specifications, and daily management experience. For example, regarding sluice gate flow monitoring, during sluice gate operation and management, the gate opening is always operated according to the safe discharge curve. When the flow exceeds the safe discharge flow, there is a risk of scouring and damage to the downstream diversion canal. Simultaneously, in energy dissipation design, 1.5 times the design flow is often used as the check flow to verify the energy dissipation facilities. Therefore, when the check flow is exceeded, both the energy dissipation facilities and the downstream diversion canal are at risk of damage. Regarding structural joint monitoring, currently, most large and medium-sized provincial engineering projects use copper sheet waterstops for structural joints. Analysis of the waterstop structure shows that one side of the copper sheet is embedded in the concrete structure, while the other side is connected to the concrete structure through an asphalt groove. Therefore, the asphalt connection is a weak point in the waterstop structure. When the asphalt elongation exceeds the asphalt design ductility, the asphalt structure fails, and the structural seepage prevention system is at risk.

[0161] Through analysis and review, the safety monitoring indicators for each safety evaluation and monitoring item of the project are summarized as follows:

[0162] Where E is the predicted value of the prediction model, and Δ is the calculated confidence interval.

[0163] Evaluation level; Based on the above analysis, and to align with the current classification of sluice gate safety levels in the safety evaluation regulations, the health status of the project is divided into four levels, and the evaluation criteria can be expressed as follows: V = {Normal, Mostly Normal, Abnormal, Abnormal}; When establishing the partitioned mapping relationship between the evaluation set V and the membership degree within the range of [0, 100], both the objective connotation of the sluice gate health concept and the subjective attributes of human thinking patterns must be considered. Objectively, from the perspective of prioritizing engineering safety, the membership degree interval corresponding to the "normal" level should be strict, while the membership degree interval corresponding to the "abnormal" level should be lenient. Subjectively, the "passing grade" in human thinking can be used as the dividing point between "normal / basically normal" and "abnormal / abnormal," and "excellent" can be used as the dividing point between "normal." Therefore, the corresponding interval between the evaluation set and the membership degree in the sluice gate health diagnosis can be determined as follows: V = {Normal, Basically Normal, Abnormal, Abnormal} = {[100, 90],(90, 75],(75, 30],(30, 0]} This monitoring platform construction focuses on the content of full-element precision monitoring, accesses the monitoring data of all pilot projects, and upgrades and develops nine functional modules: (1) Automatic acquisition module: business database upgrade, automatic acquisition of data from automatic monitoring equipment, and access to existing commonly used monitoring data. (2) Monitoring data management module: information collection for various types of observation projects, realizing functions such as display, retrieval, and data compilation. (3) Monitoring data visualization: statistical display based on relevant core and key data. (4) Engineering 3D display: large-screen display based on 3D model and engineering monitoring data. (5) Monitoring data forecast and early warning: monitoring data is monitored, all abnormal data are intelligently identified and timely warnings are issued. (6) Monitoring data analysis and evaluation: analysis and statistical display of monitoring data. (7) Monitoring data prediction and simulation: trend model prediction analysis is performed through relevant prediction models, and simulation is performed based on the prediction. (8) System parameter setting: configuration and management of relevant basic business. (9) User management and other functional modules: are the foundation of the entire system operation. The overall architecture design for the precision monitoring of element engineering is divided into four layers. The system architecture is mainly based on service applications, platform capabilities, and infrastructure. The system is permeated by a technical standard and specification system and an information security assurance system to ensure the stability and security of the system.

[0164] (1) Access layer; The access layer primarily uses PCs as the platform and browsers as the access tool. Business applications are designed to be compatible with current mainstream browsers based on customer usage habits (browser version), improving user usability and ease of use.

[0165] (2) Service application layer; The service applications are the core of the business content of this project, including basic function construction, visualization content construction, and early warning model construction. Basic function construction mainly focuses on the collection, management, analysis, reporting / early warning, and other functions related to monitoring data, combined with predictive models for forecasting, and ultimately data visualization, engineering 3D visualization, and pre-simulation visualization.

[0166] (3) Platform capability layer; The platform capability layer forms the foundation of the service application layer, primarily comprising the microservice platform and the data management platform. This system is built on a microservice architecture, providing corresponding capability interfaces for upper-layer service applications. The data management platform mainly processes data, representing the system's data processing capabilities.

[0167] (4) Infrastructure layer; The infrastructure layer includes the basic environment, hardware, network, and basic software to provide the system operating environment, including servers, databases, operating systems, communication, storage devices, networks, and other related components.

[0168] (5) Standardization and safeguards; The system is permeated by technical standards and information security systems, which regulate the system development process and the system's stability and security.

[0169] The network architecture of this technical solution is divided into three horizontal and four vertical layers. The horizontal layers are: the provincial engineering management center layer, the management office layer, and the management station layer. Engineering site data is collected by the management offices. Based on actual conditions, the control area services and management area services of the management offices are respectively connected to the corresponding areas of the management offices.

[0170] Vertically, it is divided into: a production control area, an engineering management area, and an IoT area, then connects to the internet area via a unified internet exit at the departmental network level. The production control area is divided into real-time and non-real-time zones, isolated by firewalls. Strict physical isolation is implemented in the control area. Data transmission between the production control area and the engineering management area occurs at the management level via unidirectional isolation devices as needed. The network gateway primarily uses physical isolation technology to protect the control network system equipment. This is achieved by blocking direct network connections; two networks cannot be connected to the device simultaneously. Furthermore, logical network connections are blocked, and raw data is transmitted in a non-network manner. The isolation transmission mechanism is non-programmable, while still maintaining physical reverse isolation.

[0171] The management area is interconnected via a dedicated water conservancy network. This network is logically isolated from the internet via a firewall. The provincial engineering management center layer has no control area operations.

[0172] The comprehensive precision monitoring and management platform is deployed in management areas at the provincial engineering management center level and the management station level, and the collected data is synchronously transferred to the provincial engineering management center level through the water conservancy dedicated network.

[0173] The main functions of this precision observation platform for water conservancy projects include: automatic data acquisition module, monitoring data management, monitoring data visualization, 3D engineering display, monitoring data forecasting and early warning, monitoring data analysis and evaluation, monitoring data prediction and simulation, and system management.

[0174] The automatic data acquisition module is the primary source of data, automatically collecting relevant monitoring data to lay the foundation for subsequent data application and management. Monitoring data management mainly involves managing the collected data, including managing and displaying key data from various observation items, as well as compiling data to empower business operations. Monitoring data visualization provides user-friendly visualizations of key data. Monitoring data forecasting and early warning provides forecasts and warnings for relevant monitoring data, along with an automatic "diagnosis" process. Monitoring data analysis and evaluation performs statistical analysis of key monitoring data types and provides health assessments. Monitoring data prediction and simulation utilizes relevant prediction models to predict the future state of observation data and presents this prediction through visualization simulations. System management is the foundation of this platform construction, including user management, system parameter configuration, and management of basic information.

[0175] Example 2 Figure 3 This is a schematic diagram of the structure of a precision monitoring device for water conservancy engineering elements provided in Embodiment 2 of this application, as shown below. Figure 3 As shown, the device includes: The standard dataset construction module 301 is used to acquire raw data from multi-source monitoring of water conservancy projects, and to standardize the raw data to obtain a standard monitoring dataset; wherein, the raw data from multi-source monitoring includes at least one of sensing data collected by sensing devices, environmental quantity data, and manually supplemented data; The indicator system construction module 302 is used to construct a differentiated full-element monitoring indicator system for at least two types of water conservancy monitoring objects based on the standard monitoring dataset. The weight value determination module 303 is used to receive expert weighting information and use a consistency test algorithm to determine the weight value of each monitoring indicator in the indicator system. The safety monitoring threshold determination module 304 is used to determine the safety monitoring threshold of each monitoring indicator based on the indicator system and the weight value of each monitoring indicator, using a physical cause prediction model combined with the confidence interval method. The prediction result determination module 305 is used to collect real-time monitoring data of each monitoring indicator, input the real-time monitoring data and the historical monitoring data of the corresponding indicator into a preset prediction model, and obtain the prediction result of the monitoring value of each monitoring indicator. The anomaly identification and prediction module 306 is used to identify whether there is an abnormal state of the water conservancy monitoring object based on the real-time monitoring data of each monitoring indicator and the safety monitoring threshold of each monitoring indicator; and to perform full-element indicator anomaly pre-monitoring of the water conservancy monitoring object based on the monitoring value prediction results of each monitoring indicator and the safety monitoring threshold of each monitoring indicator.

[0176] The precision monitoring device for water conservancy engineering elements provided in this embodiment corresponds to the method embodiment described above, and has corresponding functional modules and beneficial effects. To avoid repetition, it will not be described again here.

[0177] Example 3 Figure 4 This is a schematic diagram of the structure of a monitoring and management platform provided in Embodiment 3 of this application, as shown below. Figure 4 As shown, the monitoring and management platform includes a data acquisition terminal 401, a data processing terminal 402, and an anomaly response terminal 403; The data acquisition terminal 401 is used for: Acquire raw data from multi-source monitoring of water conservancy projects. The raw data from multi-source monitoring includes sensing data collected by sensing devices, environmental quantity data, and manually supplemented data. Send the raw data from multi-source monitoring to the data processing terminal. The data processing terminal 402 is used for: The multi-source monitoring raw data is standardized to obtain a standard monitoring dataset. Based on the standard monitoring dataset, a differentiated full-element monitoring indicator system is constructed for at least two types of water conservancy monitoring objects. Expert weighting information is received, and a consistency check algorithm is used to determine the weight value of each monitoring indicator in the indicator system. Based on the indicator system and the weight values ​​of each monitoring indicator, a physical cause prediction model combined with the confidence interval method is used to determine the safety monitoring threshold of each monitoring indicator. Real-time monitoring data of each monitoring indicator is collected, and the real-time monitoring data and the historical monitoring data of the corresponding indicator are input into a preset prediction model to obtain the monitoring value prediction results of each monitoring indicator. Based on the real-time monitoring data of each monitoring indicator and the safety monitoring threshold, it is identified whether there is an abnormal state of the indicator. Based on the monitoring value prediction results of each monitoring indicator and the safety monitoring threshold, full-element indicator anomaly pre-monitoring is performed on the water conservancy monitoring object. The exception response terminal 403 is used for: When an abnormal state of an indicator is identified or an abnormal trend is detected, abnormal alarm information and pre-monitoring warning information are output, and the corresponding abnormal handling and feedback process is executed.

[0178] The monitoring and management platform can be used to implement each process of the above-mentioned precise monitoring method embodiment for water conservancy project elements, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0179] Example 4 This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described method for precise monitoring of water conservancy engineering elements and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0180] The processor mentioned above is the processor in the monitoring and management platform described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0181] Example 5 This application also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described embodiment of the precise monitoring method for water conservancy engineering elements, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0182] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0183] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element. Furthermore, it should be noted that the scope of the methods and systems in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0184] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0185] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above, which are merely illustrative and not restrictive. Those skilled in the art, under the guidance of this application, can make many modifications without departing from the spirit and scope of the claims, all of which fall within the protection scope of this application.

[0186] The above description is merely a preferred embodiment and the technical principles employed in this application. This application is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions that can be made by those skilled in the art will not depart from the scope of protection of this application. Therefore, although this application has been described in detail through the above embodiments, this application is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of this application, the scope of which is determined by the scope of the claims.

Claims

1. A method for precision monitoring of a hydraulic engineering element, characterized in that, The method includes: Obtain raw data from multi-source monitoring of water conservancy projects, and standardize the raw data to obtain a standard monitoring dataset; wherein the raw data from multi-source monitoring includes at least one of sensing data collected by sensing devices, environmental quantity data, and manually supplemented data; Based on the standard monitoring dataset, a differentiated full-element monitoring indicator system is constructed for at least two types of water conservancy monitoring objects; Receive expert weighting information and use a consistency test algorithm to determine the weight value of each monitoring indicator in the indicator system; Based on the aforementioned indicator system and the weight values ​​of each monitoring indicator, the safety monitoring threshold of each monitoring indicator is determined by using a physical cause prediction model combined with the confidence interval method. Collect real-time monitoring data for each monitoring indicator, input the real-time monitoring data and the historical monitoring data of the corresponding indicator into a preset prediction model, and obtain the prediction results of the monitoring values ​​of each monitoring indicator. For the water conservancy monitoring object, based on the real-time monitoring data of each monitoring indicator and the safety monitoring threshold of each monitoring indicator, it is possible to identify whether there is an abnormal state of the indicator; and based on the monitoring value prediction results of each monitoring indicator and the safety monitoring threshold of each monitoring indicator, it is possible to conduct full-element indicator anomaly pre-monitoring of the water conservancy monitoring object.

2. The method according to claim 1, characterized in that, The aforementioned differentiated full-element monitoring indicator system, constructed for at least two types of water conservancy monitoring objects, includes: Primary indicators are extracted from the standard monitoring dataset, including at least one of the following: water level, deformation, seepage, riverbed scouring and silting, metal structure, and water turbine equipment. Secondary indicators are derived from primary indicators, including at least one of the following: vertical displacement, horizontal displacement, structural joint opening, base uplift pressure, lateral seepage, opening and closing force, and oscillation. Based on the correlation between the primary and secondary indicators and the water conservancy monitoring objects, a full-element monitoring indicator system specifically for sluice gates and a full-element monitoring indicator system specifically for pumping stations are constructed respectively.

3. The method according to claim 1, characterized in that, The process of receiving expert weighting information and using a consistency check algorithm to determine the weight values ​​of each monitoring indicator in the indicator system includes: Obtain the score sets of multiple experts for the primary and secondary indicators, and calculate the initial weights; Calculate the Kendall synergy coefficient based on the initial weights to verify the degree of consistency of expert opinions; The degree of consistency of expert opinions is determined based on the Kendall synergy coefficient. When the Kendall synergy coefficient meets the preset consistency condition, the initial weight is determined as the final weight value of each monitoring indicator.

4. The method according to claim 3, characterized in that, The method further includes: Based on the measured data and final weight values ​​of each monitoring indicator, the digital characteristics of the integrated cloud are calculated, including expectation Ex, entropy En, and hyperentropy He. The similarity between the comprehensive cloud and the four-level standard clouds (normal, basically normal, abnormal, and anomalous) is calculated. The four-level standard clouds are constructed based on the water conservancy project safety evaluation specifications and project operation experience, corresponding to the numerical ranges and cloud digital features of different health levels. Based on the principle of maximum similarity, the overall engineering health level of the water conservancy monitoring objects is determined.

5. The method according to claim 1, characterized in that, The physical cause prediction model includes a statistical regression model constructed for deformation monitoring indicators; The statistical regression model is obtained by superimposing and combining water pressure, temperature, and time-related components, and is used to output the predicted value of the deformation index.

6. The method according to claim 1, characterized in that, The method of determining the safety monitoring thresholds for each monitoring indicator using a physical cause prediction model combined with the confidence interval method includes: Based on the physical cause prediction model, the predicted values ​​of the monitoring indicators are obtained, and the residual sequence between the predicted values ​​and the measured values ​​is calculated. Calculate the mean and standard deviation from the residual sequence; The safety monitoring thresholds for the monitoring indicators are formed by using ±2 times the standard deviation of the predicted value as the early warning threshold and ±3 times the standard deviation of the predicted value as the alarm threshold.

7. The method according to claim 1, characterized in that, The pre-set prediction model is the EEMD-IPSO-BiLSTM model; The real-time monitoring data and the historical monitoring data of the corresponding indicators are input into a preset prediction model to obtain the predicted monitoring values ​​of each indicator, including: The original time series of osmotic pressure was decomposed by EEMD to obtain several IMF components and residual trend terms; The learning rate and number of neurons in the BiLSTM network are optimized using the improved particle swarm optimization algorithm IPSO. Each IMF component is input into the optimized BiLSTM network, and the predicted seepage pressure monitoring value is output.

8. The method according to claim 1, characterized in that, Based on the real-time monitoring data of each monitoring indicator and the safety monitoring threshold of each monitoring indicator, identify whether there are any abnormal states of the indicators, including: The real-time monitoring data is compared simultaneously with the baseline value, early warning threshold, and alarm threshold. Based on the range of real-time monitoring data, the indicator status is marked as normal, warning, or danger; Generate early warning information including early warning number, hydraulic structure name, measuring point number, measured value, and early warning time.

9. A precision monitoring device for water conservancy engineering elements, characterized in that, The device includes: A standard dataset construction module is used to acquire raw data from multi-source monitoring of water conservancy projects, and to standardize the raw data to obtain a standard monitoring dataset; wherein, the raw data from multi-source monitoring includes at least one of sensing data collected by sensing devices, environmental quantity data, and manually supplemented data; The indicator system construction module is used to construct a differentiated full-element monitoring indicator system for at least two types of water conservancy monitoring objects based on the standard monitoring dataset. The weight value determination module is used to receive expert weighting information and use a consistency test algorithm to determine the weight value of each monitoring indicator in the indicator system. The safety monitoring threshold determination module is used to determine the safety monitoring threshold of each monitoring indicator based on the indicator system and the weight value of each monitoring indicator, using a physical cause prediction model combined with the confidence interval method. The prediction result determination module is used to collect real-time monitoring data of each monitoring indicator, input the real-time monitoring data and the historical monitoring data of the corresponding indicator into a preset prediction model, and obtain the prediction result of the monitoring value of each monitoring indicator. The anomaly identification and prediction module is used to identify whether there is an abnormal state of the water conservancy monitoring object based on the real-time monitoring data of each monitoring indicator and the safety monitoring threshold of each monitoring indicator; and to perform full-element indicator anomaly pre-monitoring of the water conservancy monitoring object based on the monitoring value prediction results of each monitoring indicator and the safety monitoring threshold of each monitoring indicator.

10. A monitoring and management platform, characterized in that, This includes a data acquisition terminal, a data processing terminal, and an anomaly response terminal; The data acquisition terminal is used for: Acquire raw data from multi-source monitoring of water conservancy projects. The raw data from multi-source monitoring includes sensing data collected by sensing devices, environmental quantity data, and manually supplemented data. Send the raw data from multi-source monitoring to the data processing terminal. The data processing terminal is used for: The multi-source monitoring raw data is standardized to obtain a standard monitoring dataset. Based on the standard monitoring dataset, a differentiated full-element monitoring indicator system is constructed for at least two types of water conservancy monitoring objects. Expert weighting information is received, and a consistency check algorithm is used to determine the weight value of each monitoring indicator in the indicator system. Based on the indicator system and the weight values ​​of each monitoring indicator, a physical cause prediction model combined with the confidence interval method is used to determine the safety monitoring threshold of each monitoring indicator. Real-time monitoring data of each monitoring indicator is collected, and the real-time monitoring data and the historical monitoring data of the corresponding indicator are input into a preset prediction model to obtain the monitoring value prediction results of each monitoring indicator. Based on the real-time monitoring data of each monitoring indicator and the safety monitoring threshold, it is identified whether there is an abnormal state of the indicator. Based on the monitoring value prediction results of each monitoring indicator and the safety monitoring threshold, full-element indicator anomaly pre-monitoring is performed on the water conservancy monitoring object. The exception response terminal is used for: When an abnormal state of an indicator is identified or an abnormal trend is detected, abnormal alarm information and pre-monitoring warning information are output, and the corresponding abnormal handling and feedback process is executed.

Images