Intelligent monitoring method of water conservancy project data based on digital twin

Abstract

This invention discloses an intelligent monitoring method for water conservancy engineering data based on digital twins, specifically relating to the field of intelligent monitoring technology for water conservancy engineering. The method includes the following steps: acquiring control command data and real-time monitoring data of the actual water conservancy engineering, analyzing their temporal characteristics and trends, detecting whether the control command data is transmitted back and participates in system calculations, and identifying data echo effects; after detection, calculating short-term and long-term impact indices based on the convergence of prediction errors, evaluating the impact of echo effects on the current operating status and future scheduling accuracy, and determining the echo effect level as low, medium, or high based on the degree of impact, and employing corresponding optimization strategies; this invention achieves dynamic evaluation and hierarchical determination of data echo effects, improving the system's early warning and adaptability; employing fuzzy logic reasoning to optimize decision-making, enhancing the intelligence of regulation; and reconstructing models through differentiated strategies and reliable data sources to ensure the operational stability and scheduling accuracy of water conservancy engineering.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology for water conservancy projects, and more specifically, to an intelligent monitoring method for water conservancy project data based on digital twins. Background Technology

[0002] With the development of smart water conservancy and intelligent dispatching systems, the management and control of water conservancy projects are transforming from traditional manual operations and single-point monitoring to digitalization, intelligence, and automation. Digital twin technology, with its advantages of virtual-real fusion, real-time simulation, and intelligent prediction, has become one of the core technologies in modern water conservancy project management. Through digital twin models, the operational status of the physical system of a water conservancy project can be reflected in real time, assisting decision-makers in formulating scientific control strategies and achieving automated control. However, in the practical application of existing digital twin systems, the system typically relies heavily on feedback data from the physical system as input for prediction and control, and the optimization and adjustment of control strategies are often based on historical feedback information. If the system fails to effectively separate the influence of feedback data during operation, it can easily lead to abnormal loops between control commands and feedback data. When the data of the control commands is fed back and re-participates in model calculations, continuing to influence the generation of subsequent control strategies, it can easily trigger a data echo effect.

[0003] Data echo effects not only lead to the continuous accumulation of prediction errors but may also cause digital twin models to lose their ability to accurately perceive actual operating conditions, resulting in reduced effectiveness of control strategies. This can lead to abnormal water resource scheduling or unstable operation of hydraulic structures, increasing system operational risks. Current technologies lack a comprehensive echo effect detection and suppression mechanism, especially in data fusion models where feedback data carries high weight, making it difficult to accurately determine whether feedback data forms loops and affects system judgments. Furthermore, existing intelligent control systems for water conservancy projects often rely solely on error analysis at a single time scale, failing to comprehensively assess the echo effect at both the short-term fluctuation and long-term cumulative risk levels. Therefore, there is an urgent need to design a multi-level, intelligent echo effect detection and suppression method to ensure the scheduling efficiency and safety of water conservancy engineering systems. Summary of the Invention

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A method for intelligent monitoring of water conservancy project data based on digital twins includes the following steps:

[0006] During the operation of the digital twin system, control command data for feedback and regulation of water conservancy project operation parameters are acquired, and real-time monitoring data of the physical water conservancy project are collected simultaneously.

[0007] By comparing the temporal characteristics and trends of control command data and feedback data from physical water conservancy projects, it is determined whether there is a loop phenomenon where control command data is transmitted back and used as input data to participate in the calculation of the digital twin system, so as to detect whether the data echo effect occurs.

[0008] If the data echo effect is detected, a comprehensive judgment is made based on the preset test parameters. The test parameters are the convergence of the prediction error of the digital twin model, which is obtained by calculating the dynamic deviation between the current prediction value and the real-time monitoring value.

[0009] Based on the convergence of the prediction error, the short-term impact index and the long-term impact index are calculated respectively. The short-term impact index is used to assess the impact of the echo effect on the operation status of water conservancy projects in the current regulation cycle, and the long-term impact index is used to assess the cumulative impact of the echo effect on the accuracy of subsequent scheduling decisions.

[0010] Based on the short-term impact index and the long-term impact index, the level of the current data echo effect is determined. The echo effect level is divided into low, medium and high according to the degree of impact.

[0011] Based on the determined echo effect level, different data echo optimization strategies are used for processing.

[0012] In a preferred embodiment, determining whether there is a loop phenomenon where control command data is transmitted back and used as input data to participate in the system calculation again, in order to detect whether a data echo effect occurs, specifically includes the following steps:

[0013] Step a: Collect the control command dataset U(t) generated by the digital twin system at time t. U(t) includes control parameters u1, u2, and u3, which correspond to the sluice gate opening degree, flood discharge volume, and sluice gate opening and closing frequency, respectively.

[0014] Step b: Record the monitoring dataset S(t+Δt) fed back by the hydraulic engineering entity system at time t+Δt. S(t+Δt) includes the water level h, flow velocity v and gate state θ corresponding to the control parameters u1, u2 and u3, respectively.

[0015] Step c, calculate the similarity D1 between U(t) and S(t+Δt) using the dynamic time warping algorithm, with the following formula:

[0016] n is the total number of control parameter dimensions, which is also the total number of monitoring data dimensions, with a value of 3. max(|U|,|S|) is the normalization processing coefficient to ensure that D1 is dimensionless and to eliminate the influence of parameter dimensions. i is the parameter type number.

[0017] Step d: At time t+2Δt, the feedback data S(t+Δt) is input into the control command data U'(t+2Δt) generated by the digital twin system. The similarity D2 between U'(t+2Δt) and the original U(t) is calculated using the Pearson correlation coefficient R formula:

[0018] ; and Let represent the mean values ​​of each control parameter in U(t) along their respective dimensions;

[0019] Step e: When the product of D1 and D2 is greater than the set standard threshold, it is determined that there is a control command data feedback phenomenon, and the data echo effect is identified.

[0020] In a preferred embodiment, the specific steps for calculating the short-term impact index include:

[0021] Define a time window τ to evaluate the time range and divide it into multiple equally spaced time nodes to form a short-term time series. At each time node, collect three parameter values ​​predicted by the digital twin system, including water level prediction, flow velocity prediction, and gate opening and closing status prediction; at the same time, collect three monitoring parameter values ​​fed back by the physical water conservancy project system, including water level monitoring value, flow velocity monitoring value, and gate opening and closing status monitoring value.

[0022] At each time point, the predicted water level value is subtracted from the monitored water level value, and the absolute value of the difference is calculated. Then, this absolute value is divided by the monitored water level value and a very small positive number is added to avoid division by zero error. The same method is used to calculate the error of the predicted flow velocity value and the predicted gate opening and closing status value, respectively.

[0023] For each time point, three proportional coefficients are set for water level, flow velocity, and gate status, respectively reflecting their importance to the overall system safety or performance. The water level error, flow velocity error, and gate status error are weighted and averaged according to the proportional coefficients to obtain the comprehensive error value for that time point.

[0024] Within the entire time window τ, calculate the average of the comprehensive error values ​​of all time nodes. Then, for the comprehensive error value of each time node, calculate the square of the difference between it and the average value. Sum all the squared differences and divide by the total number of time nodes to obtain the error fluctuation amplitude within the short-term time period. Finally, multiply the error fluctuation amplitude by the preset sensitivity adjustment coefficient to obtain the short-term impact index.

[0025] In a preferred embodiment, the specific steps for calculating the long-term impact index include:

[0026] Define a long-term observation window T, which is much larger than the short-term window τ, and divide it into multiple equally spaced time nodes to form a long-term time series. At each time node, collect the three parameter values ​​predicted by the digital twin system, including the predicted water level, the predicted flow velocity, and the predicted gate opening and closing status. At the same time, collect the three monitoring parameter values ​​fed back by the physical water conservancy project system, including the monitored water level, the monitored flow velocity, and the monitored gate opening and closing status.

[0027] At each time point, the predicted water level is subtracted from the monitored water level, and the absolute value of the difference is calculated. This absolute value is then divided by the monitored water level, and a very small positive number is added to avoid division by zero. The same method is used to calculate the errors of the predicted flow velocity and the predicted gate opening / closing status.

[0028] For each time point, three proportional coefficients are set for water level, flow velocity, and gate status to reflect their respective impact on the long-term stability of the overall system. The water level error, flow velocity error, and gate status error are weighted and averaged according to the proportional coefficients to obtain the comprehensive error value for that time point.

[0029] The comprehensive error values ​​at all time points within a long-term period are summed and multiplied by the time interval to obtain the comprehensive error integral value, which reflects the degree of error accumulation within the long-term period.

[0030] The long-term impact index is obtained by dividing the integral value of the comprehensive error by the product of the total time length of the long-term period and the set maximum allowable error value. The long-term impact index is dimensionless. After normalization, the index is ensured to reflect the comparability between different water conservancy engineering systems, and finally the long-term impact index is obtained.

[0031] In a preferred embodiment, the level of data echo effect is determined by using fuzzy logic to infer based on the short-term impact index and the long-term impact index.

[0032] In a preferred embodiment, the principle of fuzzy logic is as follows:

[0033] The short-term and long-term impact indices corresponding to the occurrence of data echo effects are obtained and used as input variables for fuzzy logic. The corresponding levels of data echo effects are used as output variables for fuzzy logic. The input variables are fuzzified and their values ​​are converted into fuzzy sets. The output variables are also fuzzified and their values ​​are converted into fuzzy sets. Fuzzy rules are formulated to describe the degree of fit of different data echo effect corresponding levels under different combinations of data types. The fuzzified input variables are then used to infer the corresponding levels of data echo effects through fuzzy rules.

[0034] In a preferred embodiment, different data echo optimization strategies are employed for processing, including: in the low-level echo effect state, applying a data filtering algorithm to remove abnormal feedback data; in the medium-level echo effect state, adjusting the data fusion weight parameters to reduce the model's dependence on feedback data; and in the high-level echo effect state, resetting the model's input data source, forcibly introducing an external trusted monitoring data source for verification, and reconstructing the digital twin model state to completely eliminate the echo effect.

[0035] In a preferred embodiment, under the advanced echo effect level, the state reconstruction of the digital twin model is achieved by introducing a third-party trusted data source. The trusted data source includes data from UAV inspections, satellite remote sensing, and manual inspections. The reconstruction process includes: pausing the operation of the existing model, deleting the feedback link information of the current model, importing trusted data to initialize the modeling parameters, generating a new digital twin model, and verifying the model's effectiveness using historical verification data.

[0036] In a preferred embodiment, under the low echo effect level, the filtering algorithm used is adaptive Kalman filtering.

[0037] In a preferred embodiment, during the intermediate echo effect state, weight optimization adjustment is performed, specifically as follows:

[0038] The short-term impact index and the long-term impact index are normalized separately and converted into percentages.

[0039] Calculate the average of the short-term impact index and the long-term impact index as the overall degree of echo effect;

[0040] An adjustment coefficient is preset to determine the magnitude of the decrease in the fusion weight of the feedback data. The decrease value is obtained by multiplying the overall echo effect by the preset adjustment coefficient. The decrease value is then subtracted from the initial fusion weight of the feedback data, and the result is used as the new fusion weight of the feedback data.

[0041] The fusion weight of the predicted data is equal to one minus the fusion weight of the new feedback data.

[0042] The beneficial effects of the present invention are:

[0043] This invention, by introducing an analytical method combining short-term and long-term impact indices, achieves multi-dimensional, full-cycle dynamic evaluation of data echo effects in digital twin systems. This effectively solves the problem of detection lag or inaccuracy caused by relying solely on a single time scale in existing technologies. The short-term impact index can promptly capture predictive fluctuations and operational anomalies caused by data echoes within the current control cycle, while the long-term impact index comprehensively measures the error accumulation and model deviation trends caused by echo effects during long-term operation. Through the synergistic judgment of these dual indicators, the system's sensitivity to data echo effects and its early warning capabilities are enhanced, providing more timely and accurate risk identification methods for water conservancy project scheduling and management, and significantly improving the safety and stability of system operation.

[0044] This invention employs fuzzy logic reasoning, using short-term and long-term impact indices as input variables. Based on pre-defined fuzzy rules, it comprehensively judges the data echo effect level and selects differentiated response strategies accordingly. This overcomes the technical bottlenecks of existing methods, such as unclear echo effect level determination and limited treatment measures. Fuzzy logic effectively handles the fuzziness and uncertainty of the echo effect's impact, rationally classifying low-, medium-, and high-level echo effect states to ensure the system adopts optimal intervention measures under different risk levels. This mechanism significantly enhances the adaptability and autonomous control capabilities of digital twin systems to complex operating conditions, enabling intelligent management and refined control of water conservancy project operation, reducing the frequency of human intervention, and improving the scientific rigor and reliability of system decision-making.

[0045] This invention designs a hierarchical and differentiated data echo optimization strategy for different echo effect levels. Particularly in the intermediate and advanced echo effect states, it achieves accurate correction and rapid reconstruction of the digital twin system model by dynamically adjusting data fusion weights and introducing external reliable data sources. Specifically, in the intermediate state, by optimizing the fusion ratio of feedback data and prediction data, the system's dependence on echo data is reduced, maintaining the autonomy and accuracy of model predictions. In the advanced state, third-party data sources such as UAV inspections, satellite remote sensing, and manual inspections are introduced to reconstruct the digital twin model state, eliminating the model error propagation problem caused by the echo effect. This solution ensures the stable operation and precise control of the water conservancy engineering system in complex environments, effectively avoids prediction inaccuracies and scheduling anomalies caused by the echo effect, and improves the overall intelligence and automation level of engineering management. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the intelligent monitoring method for water conservancy engineering data based on digital twins in this invention. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0048] Reference Figure 1 The following examples were obtained:

[0049] Example 1: This invention is proposed against the backdrop of the ever-increasing demand for intelligent management of water conservancy projects and the widespread application of digital twin technology. With the expansion of water conservancy project scale and the increase in management complexity, traditional water conservancy project operation monitoring and control methods have gradually revealed problems such as slow response, insufficient data processing capabilities, and reliance on human experience for decision-making, making it difficult to meet the demands of modern water conservancy projects for efficient, safe, and intelligent operation. Digital twin technology, with its high degree of integration of virtual and reality, real-time dynamic simulation, and intelligent analysis and prediction capabilities, provides a novel intelligent management solution for water conservancy projects. However, in practical applications, digital twin systems, due to their high dependence on feedback data from the physical system, are prone to the echo effect of data feedback and re-participation in model calculations. This leads to abnormal loops between system control commands and feedback data, resulting in problems such as accumulated prediction errors and decision failures, seriously affecting the operational safety and control efficiency of water conservancy projects. Therefore, there is an urgent need for an intelligent monitoring method that can effectively detect and suppress the data echo effect in digital twin systems to ensure the stable operation and scientific control of water conservancy project systems.

[0050] This invention aims to provide a systematic and intelligent data echo effect detection and optimization mechanism by constructing an intelligent monitoring method for water conservancy engineering data based on a digital twin system. By collecting real-time operational data of the water conservancy engineering entity system and combining it with the prediction results of the digital twin model, the convergence degree of the system prediction error is dynamically analyzed, short-term and long-term impact indices are calculated, and the data echo effect level is comprehensively determined. Based on different levels, corresponding optimization strategies are adopted, effectively reducing the system's dependence on feedback data and avoiding system control instability and decision-making bias caused by the data echo effect. This method enables intelligent and precise monitoring and control of the operational status of water conservancy projects, improving the safety and reliability of system operation, and is widely applicable to various water conservancy engineering management scenarios such as reservoir scheduling, river management, and canal irrigation.

[0051] The method disclosed in this invention not only enables early detection and timely intervention of data echo effects, but also improves the system's adaptability to complex working conditions through fuzzy logic reasoning, and introduces diversified external reliable data sources as the basis for model reconstruction, thereby enhancing the robustness and emergency response capability of the digital twin system under extreme hydrological conditions. It has high engineering application value and promotion prospects.

[0052] A digital twin system, in the field of water conservancy engineering, refers to a comprehensive intelligent monitoring and control platform built by integrating physical entity systems with virtual simulation models and real-time data acquisition and processing technologies. This system uses the digital twin model as its core tool, providing real-time status perception, decision analysis, and feedback control functions through dynamic simulation and prediction of the water conservancy project's operation. The digital twin model is the core component of the digital twin system, specifically a virtual simulation model built based on mathematical models, physical laws, engineering parameters, and historical operational data, used to accurately reflect and predict the operational status and behavior of the water conservancy project entity system. The water conservancy project entity system refers to actual physical engineering facilities including reservoirs, dams, canals, and sluices. These facilities perform functions such as water resource allocation, flood control, irrigation, and water supply in the actual environment. Their real-time operational status is collected through sensors and monitoring equipment and transmitted to the digital twin system, achieving dynamic mapping and interaction between the virtual and real worlds.

[0053] During the operation of the digital twin system, control command data for feedback regulation of water conservancy project operating parameters is acquired, and real-time monitoring data of the physical water conservancy project is collected simultaneously. The control command data is generated by the digital twin model based on predictions and analyses of the water conservancy project system's operating status, including real-time settings for key regulation parameters such as sluice gate opening degree, flood discharge volume, and gate opening and closing frequency. Real-time monitoring data of the physical water conservancy project is collected through a multi-source sensor network deployed in hydraulic structures, channels, and water bodies, covering core feedback data such as water level, flow velocity, and gate status. This step ensures closed-loop data interaction between the digital twin system and the physical system, providing a fundamental data source for subsequent analysis of the relationship between control command data and feedback data. By achieving real-time synchronous acquisition of control commands and feedback monitoring data, the system can comprehensively grasp the dynamic correspondence between the current project operating status and control response, laying a data foundation for accurately judging data echo effects.

[0054] By comparing the temporal characteristics and trends of control command data and feedback data from the actual water conservancy project, this study determines whether there is a cyclical phenomenon where control command data is fed back and used as input data to participate in the digital twin system calculation, thus detecting the data echo effect. The process employs a dynamic time warping algorithm to analyze the temporal series matching degree of the control commands and feedback data, while combining this with the Pearson correlation coefficient to assess the consistency of the numerical trends of the two sets of data. When the similarity between the two reaches a set threshold and exhibits a closed-loop trend in time delay, it indicates that the control command data may have been fed back in the feedback loop, acting as a re-input to the digital twin model, resulting in a high correlation and synchronization between the prediction results and the actual feedback data. This step is of great significance in digital twin systems. It can effectively identify echo effects caused by improper feedback path design or excessive reliance on system feedback, prevent the system from falling into a cycle of misjudgment where prediction and feedback reinforce each other, and improve the robustness and reliability of intelligent control systems for water conservancy projects.

[0055] When a data echo effect is detected, a comprehensive judgment is made based on preset verification parameters. The verification parameter is the convergence of the prediction error of the digital twin model, obtained by calculating the dynamic deviation between the current predicted value and the real-time monitored value. When the system detects a potential data echo effect, it needs to further comprehensively assess the severity of the echo effect to determine whether intervention is necessary and what optimization strategy to adopt. This invention proposes using the convergence of prediction error as a key verification parameter. By calculating the dynamic deviation between the predicted value and the real-time monitored value of the digital twin system, the change of prediction error over time is analyzed. If the prediction error continuously increases or shows an unstable trend over multiple time periods, it indicates that the system's prediction ability may be interfered with by the data echo effect, leading to continuous error accumulation. If the prediction error rapidly converges to a small range within a short period, it indicates that the system can still effectively adjust the prediction strategy and has not been severely affected. The system can set different error convergence thresholds as standards for judging the degree of influence of the echo effect. The significance of this step is to establish a quantitative evaluation system for the degree of influence of the echo effect, enabling the system to determine whether to adjust its operating strategy based on real-time calculation results, thereby improving the adaptive control capability of the digital twin system.

[0056] Based on the convergence of prediction errors, short-term and long-term impact indices are calculated separately. The short-term impact index is used to assess the impact of echo effects on the operational status of water conservancy projects within the current control cycle, while the long-term impact index is used to assess the cumulative impact of echo effects on the accuracy of subsequent scheduling decisions. To comprehensively assess the impact of data echo effects on system operation, this invention proposes calculating short-term and long-term impact indices to quantify the degree of impact of data echo effects across different time dimensions. The short-term impact index primarily assesses the stability of the system's operational status within the current control cycle, i.e., a relatively short timeframe. A high short-term impact index indicates that the data echo effect has a significant impact on system prediction and control within the current control cycle, requiring rapid adjustment of the data fusion method or optimization of model parameters. The long-term impact index assesses the cumulative impact of echo effects on the future operation of the system. If the long-term impact index remains high for an extended period, it indicates that the data echo effect is gradually accumulating during system operation, potentially leading to increasingly larger prediction errors and ultimately affecting the scheduling accuracy of the entire water conservancy project. The significance of this step lies in enabling the system to identify not only sudden anomalies within a short period but also long-term trend deviations through dual-timescale assessment, providing comprehensive data support for subsequent optimization decisions.

[0057] Based on the short-term and long-term impact indices, the level of the current data echo effect is determined. The echo effect level is categorized into low, medium, and high levels according to the degree of impact. After calculating the short-term and long-term impact indices, the system needs to classify the severity of the data echo effect to implement corresponding optimization strategies. The echo effect levels are divided into three levels: low, medium, and high. A low-level echo effect indicates that the data echo has a relatively small impact on the system's prediction and control, requiring only appropriate adjustments to the data fusion ratio for optimization. A medium-level echo effect indicates that the data echo effect has affected the stability of the system's regulation, requiring data filtering or weight adjustments to reduce the system's dependence on feedback data. A high-level echo effect indicates that the system's prediction model has been severely disturbed, and the error cannot converge normally, necessitating more aggressive optimization measures, such as resetting the model's input data source or using reliable external data for model correction. The significance of this step is that it enables the system to take targeted optimization measures based on the degree of impact of the data echo effect, ensuring the stability of system operation and the reliability of decision-making.

[0058] Based on the determined echo effect level, different data echo optimization strategies are employed. Depending on the level of data echo effect, the system will adopt different optimization strategies to reduce the impact of data echo on system operation. Under low-level echo effect conditions, the system can apply an adaptive Kalman filter algorithm to eliminate abnormal feedback data and fine-tune the data fusion ratio to reduce the interference of data echo on control decisions. Under medium-level echo effect conditions, the system will adjust the data fusion weight parameters to reduce the participation of feedback data in the control command generation process, thereby enhancing the dominant role of the prediction model in decision-making and reducing the interference of data echo signals on the system. Under high-level echo effect conditions, the system will take more thorough optimization measures, including suspending the automated data fusion of the existing model, resetting the model input data source, and introducing a third-party trusted data source, such as UAV inspection data, satellite remote sensing data, or manual inspection data, to calibrate the model's predicted values ​​and ensure that the system can still accurately reflect the actual operating status of the water conservancy project under extreme conditions. The significance of this step is that, for different levels of data echo effect problems, the system can select the most appropriate optimization strategy to minimize the impact of data echo on the safety of water conservancy project regulation, while ensuring the adaptive capability and long-term stable operation of the digital twin system.

[0059] To determine whether there is a loop where control command data is transmitted back and used as input data in the system calculation, in order to detect the data echo effect, the following steps are included:

[0060] Step a: Collect the control command dataset U(t) generated by the digital twin system at time t. U(t) includes control parameters u1, u2, and u3, which correspond to the sluice gate opening degree, flood discharge volume, and sluice gate opening and closing frequency, respectively.

[0061] Step b: Record the monitoring dataset S(t+Δt) fed back by the hydraulic engineering entity system at time t+Δt. S(t+Δt) includes the water level h, flow velocity v and gate state θ corresponding to the control parameters u1, u2 and u3, respectively.

[0062] Step c, calculate the similarity D1 between U(t) and S(t+Δt) using the dynamic time warping algorithm, with the following formula:

[0063] n is the total number of control parameter dimensions, which is also the total number of monitoring data dimensions, with a value of 3. max(|U|,|S|) is the normalization processing coefficient to ensure that D1 is dimensionless and to eliminate the influence of parameter dimensions. i is the parameter type number.

[0064] Step d: At time t+2Δt, the feedback data S(t+Δt) is input into the control command data U'(t+2Δt) generated by the digital twin system. The similarity D2 between U'(t+2Δt) and the original U(t) is calculated using the Pearson correlation coefficient R formula:

[0065] ; and Let represent the mean values ​​of each control parameter in U(t) along their respective dimensions;

[0066] Step e: When the product of D1 and D2 is greater than the set standard threshold, it is determined that there is a control command data feedback phenomenon, and the data echo effect is identified.

[0067] In the dynamic interaction between a digital twin system and the physical system of a water conservancy project, the digital twin system continuously corrects and generates new control commands based on the currently collected feedback data. However, when the system fails to effectively isolate the direct coupling between feedback data and control commands, it may lead to the system continuously using feedback data to influence the generation of control commands, thus forming an abnormal loop between control command data and feedback data. This looping phenomenon, known as the data echo effect, causes the system's predictive model to lose its true perception of the actual state of the physical system, causing control commands and feedback data to fall into a closed loop of self-reinforcement. In severe cases, it may lead to the continuous accumulation of system prediction errors, loss of control over decision-making, and affect the safety and efficiency of water conservancy project scheduling.

[0068] By collecting control command data generated by the digital twin system and monitoring data fed back from the physical hydraulic engineering system within a preset time interval, and employing time-series similarity analysis, it is possible to effectively detect whether control command data is being transmitted back and re-input into the system. This detection method not only promptly identifies data echo effects but also provides a basis for subsequent control strategy adjustments and optimizations, ensuring the independence and stability of the interaction between the digital twin system and the physical system.

[0069] In this invention, to evaluate the correlation between control command data at different time points, the system needs to compare the similarity of control command datasets generated at different times. Specifically, when the system detects whether feedback data affects the generation of subsequent control command data, it needs to analyze the changing patterns of control parameters generated at different time points.

[0070] Dimensional mean calculation refers to calculating the average value of each control parameter in a control command dataset separately. Since control command datasets typically contain multiple different types of parameters, such as sluice gate opening degree, flood discharge capacity, and gate opening / closing frequency, these parameters have different dimensions and numerical ranges, making it impossible to directly calculate the overall mean. To eliminate interference between different parameter dimensions, statistical analysis must be performed on each parameter individually.

[0071] Specifically, within the system's defined time window, control command data is collected at multiple time points. For the sluice gate opening parameter, the system extracts the parameter's value at all time points, calculates the arithmetic mean of these values, and obtains the sluice gate opening's mean value over all dimensions. The same method is used to calculate the mean values ​​of the flood discharge and gate opening / closing frequency parameters. In this way, each control parameter has an independent average value, reflecting its overall level or trend over a specific time period.

[0072] In the Pearson correlation calculation process, the mean value for each dimension plays a role in eliminating data bias and revealing the true correlation between variables. By subtracting the mean value for each dimension, the system can more accurately assess the degree of linear correlation between control command data at different time points, avoiding the influence of numerical bias or different data scales on the correlation judgment.

[0073] This invention accurately detects the occurrence of data echo effects by comparing the temporal characteristics of control command data and feedback data, combined with dynamic time warping algorithm and Pearson correlation analysis. The dynamic time warping algorithm ensures the effectiveness of alignment and comparison of data across different time dimensions, while normalization and dimensionless calculation ensure the comparability of data with different parameter types. Pearson correlation analysis, through the introduction of dimension-wise means, makes the correlation determination between different control command datasets more objective and accurate. When the product of the similarity analysis results of control command data and feedback data, combined with the Pearson correlation analysis results, exceeds a set threshold, the system can accurately determine the existence of data echo effects. This judgment criterion considers the strong correlation between control and feedback, as well as the feedback loop characteristics of the system itself, effectively avoiding misjudgments and ensuring high reliability and security of the prediction and control decisions of the digital twin system.

[0074] The specific steps for calculating the short-term impact index include:

[0075] Define a time window τ to evaluate the time range and divide it into multiple equally spaced time nodes to form a short-term time series. At each time node, collect three parameter values ​​predicted by the digital twin system, including water level prediction, flow velocity prediction, and gate opening and closing status prediction; at the same time, collect three monitoring parameter values ​​fed back by the physical water conservancy project system, including water level monitoring value, flow velocity monitoring value, and gate opening and closing status monitoring value.

[0076] At each time point, the predicted water level value is subtracted from the monitored water level value, and the absolute value of the difference is calculated. Then, this absolute value is divided by the monitored water level value and a very small positive number is added to avoid division by zero error. The same method is used to calculate the error of the predicted flow velocity value and the predicted gate opening and closing status value, respectively.

[0077] For each time point, three proportional coefficients are set for water level, flow velocity, and gate status, respectively reflecting their importance to the overall system safety or performance. The water level error, flow velocity error, and gate status error are weighted and averaged according to the proportional coefficients to obtain the comprehensive error value for that time point.

[0078] Within the entire time window τ, calculate the average of the comprehensive error values ​​of all time nodes. Then, for the comprehensive error value of each time node, calculate the square of the difference between it and the average value. Sum all the squared differences and divide by the total number of time nodes to obtain the error fluctuation amplitude within the short-term time period. Finally, multiply the error fluctuation amplitude by the preset sensitivity adjustment coefficient to obtain the short-term impact index.

[0079] In this invention, the method for calculating the short-term impact index effectively reflects the influence of data echo effects on the operational status of the water conservancy project within a short period by dynamically evaluating the deviation fluctuations between the predicted values ​​of the digital twin system and the feedback values ​​of the actual water conservancy project within a set time window. Specifically, the sluice gate opening degree, flood discharge volume, and gate opening and closing frequency among the control parameters determine the water level, flow velocity, and gate status of the actual water conservancy project system, respectively. Among them, the sluice gate opening degree directly affects the water level regulation of the reservoir area, the flood discharge volume determines the flow velocity change of the downstream river channel, and the gate opening and closing frequency reflects the dynamic status and execution frequency of the hydraulic structure operation. Therefore, the digital twin system needs to focus on the real-time changes of the three core feedback parameters—water level, flow velocity, and gate status—during the predictive control process. The Short-Term Impact Index calculates the errors in water level prediction and monitoring, flow velocity prediction and monitoring, and gate status prediction and monitoring separately. At each time point, different weighting coefficients are assigned to these three errors, and a weighted average is calculated. The weighting coefficients are assigned using an assignment method, pre-assigning different weight values ​​to water level, flow velocity, and gate status based on the safety requirements and operational management rules of different water conservancy projects, reflecting their degree of influence on system stability. In this way, the Short-Term Impact Index can quantify the impact of data echo effects on the short-term stability and reliability of water conservancy project control systems, providing a scientific basis for timely optimization of control decisions.

[0080] The specific steps for calculating the long-term impact index include:

[0081] Define a long-term observation window T, which is much larger than the short-term window τ, and divide it into multiple equally spaced time nodes to form a long-term time series. At each time node, collect the three parameter values ​​predicted by the digital twin system, including the predicted water level, the predicted flow velocity, and the predicted gate opening and closing status. At the same time, collect the three monitoring parameter values ​​fed back by the physical water conservancy project system, including the monitored water level, the monitored flow velocity, and the monitored gate opening and closing status.

[0082] At each time point, the predicted water level is subtracted from the monitored water level, and the absolute value of the difference is calculated. This absolute value is then divided by the monitored water level, and a very small positive number is added to avoid division by zero. The same method is used to calculate the errors of the predicted flow velocity and the predicted gate opening / closing status.

[0083] For each time point, three proportional coefficients are set for water level, flow velocity, and gate status to reflect their respective impact on the long-term stability of the overall system. The water level error, flow velocity error, and gate status error are weighted and averaged according to the proportional coefficients to obtain the comprehensive error value for that time point.

[0084] The comprehensive error values ​​at all time points within a long-term period are summed and multiplied by the time interval to obtain the comprehensive error integral value, which reflects the degree of error accumulation within the long-term period.

[0085] The long-term impact index is obtained by dividing the integral value of the comprehensive error by the product of the total time length of the long-term period and the set maximum allowable error value. The long-term impact index is dimensionless. After normalization, the index is ensured to reflect the comparability between different water conservancy engineering systems, and finally the long-term impact index is obtained.

[0086] In this invention, the long-term impact index is calculated by continuously evaluating the cumulative error between the predicted values ​​of the digital twin system and the feedback values ​​of the hydraulic engineering entity system within a set long-term observation window. This reflects the impact of data echo effects on the overall operational accuracy and reliability of control strategies over a longer timescale. The long-term impact index measures the continuous accumulation trend of system prediction errors using error integrals, identifying the long-term accumulation of prediction model biases caused by data echo effects, thus indicating potential risks such as decision failures and control imbalances. This index eliminates differences in scale and parameter dimensions among different hydraulic engineering systems through normalization, ensuring the uniformity and comparability of analysis results across different application scenarios, and providing a standardized evaluation basis for the governance and optimization of data echo effects across engineering systems.

[0087] The three proportional coefficients used in the calculation of the long-term impact index are basically consistent with those used in the short-term impact index, reflecting the importance of water level, flow velocity, and gate status in the long-term stability analysis of the system. These proportional coefficients are set by assignment, based on factors such as the management needs of the water conservancy project, the safety level, and the operational characteristics of the project. For example, in flood control scheduling scenarios, the proportional coefficient of the water level parameter can be given a higher weight, while in water supply scheduling or irrigation management, the proportional coefficients of flow velocity and gate status may be more prominent. Although the proportional coefficients are consistent in the short-term and long-term impact indices, in specific applications, the setting value of the long-term proportional coefficient can be adjusted separately according to different needs of long-term operational risk assessment to enhance the sensitivity and responsiveness to the risk of long-term prediction error accumulation.

[0088] When determining the level of data echo effect, fuzzy logic is used to infer the results based on short-term and long-term impact indices. The principle of fuzzy logic is as follows: The short-term and long-term impact indices corresponding to the occurrence of a data echo effect are obtained and used as input variables for the fuzzy logic. The level of the data echo effect is used as the output variable. The input variables are fuzzified, converting their values ​​into fuzzy sets. Similarly, the output variables are fuzzified, converting their values ​​into fuzzy sets. Fuzzy rules are then formulated to describe the degree of fit between different data type combinations and their corresponding levels of data echo effect. The fuzzified input variables are then used to infer the corresponding level of the data echo effect through these fuzzy rules.

[0089] First, the two input variables, the short-term impact index and the long-term impact index, are fuzzified, and their value ranges are divided into several fuzzy linguistic value intervals, such as low, medium, and high. Each interval corresponds to a certain membership function to reflect the membership relationship of the current input data in different fuzzy intervals. Simultaneously, the output level of the data echo effect is also fuzzified, set to low, medium, and high levels, and fuzzy sets are established for each output variable. Based on the water conservancy project control logic and expert experience, the system formulates a fuzzy rule base, which describes the data echo effect level to be determined under different combinations of short-term and long-term impact indices. For example, when both the short-term and long-term impact indices are high, it is determined to be a high-level echo effect; when the short-term impact index is low and the long-term impact index is medium, it is determined to be a medium-level echo effect; if both are low, it is determined to be a low-level echo effect.

[0090] During the inference phase, the system matches fuzzy rules based on the fuzzy set membership degree corresponding to the current input data. Combined with the inference engine, all valid rules are comprehensively analyzed to ultimately obtain a fuzzy output of the data echo effect level. Finally, through a defuzzification process, the fuzzy output variables are converted into explicit level judgment results, guiding the system to adopt corresponding optimization strategies. This fuzzy logic inference process fully embodies the intelligent characteristics of multi-factor comprehensive evaluation, effectively improving the system's ability to judge complex echo effect phenomena and its response accuracy, ensuring the safety and stability of water conservancy project regulation.

[0091] The Short-Term Impact Index represents the immediate impact of data echo effects on the current operational status of a water conservancy project's physical system within a short timeframe. This index dynamically analyzes the error fluctuations between the digital twin system's predicted values ​​and the physical system's feedback values ​​within a short time window, reflecting the system's prediction accuracy and the stability of control command execution within a short period. The Short-Term Impact Index primarily measures whether data echo effects cause abnormal fluctuations in the system's state during the current control period, revealing whether the system faces risks such as increased instantaneous prediction deviations, control instability, or command failures due to data echoes. It is mainly used to assist in the rapid diagnosis and immediate adjustment of the current operational status during water conservancy project control, ensuring the system's safety and effectiveness within a short period.

[0092] The Long-Term Impact Index represents the cumulative impact of data echo effects on the overall predictive capability of digital twin systems and the accuracy of water conservancy project scheduling decisions over a longer time scale. This index analyzes the cumulative error between system predictions and feedback values ​​within a long-term observation window, integrating these errors to reflect whether the prediction model exhibits a trend of deviation from actual conditions due to data echo effects during long-term operation. The Long-Term Impact Index primarily measures the cumulative degree of system prediction errors and whether long-term control strategies lead to systemic instability or a continuous expansion of decision-making biases due to echo effects. It is mainly used to guide the long-term operational optimization, model correction, and adjustment of decision parameters for water conservancy projects, ensuring the overall stability and reliability of the system's control over long periods.

[0093] Different data echo optimization strategies are employed to address the following: in the low-level echo effect state, data filtering algorithms are applied to remove abnormal feedback data; in the medium-level echo effect state, data fusion weight parameters are adjusted to reduce the model's dependence on feedback data; in the high-level echo effect state, the model's input data source is reset, an external trusted monitoring data source is forcibly introduced for verification, and the digital twin model state is reconstructed to completely eliminate the echo effect.

[0094] In the advanced echo effect level, the state reconstruction of the digital twin model is achieved by introducing a third-party trusted data source. The trusted data source includes data from UAV inspection, satellite remote sensing, and manual inspection. The reconstruction process includes: pausing the operation of the existing model, deleting the current model feedback link information, importing trusted data to initialize the modeling parameters, generating a new digital twin model, and verifying the model's effectiveness using historical validation data.

[0095] Under low-level echo effects, the adaptive Kalman filter is employed. At this level, the adaptive Kalman filter effectively identifies and eliminates anomalous feedback data caused by data echoes. By dynamically adjusting the prediction error covariance and measurement error covariance in the system state estimation, the filter automatically optimizes its parameters based on real-time changes in the feedback data. When feedback data exhibits short-term anomalies or slight deviations due to low-level echo effects, the adaptive Kalman filter utilizes the dynamic error balancing mechanism between the system's predicted and observed values ​​to automatically reduce the level of trust in the anomalous feedback data, suppressing the impact of data echo effects on system state estimation. This ensures the stability and reliability of the control commands output by the digital twin model, preventing system control failures caused by small-scale echo interference.

[0096] In the intermediate echo effect state, the weights are optimized and adjusted as follows:

[0097] The short-term impact index and the long-term impact index are normalized separately and converted into percentages.

[0098] Calculate the average of the short-term impact index and the long-term impact index as the overall degree of echo effect;

[0099] An adjustment coefficient is preset to determine the magnitude of the decrease in the fusion weight of the feedback data. The decrease value is obtained by multiplying the overall echo effect by the preset adjustment coefficient. The decrease value is then subtracted from the initial fusion weight of the feedback data, and the result is used as the new fusion weight of the feedback data.

[0100] The fusion weight of the predicted data is equal to one minus the fusion weight of the new feedback data.

[0101] The range of adjustment coefficients is typically set or dynamically optimized based on the operational characteristics of the hydraulic engineering system, the sensitivity of the digital twin model, and historical operational data. Generally, the range of adjustment coefficients is set between zero and one, where zero indicates no adjustment of the fusion weights, while one indicates that when the echo effect reaches its limit, the system is allowed to reduce the fusion weights of the feedback data to the maximum extent possible, thereby strongly suppressing the echo effect. The adjustment coefficients can be determined in the following ways: First, based on historical data analysis, statistical analysis of the system's historical operational data is used to evaluate the system's response capability and control effect under different echo effect levels, forming an empirical adjustment coefficient range; second, using an expert knowledge base method, combining the judgments of experts in the field of hydraulic engineering on the system's control risks and the trustworthiness of the feedback data, to set weight adjustment strategies and adjustment coefficient ranges that meet the actual engineering needs.

[0102] Under intermediate echo effect conditions, optimizing the fusion weights of feedback and prediction data can effectively reduce the system's dependence on echo source and suppress further spread and accumulation of echo effects. Dynamically calculating the decrease in feedback data fusion weights allows for flexible control over the reliability of feedback data. By reducing the proportion of feedback data in the fusion process, the decision-making dominance of the prediction model is enhanced, effectively reducing the interference of echo data on the system's prediction and control accuracy, while maintaining the model's autonomy and robustness. This ensures the stable operation and safe regulation of the digital twin system under intermediate echo effect conditions.

[0103] Example 2: In the real-time operation monitoring of a large reservoir, a digital twin system was used to predict reservoir water level changes and regulate sluice gate openings. Under normal operating conditions, the system relied heavily on feedback data for real-time control. However, during heavy rain, the reservoir inflow and water level fluctuated drastically, and the feedback data changed rapidly. The system detected an abnormally high similarity between control command data and feedback data. By comparing the temporal characteristics of the control command data and feedback data using a dynamic time warping algorithm, a data echo effect was found. The system activated the echo effect detection mechanism, calculating short-term and long-term impact indices based on the prediction error convergence. Based on the index results, the system determined the current echo effect level to be medium, and subsequently reduced the fusion weight of feedback data in the control command generation process, enhancing the dominant role of the prediction model in decision-making. By adjusting the data fusion ratio, the risk of misjudgment caused by feedback data loops was successfully reduced, ensuring the safety and accuracy of reservoir flood discharge scheduling.

[0104] Example 3: In a multi-functional water conservancy project, a digital twin system monitors the opening and closing status of sluice gates and the downstream flow regulation effect over a long period. After running for a period of time, analysis of short-term and long-term impact indices revealed a persistent convergence problem in the long-term prediction error. Further testing showed an abnormally high Pearson correlation between control command data and feedback monitoring data, with the product judgment value exceeding a set threshold, confirming that the system had entered an advanced echo effect state. To prevent a continuous self-excited loop between the model and the feedback system, the system automatically implemented an advanced echo effect optimization strategy. The system suspended the existing model feedback link, interrupted feedback data input, and temporarily introduced a trusted third-party data source for status verification, including images captured by drone inspections, remote sensing data on water surface changes, and data obtained from manual inspections. The digital twin model state was reconstructed using this external data, and the prediction parameters were calibrated with the assistance of historical validation data, ultimately restoring the model to normal operation and effectively eliminating the echo effect.

[0105] Example 4: In a regional canal irrigation management system, a digital twin model is used to predict and control irrigation water flow, ensuring efficient utilization of farmland water resources. During the flood season, the system continuously monitors the error trends between the model and actual feedback data using short-term and long-term impact indices. The system detects a sudden increase in the short-term impact index, while the long-term impact index remains at a moderate level. Based on fuzzy logic reasoning, the system determines the current echo effect level to be moderate and initiates an echo effect optimization process. First, the system dynamically adjusts the data fusion weight parameters, reducing the participation of feedback data in predictive control decisions and shifting the fusion weight to be dominated by the predictive model. Second, adaptive Kalman filtering is performed on the real-time feedback data to eliminate detected abnormal feedback data points, preventing them from affecting control decisions. This optimization process ensures that the canal control system can operate stably and efficiently in the face of sudden changes in water conditions, avoiding water resource control chaos caused by data echo effects.

[0106] The above three implementation methods describe in detail the technical path and actual operation process of the present invention in solving the data echo effect problem from different water conservancy engineering scenarios such as reservoir scheduling, hub project management, and canal irrigation control, highlighting the application value and innovation of the invention.

[0107] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0108] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0109] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0110] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0111] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for intelligent monitoring of water conservancy engineering data based on digital twins, characterized in that, Includes the following steps: During the operation of the digital twin system, control command data for feedback and regulation of water conservancy project operation parameters are acquired, and real-time monitoring data of the physical water conservancy project are collected simultaneously. By comparing the temporal characteristics and trends of control command data and feedback data from physical water conservancy projects, it is determined whether there is a loop phenomenon where control command data is transmitted back and used as input data to participate in the calculation of the digital twin system, so as to detect whether the data echo effect occurs. The process of determining whether control command data is being transmitted back and used as input data in a loop to detect data echo effects includes the following steps: Step a: Collect the control command dataset U(t) generated by the digital twin system at time t. U(t) includes control parameters u1, u2, and u3, which correspond to the sluice gate opening degree, flood discharge volume, and sluice gate opening and closing frequency, respectively. Step b: Record the monitoring dataset S(t+Δt) fed back by the hydraulic engineering entity system at time t+Δt. S(t+Δt) includes the water level h, flow velocity v and gate state θ corresponding to the control parameters u1, u2 and u3, respectively. Step c, calculate the similarity D1 between U(t) and S(t+Δt) using the dynamic time warping algorithm, with the following formula: n is the total number of control parameter dimensions, which is also the total number of monitoring data dimensions, with a value of 3. max(|U|,|S|) is the normalization processing coefficient to ensure that D1 is dimensionless and to eliminate the influence of parameter dimensions. i is the parameter type number. Step d: At time t+2Δt, the feedback data S(t+Δt) is input into the control command data U'(t+2Δt) generated by the digital twin system. The similarity D2 between U'(t+2Δt) and the original U(t) is calculated using the Pearson correlation coefficient R formula: ; and Let represent the mean values ​​of each control parameter in U(t) along their respective dimensions; Step e: When the product of D1 and D2 is greater than the set standard threshold, it is determined that there is a control command data feedback phenomenon, and the data echo effect is identified. If the data echo effect is detected, a comprehensive judgment is made based on the preset test parameters. The test parameters are the convergence of the prediction error of the digital twin model, which is obtained by calculating the dynamic deviation between the current prediction value and the real-time monitoring value. Based on the convergence of the prediction error, the short-term impact index and the long-term impact index are calculated respectively. The short-term impact index is used to assess the impact of the echo effect on the operation status of water conservancy projects in the current regulation cycle, and the long-term impact index is used to assess the cumulative impact of the echo effect on the accuracy of subsequent scheduling decisions. Based on the short-term and long-term impact indices, the level of the current data echo effect is determined. The echo effect level is divided into low, medium, and high levels according to the degree of impact. Different data echo optimization strategies are adopted for processing based on the determined echo effect level.

2. The intelligent monitoring method for water conservancy engineering data based on digital twins according to claim 1, characterized in that, The specific steps for calculating the short-term impact index include: Define a time window τ to evaluate the time range and divide it into multiple equally spaced time nodes to form a short-term time series. At each time node, collect three parameter values ​​predicted by the digital twin system, including water level prediction, flow velocity prediction, and gate opening and closing status prediction; at the same time, collect three monitoring parameter values ​​fed back by the physical water conservancy project system, including water level monitoring value, flow velocity monitoring value, and gate opening and closing status monitoring value. At each time point, the predicted water level value is subtracted from the monitored water level value, and the absolute value of the difference is calculated. Then, this absolute value is divided by the monitored water level value and a very small positive number is added to avoid division by zero error. The same method is used to calculate the error of the predicted flow velocity value and the predicted gate opening and closing status value, respectively. For each time point, three proportional coefficients are set for water level, flow velocity, and gate status, respectively reflecting their importance to the overall system safety or performance. The water level error, flow velocity error, and gate status error are weighted and averaged according to the proportional coefficients to obtain the comprehensive error value for that time point. Within the entire time window τ, calculate the average of the comprehensive error values ​​of all time nodes. Then, for the comprehensive error value of each time node, calculate the square of the difference between it and the average value. Sum all the squared differences and divide by the total number of time nodes to obtain the error fluctuation amplitude within the short-term time period. Finally, multiply the error fluctuation amplitude by the preset sensitivity adjustment coefficient to obtain the short-term impact index.

3. The intelligent monitoring method for water conservancy engineering data based on digital twins according to claim 2, characterized in that, The specific steps for calculating the long-term impact index include: Define a long-term observation window T, which is divided into multiple equally spaced time nodes to form a long-term time series. At each time node, collect the three parameter values ​​predicted by the digital twin system, including the predicted water level, the predicted flow velocity, and the predicted gate opening and closing status. At the same time, collect the three monitoring parameter values ​​fed back by the physical water conservancy project system, including the monitored water level, the monitored flow velocity, and the monitored gate opening and closing status. At each time point, the predicted water level value is subtracted from the monitored water level value, and the absolute value of the difference is calculated. Then, the absolute value is divided by the monitored water level value and a very small positive number is added to avoid division by zero error. The same method is used to calculate the error of the predicted flow velocity value and the predicted gate opening and closing status value respectively. For each time point, three proportional coefficients are set for water level, flow velocity, and gate status to reflect their respective impact on the long-term stability of the overall system. The water level error, flow velocity error, and gate status error are weighted and averaged according to the proportional coefficients to obtain the comprehensive error value for that time point. The comprehensive error values ​​at all time points within a long-term period are summed and multiplied by the time interval to obtain the comprehensive error integral value, which reflects the degree of error accumulation within the long-term period. The long-term impact index is obtained by dividing the integral value of the comprehensive error by the product of the total time length of the long-term period and the set maximum allowable error value. The long-term impact index is dimensionless. After normalization, the index is ensured to reflect the comparability between different water conservancy engineering systems, and finally the long-term impact index is obtained.

4. The intelligent monitoring method for water conservancy engineering data based on digital twins according to claim 3, characterized in that, When determining the level of data echo effect, fuzzy logic is used to infer the results based on the short-term impact index and the long-term impact index.

5. The intelligent monitoring method for water conservancy engineering data based on digital twins according to claim 4, characterized in that, The principle of fuzzy logic is as follows: The short-term and long-term impact indices corresponding to the occurrence of data echo effects are obtained and used as input variables for fuzzy logic. The corresponding levels of data echo effects are used as output variables for fuzzy logic. The input variables are fuzzified and their values ​​are converted into fuzzy sets. The output variables are also fuzzified and their values ​​are converted into fuzzy sets. Fuzzy rules are formulated to describe the degree of fit of different data echo effect corresponding levels under different combinations of data types. The fuzzified input variables are then used to infer the corresponding levels of data echo effects through fuzzy rules.

6. The intelligent monitoring method for water conservancy engineering data based on digital twins according to claim 5, characterized in that, Different data echo optimization strategies are employed to address the following: in the low-level echo effect state, data filtering algorithms are applied to remove abnormal feedback data; in the medium-level echo effect state, data fusion weight parameters are adjusted to reduce the model's dependence on feedback data; in the high-level echo effect state, the model's input data source is reset, an external trusted monitoring data source is forcibly introduced for verification, and the digital twin model state is reconstructed to completely eliminate the echo effect.

7. The intelligent monitoring method for water conservancy engineering data based on digital twins according to claim 6, characterized in that, In the advanced echo effect level, the state reconstruction of the digital twin model is achieved by introducing a third-party trusted data source. The trusted data source includes data from UAV inspection, satellite remote sensing, and manual inspection. The reconstruction process includes: pausing the operation of the existing model, deleting the current model feedback link information, importing trusted data to initialize the modeling parameters, generating a new digital twin model, and verifying the model's effectiveness using historical validation data.

8. The intelligent monitoring method for water conservancy engineering data based on digital twins according to claim 7, characterized in that, At the low echo level, the filtering algorithm used is adaptive Kalman filtering.

9. The intelligent monitoring method for water conservancy engineering data based on digital twins according to claim 8, characterized in that, In the intermediate echo effect state, the weights are optimized and adjusted as follows: The short-term impact index and the long-term impact index are normalized separately and converted into percentages. Calculate the average of the short-term impact index and the long-term impact index as the overall degree of echo effect; An adjustment coefficient is preset to determine the magnitude of the decrease in the fusion weight of the feedback data. The decrease value is obtained by multiplying the overall echo effect by the preset adjustment coefficient. The decrease value is then subtracted from the initial fusion weight of the feedback data, and the result is used as the new fusion weight of the feedback data. The fusion weight of the predicted data is equal to one minus the fusion weight of the new feedback data.

Images