Channel siltation probability early warning method and system based on water flow feature recognition

Abstract

The invention discloses a channel siltation probability early warning method and system for water flow feature recognition, and relates to the technical field of channel maintenance, and the method comprises the steps: obtaining channel characteristic parameters, inputting the channel characteristic parameters into an interaction knowledge graph, and determining a key water flow feature index set; performing random feature combination based on the index set, screening K coordination feature groups in combination with historical operation data, training K reference deposition prediction models, and performing integration to generate a deposition probability prediction model; collecting real-time water flow data through a sensor, processing the data and inputting the data into the model to predict the channel siltation probability; if the prediction result meets the early warning condition, siltation early warning is triggered. Therefore, the technical effects of improving the early warning perspectiveness, reducing the misjudgment rate and enhancing the complex environment adaptability are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of channel maintenance, and particularly relates to a method and system for early warning of channel siltation probability based on water flow feature recognition. Background Art

[0002] Traditional channel monitoring methods mainly focus on the monitoring of single parameters such as water level and flow velocity of basic channel parameters, and often lack a comprehensive analysis of the mutual relationship between these parameters and the deep connection with the channel siltation condition. For example, some existing channel monitoring systems only obtain single-point water level and flow velocity data by installing simple water level gauges and flow meters in the channel, and then make simple siltation judgments according to empirical thresholds. This method is difficult to accurately identify early and slight channel siltation situations, and often can only be discovered when the siltation is relatively serious and has a greater impact on the normal function of the channel.

[0003] There are many deficiencies in the existing technology for channel siltation monitoring. On the one hand, relying only on a small number of single water flow parameters for analysis cannot comprehensively reflect the complex water flow behavior in the channel, and is easily interfered by local water flow conditions or accidental factors, resulting in frequent misjudgments. On the other hand, the lack of effective utilization and in-depth mining of historical data makes the prediction ability and adaptability of the model poor, and cannot meet the urgent needs of modern water conservancy projects for early and accurate early warning of channel siltation. Summary of the Invention

[0004] The present invention provides a method and system for early warning of channel siltation probability based on water flow feature recognition to solve the technical problems in the existing technology, such as the lack of early warning ability, high misjudgment rate, and inability to adapt to complex and changeable channel operation conditions, and achieve the technical effects of improving the foresight of early warning, reducing the misjudgment rate, and enhancing the adaptability to complex environments.

[0005] In the first aspect, the present invention provides a method for early warning of channel siltation probability based on water flow feature recognition, wherein the method for early warning of channel siltation probability based on water flow feature recognition includes: Obtain the characteristic parameters of the target channel, and input the characteristic parameters into a pre-constructed interactive knowledge graph to determine a key water flow feature index set for characterizing the water flow behavior of the channel.

[0006] Taking the key water flow feature index set as the feature space, perform random feature combinations with replacement, and screen the feature combination results based on historical channel operation data to determine K coordinated feature groups.

[0007] Taking the K coordinated feature groups as input variable groups respectively, construct and train K benchmark siltation prediction models, and integrate the K benchmark siltation prediction models to obtain a siltation probability prediction model.

[0008] Based on the key water flow characteristic index set, sensors are arranged and water flow is monitored to obtain real-time water flow monitoring data. After being processed by feature engineering, the data is input into the sedimentation probability prediction model to predict the sedimentation probability of the target channel.

[0009] According to the sedimentation probability prediction result, it is judged whether there is a sedimentation risk in the target channel, and a sedimentation probability warning is issued when the risk meets the preset conditions.

[0010] In a feasible implementation manner, the characteristic parameters of the target channel are obtained and input into a pre-constructed interactive knowledge graph to determine a key water flow characteristic index set for characterizing the water flow behavior of the channel, including: The channel structure parameters and hydraulic design parameters in the target area are obtained and input into the interactive knowledge graph to trigger a knowledge graph reasoning process based on semantic reasoning and entity relationship matching.

[0011] According to the knowledge graph reasoning result, the key water flow behavior characteristic indexes corresponding to the target channel are identified to form the key water flow characteristic index set.

[0012] In a feasible implementation manner, taking the key water flow characteristic index set as the feature space, a random feature combination with replacement is performed, and the feature combination result is screened based on the historical channel operation data to determine K coordinated feature groups, including: Perform principal component analysis on the key water flow characteristic index set, and define the number of feature indexes that meet the preset contribution degree threshold as the random selection constraint.

[0013] Configure a random number generator based on the random selection constraint.

[0014] Activate the random number generator to generate a single selection quantity n, and randomly select n key water flow characteristic indexes from the key water flow characteristic index set and store them as a feature combination.

[0015] Iteratively obtain N such feature combinations, and evaluate the combination synergy degree of the N feature combinations in combination with the historical channel operation data, and select the top K feature combinations with high combination synergy degree and output them as K coordinated feature groups, where N is greater than n, and K is less than N.

[0016] In a feasible implementation manner, taking K coordinated feature groups as input variable groups respectively, K benchmark sedimentation prediction models are constructed and trained, and the K benchmark sedimentation prediction models are integrated to obtain a sedimentation probability prediction model, including: Extract the K coordinated promotion degrees of the K coordinated feature groups.

[0017] Based on the K coordination promotion degrees, combined with the historical channel operation data, K groups of lightweight prediction models are correspondingly constructed and trained.

[0018] Select the lightweight prediction model with the best performance in each group, obtain the K benchmark sedimentation prediction models for ensemble learning, and construct the sedimentation probability prediction model.

[0019] In a feasible implementation manner, based on the key water flow feature index set, sensor layout and water flow monitoring are carried out to obtain real-time water flow monitoring data, including: Based on the key water flow feature index set, determine the monitoring requirements of each feature index, configure corresponding types of sensors, and establish.

[0020] Obtain the flow velocity data of each profile direction of the target channel through a flow velocity sensor, obtain the water level change data of each typical cross-section through a water level sensor, collect pressure pulsation signals through a turbulence sensor to construct a turbulent dynamic feature sequence, and obtain the suspended sediment concentration and bed sediment concentration in the water body through a sediment sensor, and output it as the real-time water flow monitoring data.

[0021] In a feasible implementation manner, after being processed by feature engineering, it is input into the sedimentation probability prediction model for predicting the sedimentation probability of the target channel, including: Decompose the key water flow feature index set and classify it into intuitive water flow feature indexes and deduced water flow feature indexes.

[0022] Based on the preprocessed real-time water flow monitoring data, extract the intuitive water flow feature parameters corresponding to the intuitive water flow feature indexes.

[0023] Calculate the deduced water flow feature parameters corresponding to the deduced water flow feature indexes according to the intuitive water flow feature parameters.

[0024] Fuse the intuitive water flow feature parameters and the deduced water flow feature parameters, and process the fusion result through feature engineering methods to obtain real-time input variables.

[0025] Transmit the real-time input variables to the input end of the sedimentation probability prediction model, and obtain the sedimentation probability prediction result from the output end.

[0026] In a feasible implementation manner, according to the sedimentation probability prediction result, judge whether there is a sedimentation risk in the target channel, and issue a sedimentation probability warning when the risk meets the preset conditions, including: Set a sedimentation probability risk grading threshold adapted to the channel operation safety level, and establish a sedimentation risk discrimination criterion.

[0027] Compare the sedimentation probability prediction result with the sedimentation risk discrimination criterion.

[0028] When the sedimentation probability result exceeds the corresponding threshold of the discrimination criterion, it is determined that there is a sedimentation risk.

[0029] Generate a warning instruction according to the sedimentation risk level, and send the sedimentation probability warning information to the channel operation management terminal through the warning interface.

[0030] In a second aspect, the present invention also provides a channel sedimentation probability warning system for identifying water flow characteristics. Among them, the channel sedimentation probability warning system for identifying water flow characteristics includes: A characteristic index determination module, configured to obtain the characteristic parameters of the target channel, and input the characteristic parameters into a pre-constructed interactive knowledge graph to determine a key water flow characteristic index set for characterizing the water flow behavior of the channel.

[0031] A coordinated feature group screening module, configured to use the key water flow characteristic index set as a feature space to perform random feature combinations with replacement, and screen the feature combination results based on historical channel operation data to determine K coordinated feature groups.

[0032] A prediction model construction and integration module, configured to respectively use K of the coordinated feature groups as input variable groups, construct and train K baseline sedimentation prediction models, and integrate the K baseline sedimentation prediction models to obtain a sedimentation probability prediction model.

[0033] A water flow monitoring and sedimentation probability prediction module, configured to arrange sensors and monitor water flow based on the key water flow characteristic index set, obtain real-time water flow monitoring data, and input it into the sedimentation probability prediction model after feature engineering processing to predict the sedimentation probability of the target channel.

[0034] A sedimentation risk judgment and warning module, configured to judge whether there is a sedimentation risk in the target channel according to the sedimentation probability prediction result, and issue a sedimentation probability warning when the risk meets the preset conditions.

[0035] The present invention discloses a method and system for early warning of channel siltation probability based on water flow feature recognition, including: obtaining characteristic parameters of a target channel, inputting the characteristic parameters into a pre-constructed interactive knowledge graph, and determining a key water flow feature index set for characterizing the water flow behavior of the channel; using the key water flow feature index set as a feature space, performing replaceable random feature combinations, and screening the feature combination results based on historical channel operation data to determine K coordinated feature groups; respectively using the K coordinated feature groups as input variable groups, constructing and training K benchmark siltation prediction models, and integrating the K benchmark siltation prediction models to obtain a siltation probability prediction model; arranging sensors and monitoring water flow based on the key water flow feature index set, obtaining real-time water flow monitoring data, and inputting the data into the siltation probability prediction model after feature engineering processing to predict the siltation probability of the target channel; judging whether there is a siltation risk in the target channel according to the siltation probability prediction result, and issuing an early warning of siltation probability when the risk meets a preset condition. The method and system for early warning of channel siltation probability based on water flow feature recognition disclosed by the present invention solve the technical problems of lack of early warning ability, high misjudgment rate, and inability to adapt to complex and changeable channel operation conditions, and achieve the technical effects of improving the foresight of early warning, reducing the misjudgment rate, and enhancing the adaptability to complex environments. Description of the Drawings

[0036] Figure 1 It is a schematic flow chart of a method for early warning of channel siltation probability based on water flow feature recognition according to the present invention.

[0037] Figure 2 It is a schematic structural diagram of a system for early warning of channel siltation probability based on water flow feature recognition according to the present invention.

[0038] Description of the reference numerals: Feature index determination module 11, Coordinated feature group screening module 12, Prediction model construction and integration module 13, Water flow monitoring and siltation probability prediction module 14, Siltation risk judgment and early warning module 15. Detailed Embodiments

[0039] The following will combine the description of the drawings in the specification and specific embodiments to elaborate on the above technical solutions in detail to better understand the above technical solutions. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments only for explaining the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention. In addition, it should be noted that for the sake of description, only parts related to the present invention are shown in the drawings, rather than all of them.

[0040] Embodiment 1, as Figure 1The figure is a schematic flowchart of a method for early warning of channel siltation probability based on water flow feature recognition. Among them, the method for early warning of channel siltation probability based on water flow feature recognition includes: Obtain the characteristic parameters of the target channel, and input the characteristic parameters into a pre-constructed interactive knowledge graph to determine a set of key water flow feature indicators for characterizing the water flow behavior of the channel.

[0041] Specifically, the "channel characteristic parameters" refer to a set of parameters that can reflect the physical structure and operation status of the target channel, including but not limited to channel cross-section shape, bottom slope of the channel, channel width, channel roughness, water depth, water flow velocity, flow rate, water level, sediment concentration, etc.

[0042] Specifically, the "interactive knowledge graph" refers to a knowledge network structure with multiple entities and multiple relationships constructed based on channel operation data, monitoring data, and historical siltation events. Among them, nodes represent entities such as channels, monitoring points, sensors, and historical events, and edges represent physical, logical, or causal associations between entities, and various characteristic parameters can be stored and retrieved. By taking various characteristic parameters of the channel and water body as entity nodes and defining the relationships between these entities (such as semantic connections like "influence" and "association"), a semantic network capable of logical reasoning is formed.

[0043] Specifically, the "set of key water flow feature indicators" refers to a set of feature indicators that can sensitively reflect changes in the water flow behavior of the channel and have a high correlation with siltation risk, such as flow velocity distribution, water level change rate, pulsation intensity of turbulence, sediment concentration, etc.

[0044] Through the above-mentioned interactive knowledge graph, the efficient screening from a large amount of channel characteristic parameters to accurate key water flow features is realized, avoiding the one-sidedness of manual experience judgment, providing a data basis with high correlation and high value for subsequent model construction based on these features, thereby ensuring the accuracy and reliability of the entire siltation probability early warning method.

[0045] In some embodiments, obtaining the characteristic parameters of the target channel and inputting the characteristic parameters into a pre-constructed interactive knowledge graph to determine a set of key water flow feature indicators for characterizing the water flow behavior of the channel includes: Obtain the channel structure parameters and hydraulic design parameters in the target area and input them into the interactive knowledge graph to trigger a knowledge graph reasoning process based on semantic reasoning and entity relationship matching; according to the knowledge graph reasoning result, identify the key water flow behavior characteristic indicators corresponding to the target channel to form the set of key water flow feature indicators.

[0046] Specifically, the channel structure parameters refer to the parameters related to the geometric shape of the channel, such as the width, depth, and side slope coefficient of the channel. The hydraulic design parameters cover the design flow rate, design velocity, roughness coefficient, etc. of the channel. The channel structure parameters and hydraulic design parameters are used to describe the static characteristics of the channel and the expected water flow dynamic characteristics from different dimensions.

[0047] Specifically, the interactive knowledge graph contains a vast amount of water conservancy knowledge and logical rules, and has sorted out and stored various water conservancy concepts and their associated relationships during pre-construction, such as the relationship between the bottom slope of the channel and the water flow velocity, the correlation between the flow rate and the water passing capacity of the channel, etc. Through the interactive knowledge graph, the characteristic indicators suitable for channels of different types and parameters can be efficiently determined.

[0048] Specifically, first obtain the structure parameters of the channel in the target area, such as the channel width is 5m, the depth is 3m, the side slope coefficient is 1.2, etc. At the same time, obtain the hydraulic design parameters, such as the design flow rate is 10m³ / s, the design velocity is 1.5m / s, the roughness coefficient is 0.02, etc., and input the above parameters as attribute information into the pre-constructed interactive knowledge graph. After receiving this information, the knowledge graph automatically retrieves the historical data, expert rules, and typical water flow behavior patterns associated with the above parameters through semantic reasoning and entity relationship matching mechanisms, triggering the reasoning process of the knowledge graph. For example, based on the relationship between "small channel bottom slope + large roughness coefficient" and "high incidence of historical sedimentation", the system conducts semantic reasoning and judges based on the stored water conservancy knowledge rules that under this channel width, depth, and design velocity, which positions are prone to abnormal reduction in velocity leading to sediment deposition, and further determines key water flow behavior characteristic indicators such as the velocity distribution characteristics at the bend and the water level fluctuation frequency at a specific cross-section, forming a set of key water flow characteristic indicators, providing a characteristic basis for subsequent sedimentation probability early warning.

[0049] Through the above process, it is possible to achieve structured management and semantic association of the channel structure parameters and hydraulic design parameters, and use the reasoning ability of the knowledge graph to automatically screen and output a set of key water flow characteristic indicators highly relevant to the channel sedimentation risk. This helps to improve the intelligence and scientific nature of channel water flow behavior characteristic recognition, provides efficient and accurate characteristic input for the channel sedimentation probability early warning model, and enhances the accuracy and reliability of sedimentation risk early warning.

[0050] Using the set of key water flow characteristic indicators as the feature space, perform random feature combinations with replacement, and screen the feature combination results based on historical channel operation data to determine K coordinated feature groups.

[0051] Specifically, the key water flow feature index set is a group of parameters that are screened out by the interactive knowledge graph and have a significant representative effect on the channel water flow behavior. These parameters can effectively reflect the channel water flow state and the potential degree of siltation risk, such as the velocity distribution characteristic value, the water level change rate, the mean value of turbulent pulsation intensity, etc.

[0052] Specifically, the feature space is a multi-dimensional parameter space constructed with the key water flow feature indexes as dimensions, that is, the selection space for the random feature combinations with replacement; when constructing the feature combinations, the same feature index is allowed to be selected multiple times in different combinations, so as to generate diverse feature combinations from the limited feature indexes, thereby increasing the probability of screening out high-quality feature combinations.

[0053] Specifically, the coordinated feature groups refer to those feature combinations that perform excellently in the evaluation of the combination synergy degree. Their combinations can show good representative ability and prediction potential for the channel siltation condition under the verification of the historical channel operation data.

[0054] In some embodiments, using the key water flow feature index set as the feature space, perform random feature combinations with replacement, and screen the feature combination results based on the historical channel operation data to determine K coordinated feature groups, including: Perform principal component analysis on the key water flow feature index set, define the number of feature indexes that meet the preset contribution threshold as the random selection constraint; configure a random number generator based on the random selection constraint; activate the random number generator to generate a single selection quantity n, and randomly select n key water flow feature indexes from the key water flow feature index set and store them as a feature combination; iteratively obtain N such feature combinations, and evaluate the combination synergy degree of the N feature combinations in combination with the historical channel operation data, and select the top K feature combinations with high combination synergy degree and output them as K coordinated feature groups, where N is greater than n, and K is less than N.

[0055] Specifically, the feature combination synergy degree refers to the collaborative improvement effect of each feature combination in the channel siltation probability prediction task, usually measured by the proportion of the improvement in the prediction performance after combination; the K coordinated feature groups refer to the top K subsets of feature indexes that are screened out from all random feature combinations and show high synergy degree on the historical channel operation data.

[0056] Specifically, in the specific implementation process, first, principal component analysis is performed on the key water flow characteristic index set (such as including 8 items: flow variation coefficient, water flow velocity distribution, water level gradient, sediment concentration, etc.), and the contribution degree of each characteristic is calculated. If the cumulative contribution degree reaches 90% and the required number of characteristics is 5, then "n ≤ 5" is used as the random selection constraint. Then, based on the above constraints, a random number generator is configured, and each time it is activated to generate a selection quantity n (such as n = 3, 4, or 5), and n indicators are randomly selected from the key water flow characteristic index set to form a characteristic combination. The selected indicators can continue to be used for subsequent selections, that is, the same characteristic is allowed to participate in different combinations multiple times; repeat this process to iteratively obtain N characteristic combinations (such as N = 50, n = 3 - 5).

[0057] Furthermore, using the historical channel operation data, the performance of each characteristic combination in the sedimentation probability prediction model is evaluated, and the combination synergy is calculated. Among them, the combination synergy can be measured by metrics such as area under the curve improvement, accuracy improvement, correlation coefficient, etc. Finally, according to the above combination synergy evaluation results and meeting the screening constraints (such as selecting the top K characteristic combinations with the highest synergy), K coordinated characteristic groups are output. These coordinated characteristic groups can reflect the internal relationship between the channel water flow characteristics and the sedimentation status to the greatest extent.

[0058] Through the above process, using the random characteristic combination method with replacement, it is possible to efficiently explore various potential characteristic combination patterns in the key water flow characteristic index set, and through the screening and verification of historical data, accurately determine the characteristic groups that have high coordination and representativeness for channel sedimentation prediction. It not only enriches the diversity of characteristic combinations, improves the comprehensiveness of feature screening, but also provides high-quality and multi-dimensional feature inputs for the subsequent construction of the benchmark sedimentation prediction model, helps to improve the prediction accuracy and generalization ability of the model, and thus realizes early and accurate warning of sedimentation risks under complex and changeable channel operation conditions.

[0059] Using the K coordinated characteristic groups as input variable groups respectively, K benchmark sedimentation prediction models are constructed and trained, and the K benchmark sedimentation prediction models are integrated to obtain a sedimentation probability prediction model.

[0060] Specifically, the coordinated characteristic group refers to the characteristic group obtained through random characteristic combination and historical data screening. The characteristics within each group have high synergy and complementarity in characterizing the channel sedimentation status, and can comprehensively reflect the relationship between the channel water flow behavior and the sedimentation risk from different dimensions. In other words, the characteristic combination relationship of the coordinated characteristic group helps to improve the prediction accuracy.

[0061] Specifically, the benchmark sedimentation prediction model is a prediction model constructed based on a single coordinated feature group, which uses lightweight machine learning algorithms (such as logistic regression, decision tree, K-nearest neighbor, shallow neural network, etc.) to ensure the fast training and efficient prediction ability of the model. By comprehensively integrating the prediction results of multiple benchmark models (such as weighted average, voting mechanism, etc.), a comprehensive sedimentation probability prediction model can be obtained. This sedimentation probability prediction model improves the accuracy and stability of the prediction through the fusion of diverse models.

[0062] In some embodiments, using the K coordinated feature groups as input variable groups respectively, K benchmark sedimentation prediction models are constructed and trained, and the K benchmark sedimentation prediction models are integrated to obtain a sedimentation probability prediction model, including: Extracting the K coordinated promotion degrees of the K coordinated feature groups; based on the K coordinated promotion degrees, combined with the historical channel operation data, correspondingly constructing and training K groups of lightweight prediction models; selecting the lightweight prediction model with the best performance in each group, obtaining the K benchmark sedimentation prediction models for ensemble learning, and constructing the sedimentation probability prediction model.

[0063] Specifically, first, calculate the coordinated promotion degree of each coordinated feature group. Assume K = 3, and the three coordinated feature groups are Group 1 (flow velocity, water level), Group 2 (sediment concentration, turbulence intensity), and Group 3 (flow velocity, sediment concentration). By training a preliminary model with historical data and calculating the explained variance contribution rate or other performance indicators of each group for sedimentation prediction, the coordinated promotion degree of Group 1 is obtained as 0.4, Group 2 is 0.3, and Group 3 is 0.3.

[0064] Next, based on the obtained coordinated promotion degrees, combined with the historical channel operation data, respectively construct and train multiple lightweight prediction models for each coordinated feature group. Among them, the number of lightweight prediction models corresponding to each coordinated feature group is determined by the coordinated promotion degree. The higher the promotion degree, the fewer the configurable number, which helps to reduce the model cost.

[0065] Then, perform performance evaluation (such as accuracy, recall, etc.) on the lightweight models within each group, and select the model with the best performance in each group as a benchmark sedimentation prediction model. Suppose the decision tree model has the best performance in Group 1, the support vector machine model is the best in Group 2, and the logistic regression model is the best in Group 3. At this time, 3 benchmark sedimentation prediction models are obtained. Finally, these 3 models are subjected to ensemble learning (such as, using the weighted average method) to construct a sedimentation probability prediction model.

[0066] Optionally, when using the weighted average method for ensemble learning, the weights can be normalized and assigned according to the coordinated promotion degrees.

[0067] Through the above process, the complementary information of different feature groups can be fully utilized, thereby enhancing the sensitivity and recognition accuracy of the model to the risk of channel siltation, reducing the risk of overfitting or feature bias of a single model, and providing more reliable decision-making support for channel operation management.

[0068] Based on the set of key water flow feature indicators, sensor layout and water flow monitoring are carried out to obtain real-time water flow monitoring data, and after being processed by feature engineering, the data is input into the siltation probability prediction model to predict the siltation probability of the target channel.

[0069] Specifically, according to the monitoring requirements of the set of key water flow feature indicators, appropriate sensor types (such as flow velocity sensors, water level gauges, etc.) can be selected and their reasonable installation positions on the channel can be determined.

[0070] Among them, the sensors may include: a flow velocity profile monitoring unit, such as an ADCP, an electromagnetic flowmeter array, to collect the vertical and lateral flow velocity distributions; a water level monitoring unit, such as a high-precision pressure sensor and a radar water level gauge, to obtain real-time water level changes; a turbulence monitoring unit, such as a high-frequency pressure pulsation sensor, to analyze the turbulent kinetic energy and vortex frequency spectrum characteristics; and a sediment monitoring unit, such as a turbidimeter and a laser particle size analyzer.

[0071] Specifically, the real-time water flow monitoring data refers to various data reflecting the water flow state collected by sensors in real time during the operation of the channel. Feature engineering processing is to perform operations such as preprocessing, feature extraction, and transformation on the collected original monitoring data to form feature inputs that can be efficiently utilized by the prediction model.

[0072] In some embodiments, based on the set of key water flow feature indicators, sensor layout and water flow monitoring are carried out to obtain real-time water flow monitoring data, including: Based on the set of key water flow feature indicators, determine the monitoring requirements for each feature indicator, configure corresponding types of sensors, and establish; obtain the flow velocity data of each profile direction of the target channel through a flow velocity sensor, obtain the water level change data of each typical cross-section through a water level sensor, collect pressure pulsation signals through a turbulence sensor to construct a turbulent dynamic feature sequence, and obtain the suspended sediment concentration and bed sediment concentration in the water body through a sediment sensor, and output as the real-time water flow monitoring data.

[0073] Specifically, the monitoring requirements refer to determining the required monitoring accuracy, frequency, position, etc. according to the characteristics of the set of key water flow feature indicators to ensure that the collected data can accurately reflect the water flow behavior.

[0074] Specifically, the sensor types include flow velocity sensors, water level sensors, turbulence sensors, sediment sensors, etc. They are respectively used to measure the water flow velocity, water level changes, turbulent pulsation signals, and the concentration distribution of sediment in the water body. The real-time water flow monitoring data refers to the original data collected and transmitted in real time by the above sensors during the operation of the channel, which is an important basis for subsequent sedimentation probability prediction.

[0075] Specifically, based on the key water flow characteristic index set (such as flow velocity, water level, turbulence, suspended sediment concentration, bed sediment concentration), analyze the spatial distribution and dynamic change requirements of each characteristic, and determine the monitoring cross-section and sampling frequency. For example: The flow velocity needs to be measured at multiple points in the upper, middle, and lower reaches and key cross-sections, with a sampling frequency of 1 Hz; The turbulence characteristics need to be measured by high-frequency turbulence sensors near the turning section or structures where turbulence is likely to occur, with a sampling frequency above 10 Hz; The suspended sediment and bed sediment concentrations need to be measured by sediment sensors at typical cross-sections, with a sampling frequency of 0.1 Hz.

[0076] Specifically, obtain the flow velocity data in each profile direction of the target channel through a flow velocity sensor (such as an ultrasonic flowmeter) to form a profile flow velocity distribution matrix; obtain the water level change data of each typical cross-section through a water level sensor (such as a pressure water level gauge, radar water level gauge) to form a water level time series curve; collect pressure pulsation signals through a turbulence sensor (such as a pressure pulsation sensor, particle image velocimetry), and construct a turbulent dynamic characteristic sequence (such as turbulence intensity, energy spectrum, etc.) through signal processing (such as wavelet analysis, Fourier transform); obtain the suspended sediment concentration and bed sediment concentration in the water body through a sediment sensor (such as an optical turbidity meter or ultrasonic sediment meter), and output the sediment spatio-temporal distribution data.

[0077] Furthermore, integrate the multi-source data collected by the above various sensors and output it as the real-time water flow monitoring data, providing high-quality input for subsequent feature engineering and sedimentation probability prediction models.

[0078] Through scientific sensor layout and multi-dimensional water flow monitoring based on the key water flow characteristic index set, it is possible to comprehensively, dynamically, and finely obtain the water flow and sediment transport states in the channel, providing a rich and accurate data basis for subsequent feature engineering processing and sedimentation probability prediction, thereby improving the timeliness and reliability of channel sedimentation risk monitoring and intelligent early warning.

[0079] In some embodiments, after being processed by feature engineering and input into the sedimentation probability prediction model for predicting the sedimentation probability of the target channel, it includes: Decompose the key water flow characteristic index set, and classify it into intuitive water flow characteristic indexes and deduced water flow characteristic indexes; based on the preprocessed real-time water flow monitoring data, extract the intuitive water flow characteristic parameters corresponding to the intuitive water flow characteristic indexes; calculate the deduced water flow characteristic parameters corresponding to the deduced water flow characteristic indexes according to the intuitive water flow characteristic parameters; fuse the intuitive water flow characteristic parameters and the deduced water flow characteristic parameters, and process the fusion result by feature engineering methods to obtain real-time input variables; transmit the real-time input variables to the input end of the sedimentation probability prediction model, and obtain the sedimentation probability prediction result from the output end.

[0080] Specifically, intuitive water flow characteristic indexes refer to water flow characteristic parameters that can be directly measured by sensors, such as flow velocity, water level, turbidity, pressure pulsation, etc. These indexes can directly reflect the basic state of water flow. Deduced water flow characteristic indexes are indirect characteristics derived from intuitive water flow characteristic parameters through mathematical models, physical formulas or empirical relationships, such as kinetic energy of water flow, shear force, sediment transport rate, Reynolds number, Froude number, etc., which are used to provide more in-depth information on water flow behavior. Exemplarily, for example, the flow rate can be calculated based on the flow velocity and cross-sectional area; the shear force can be calculated based on the flow velocity gradient and fluid viscosity; dimensionless parameters such as Reynolds number and Froude number can be calculated based on the flow velocity and water depth; the sediment transport rate can be deduced based on the flow velocity and sediment concentration, etc.

[0081] In the specific implementation process, first, decompose the key water flow characteristic index set and classify it into intuitive water flow characteristic indexes and deduced water flow characteristic indexes. For example, parameters that can be directly measured, such as flow velocity and water level, are classified as intuitive water flow characteristic indexes; turbulent intensity calculated based on flow velocity and flow rate calculated based on water level and flow velocity are classified as deduced water flow characteristic indexes. Then, based on the preprocessed real-time water flow monitoring data, extract the intuitive water flow characteristic parameters corresponding to the intuitive water flow characteristic indexes. For example, extract the flow velocity values in each profile direction from the data collected by the flow velocity sensor, and extract the water level values of each typical cross-section from the data collected by the water level sensor.

[0082] Furthermore, calculate the deduced water flow characteristic parameters corresponding to the deduced water flow characteristic indexes according to the extracted intuitive water flow characteristic parameters. Fuse the intuitive water flow characteristic parameters and the deduced water flow characteristic parameters, and process the fusion result by feature engineering methods to obtain real-time input variables. For example, standardize parameters such as flow velocity, water level, turbulent intensity, and flow rate to eliminate the influence of dimension, and then perform feature construction (such as feature splicing) to form a feature vector or feature matrix that can comprehensively characterize the water flow state of the channel.

[0083] Finally, the processed real-time input variables are transmitted to the input end of the sedimentation probability prediction model. The model calculates based on the input feature vectors and obtains the sedimentation probability prediction result of the target channel from the output end, such as the output sedimentation probability P (e.g., P = 0.82, indicating that the current sedimentation probability of the channel is 82%).

[0084] Through the scientific decomposition, parameter extraction and fusion of the key water flow feature index set, combined with the feature engineering method, the above process can significantly improve the representativeness and effectiveness of the model input variables, enhance the model's ability to depict complex hydrodynamic and sediment transport processes, and thus greatly improve the accuracy and timeliness of sedimentation probability prediction.

[0085] According to the sedimentation probability prediction result, judge whether there is a sedimentation risk in the target channel, and issue a sedimentation probability warning when the risk meets the preset conditions.

[0086] In some embodiments, according to the sedimentation probability prediction result, judging whether there is a sedimentation risk in the target channel, and issuing a sedimentation probability warning when the risk meets the preset conditions, includes: Setting sedimentation probability risk grading thresholds adapted to the channel operation safety level, and establishing sedimentation risk discrimination criteria; comparing the sedimentation probability prediction result with the sedimentation risk discrimination criteria; when the sedimentation probability result exceeds the corresponding threshold of the discrimination criteria, it is determined that there is a sedimentation risk; generating a warning instruction according to the sedimentation risk level, and sending a sedimentation probability warning message to the channel operation management end through the warning interface.

[0087] Specifically, the sedimentation probability risk grading threshold refers to the probability threshold preset according to the channel operation safety level for distinguishing different sedimentation risk levels, such as low risk (0% - 30%), medium risk (30% - 60%), high risk (60% - 90%) and extremely high risk (above 90%). The sedimentation risk discrimination criterion refers to the judgment criterion based on the sedimentation probability risk grading threshold for determining whether there is a sedimentation risk in the target channel and its level currently.

[0088] Specifically, the warning interface is a communication interface for sending sedimentation risk warning information to the channel operation management end, and can be implemented in various forms such as wired, wireless or platform API. The warning instruction is automatically generated by the system when it is determined that there is a sedimentation risk, and the risk prompt information is pushed through the warning interface.

[0089] Specifically, the predicted sedimentation probability results output by the model are compared with the pre-set discrimination criteria. For example, if the predicted sedimentation probability of a certain cross-section by the model is 65%, and the high-risk threshold of this channel is 50%, it is determined that there is a high-risk sedimentation at this cross-section. Furthermore, an early warning instruction is generated according to the determined risk level, and the sedimentation probability early warning information is sent to the channel operation management terminal through the early warning interface to remind the management personnel to take corresponding measures. Exemplarily, the early warning information may include fields such as channel number, time, current sedimentation probability, risk level, recommended measures, etc.

[0090] Through the above process, the risk classification thresholds adapted to the channel safety level are set, and a scientific risk discrimination criterion is established, realizing the automatic, hierarchical determination and real-time early warning of the channel sedimentation risk. It can effectively improve the risk perception ability and response efficiency of channel operation and maintenance, significantly reduce the operation safety hazards caused by sedimentation in the channel, and ensure the safe and stable operation of the channel system.

[0091] In summary, a method for early warning of channel sedimentation probability by identifying water flow characteristics provided by the present invention has the following technical effects: By obtaining the characteristic parameters of the target channel and inputting the characteristic parameters into a pre-constructed interactive knowledge graph, a key water flow characteristic index set for characterizing the water flow behavior of the channel is determined; using the key water flow characteristic index set as the feature space, random feature combinations with replacement are performed, and the feature combination results are screened based on historical channel operation data to determine K coordinated feature groups; respectively using the K coordinated feature groups as input variable groups, K benchmark sedimentation prediction models are constructed and trained, and the K benchmark sedimentation prediction models are integrated to obtain a sedimentation probability prediction model; based on the key water flow characteristic index set, sensor layout and water flow monitoring are carried out to obtain real-time water flow monitoring data, and after being processed by feature engineering, they are input into the sedimentation probability prediction model to predict the sedimentation probability of the target channel; according to the sedimentation probability prediction result, it is judged whether there is a sedimentation risk in the target channel, and a sedimentation probability early warning is issued when the risk meets the preset conditions, so as to achieve the technical effects of improving the forward-looking of the early warning, reducing the misjudgment rate, and enhancing the adaptability to complex environments.

[0092] Embodiment 2, as Figure 2 is a schematic structural diagram of a system for early warning of channel sedimentation probability by identifying water flow characteristics of the present invention. For example, Figure 1 In the flow chart of a method for early warning of channel sedimentation probability by identifying water flow characteristics of the present invention can be realized by a structure such as Figure 2 shown.

[0093] Based on the same concept as the method for early warning of channel sedimentation probability by identifying water flow characteristics in the above embodiment, the present invention also provides a system for early warning of channel sedimentation probability by identifying water flow characteristics, including: The characteristic index determination module 11 is configured to obtain the characteristic parameters of the target channel, and input the characteristic parameters into a pre-constructed interactive knowledge graph to determine a key water flow characteristic index set for characterizing the channel water flow behavior.

[0094] The coordinated feature group screening module 12 is configured to perform replacement random feature combinations with the key water flow characteristic index set as the feature space, and screen the feature combination results based on historical channel operation data to determine K coordinated feature groups.

[0095] The prediction model construction and integration module 13 is configured to respectively use the K coordinated feature groups as input variable groups, construct and train K benchmark sedimentation prediction models, and integrate the K benchmark sedimentation prediction models to obtain a sedimentation probability prediction model.

[0096] The water flow monitoring and sedimentation probability prediction module 14 is configured to perform sensor layout and water flow monitoring based on the key water flow characteristic index set, obtain real-time water flow monitoring data, and input it into the sedimentation probability prediction model after feature engineering processing to predict the sedimentation probability of the target channel.

[0097] The sedimentation risk judgment and early warning module 15 is configured to judge whether there is a sedimentation risk in the target channel according to the sedimentation probability prediction result, and issue a sedimentation probability early warning when the risk meets the preset conditions.

[0098] In some embodiments, the characteristic index determination module 11 includes: The channel parameter acquisition and knowledge graph input unit is configured to acquire the channel structure parameters and hydraulic design parameters in the target area, and input them into the interactive knowledge graph to trigger a knowledge graph reasoning process based on semantic reasoning and entity relationship matching.

[0099] The key water flow characteristic index recognition unit is configured to identify the key water flow behavior characteristic indexes corresponding to the target channel according to the knowledge graph reasoning result, and form the key water flow characteristic index set.

[0100] In some embodiments, the coordinated feature group screening module 12 includes: The principal component analysis and random selection constraint definition unit is configured to perform principal component analysis on the key water flow characteristic index set, and define the number of characteristic indexes that meet the preset contribution degree threshold as the random selection constraint.

[0101] The random number generator configuration unit is configured to configure a random number generator based on the random selection constraint.

[0102] The feature combination generation and storage unit is configured to activate the random number generator to generate a single selection quantity n, and randomly select n key water flow characteristic indexes from the key water flow characteristic index set and store them as a feature combination.

[0103] A feature combination iteration and synergy evaluation unit for iteratively obtaining N of the feature combinations, evaluating the combination synergy of the N feature combinations in combination with the historical channel operation data, and selecting the top K of the feature combinations with high combination synergy, and outputting them as K coordinated feature groups, where N is greater than n and K is less than N.

[0104] In some embodiments, the prediction model construction and integration module 13 includes: A coordination promotion degree extraction unit for extracting the K coordination promotion degrees of the K coordinated feature groups.

[0105] A lightweight prediction model construction and training unit for correspondingly constructing and training K groups of lightweight prediction models based on the K coordination promotion degrees in combination with the historical channel operation data.

[0106] A benchmark sedimentation prediction model integration and prediction model construction unit for selecting the lightweight prediction model with the optimal performance in each group, obtaining K of the benchmark sedimentation prediction models for ensemble learning, and constructing the sedimentation probability prediction model.

[0107] In some embodiments, the water flow monitoring and sedimentation probability prediction module 14 includes: A monitoring requirement determination and sensor configuration unit for determining the monitoring requirements of each feature index based on the key water flow feature index set, configuring corresponding types of sensors, and establishing a monitoring system.

[0108] A real-time water flow monitoring data acquisition unit for obtaining the flow velocity data of each profile direction of the target channel through a flow velocity sensor, obtaining the water level change data of each typical cross-section through a water level sensor, collecting pressure pulsation signals through a turbulence sensor to construct a turbulent dynamic feature sequence, and obtaining the suspended sediment concentration and bed sediment concentration in the water body through a sediment sensor, and outputting the real-time water flow monitoring data.

[0109] In some embodiments, the water flow monitoring and sedimentation probability prediction module 14 further includes: A key water flow feature index decomposition and classification unit for decomposing the key water flow feature index set and classifying it into intuitive water flow feature indexes and deduced water flow feature indexes.

[0110] An intuitive water flow feature parameter extraction unit for extracting the intuitive water flow feature parameters corresponding to the intuitive water flow feature indexes based on the preprocessed real-time water flow monitoring data.

[0111] A deduced water flow feature parameter calculation unit for calculating the deduced water flow feature parameters corresponding to the deduced water flow feature indexes according to the intuitive water flow feature parameters.

[0112] A real-time input variable acquisition unit, configured to fuse the intuitive water flow characteristic parameters and the estimated water flow characteristic parameters, and process the fusion result by means of feature engineering methods to obtain real-time input variables.

[0113] A sedimentation probability prediction result acquisition unit, configured to transmit the real-time input variables to the input end of the sedimentation probability prediction model, and obtain the sedimentation probability prediction result from the output end.

[0114] In some embodiments, the sedimentation risk judgment and early warning module 15 includes: A sedimentation probability risk grading threshold setting and discrimination criterion establishing unit, configured to set a sedimentation probability risk grading threshold adapted to the channel operation safety level, and establish a sedimentation risk discrimination criterion.

[0115] A sedimentation probability prediction result comparison unit, configured to compare the sedimentation probability prediction result with the sedimentation risk discrimination criterion.

[0116] A sedimentation risk determination unit, configured to determine that there is a sedimentation risk when the sedimentation probability result exceeds the corresponding threshold of the discrimination criterion.

[0117] An early warning instruction generation and early warning information sending unit, configured to generate an early warning instruction according to the sedimentation risk level, and send a sedimentation probability early warning information to the channel operation management end through an early warning interface.

[0118] It should be understood that the embodiments mentioned in this specification focus on their differences from other embodiments. The specific embodiments in the foregoing Embodiment 1 are equally applicable to the channel sedimentation probability early warning system for water flow feature recognition described in Embodiment 2. For the sake of brevity of the specification, no further elaboration will be made here.

[0119] It should be understood that the disclosed embodiments of the present invention and the above descriptions can enable those skilled in the art to implement the present invention using the present invention. At the same time, the present invention is not limited to the above-mentioned part of the embodiments. It should be understood that those of ordinary skill in the art can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for early warning of channel siltation probability by identifying water flow characteristics, characterized in that Including: Obtain the characteristic parameters of the target channel, input the characteristic parameters into a pre-constructed interactive knowledge graph, and determine a key water flow feature index set for characterizing the water flow behavior of the channel; Taking the key water flow feature index set as the feature space, perform random feature combinations with replacement, and screen the feature combination results based on historical channel operation data to determine K coordinated feature groups; Respectively taking the K coordinated feature groups as input variable groups, construct and train K benchmark siltation prediction models, and integrate the K benchmark siltation prediction models to obtain a siltation probability prediction model; Based on the key water flow feature index set, arrange sensors and monitor water flow, obtain real-time water flow monitoring data, and input it into the siltation probability prediction model after feature engineering processing to predict the siltation probability of the target channel; According to the siltation probability prediction result, judge whether there is a siltation risk in the target channel, and issue a siltation probability warning when the risk meets the preset conditions.

2. The method for warning of channel sedimentation probability for identifying water flow characteristics according to claim 1, wherein Obtain the characteristic parameters of the target channel, input the characteristic parameters into a pre-constructed interactive knowledge graph, and determine a key water flow feature index set for characterizing the water flow behavior of the channel, including: Obtain the channel structure parameters and hydraulic design parameters in the target area, and input them into the interactive knowledge graph to trigger a knowledge graph reasoning process based on semantic reasoning and entity relationship matching; According to the knowledge graph reasoning result, identify the key water flow behavior feature indexes corresponding to the target channel, and form the key water flow feature index set.

3. The channel siltation probability early warning method for water flow feature recognition according to claim 2, wherein Taking the key water flow feature index set as the feature space, perform random feature combinations with replacement, and screen the feature combination results based on historical channel operation data to determine K coordinated feature groups, including: Perform principal component analysis on the key water flow feature index set, and define the number of feature indexes that meet the preset contribution threshold as the random selection constraint; Configure a random number generator based on the random selection constraint; Activate the random number generator to generate a single selection quantity n, and randomly select n key water flow feature indexes from the key water flow feature index set and store them as a feature combination; Iteratively obtain N such feature combinations, and evaluate the combination synergy of the N feature combinations in combination with the historical channel operation data, select the top K feature combinations with high combination synergy, and output them as the K coordinated feature groups, where N is greater than n and K is less than N.

4. The channel sedimentation probability early warning method for water flow feature recognition according to claim 3, characterized in that, Respectively taking the K coordinated feature groups as input variable groups, construct and train K benchmark siltation prediction models, and integrate the K benchmark siltation prediction models to obtain a siltation probability prediction model, including: Extract the K coordinated promotion degrees of the K coordinated feature groups; Based on the K coordinated promotion degrees, in combination with the historical channel operation data, correspondingly construct and train K groups of lightweight prediction models; Select the lightweight prediction model with the best performance in each group, obtain the K benchmark siltation prediction models for ensemble learning, and construct the siltation probability prediction model.

5. The channel siltation probability early warning method for water flow feature recognition according to claim 4, characterized in that Based on the key water flow feature index set, arrange sensors and monitor water flow, obtain real-time water flow monitoring data, including: Determine the monitoring requirements for each characteristic index based on the key water flow characteristic index set, configure corresponding types of sensors, and establish; Obtain the flow velocity data in each profile direction of the target channel through a flow velocity sensor, obtain the water level change data of each typical cross-section through a water level sensor, collect pressure pulsation signals through a turbulence sensor to construct a turbulent dynamic characteristic sequence, and obtain the suspended sediment concentration and bed sediment concentration in the water body through a sediment sensor, and output as the real-time water flow monitoring data.

6. The channel sedimentation probability warning method for water flow feature recognition according to claim 5, characterized in that, After being processed by feature engineering, input it into the sedimentation probability prediction model for predicting the sedimentation probability of the target channel, including: Decompose the key water flow characteristic index set and classify it into intuitive water flow characteristic indexes and deduced water flow characteristic indexes; Based on the preprocessed real-time water flow monitoring data, extract the intuitive water flow characteristic parameters corresponding to the intuitive water flow characteristic indexes; Calculate the deduced water flow characteristic parameters corresponding to the deduced water flow characteristic indexes according to the intuitive water flow characteristic parameters; Fuse the intuitive water flow characteristic parameters and the deduced water flow characteristic parameters, and process the fusion result by feature engineering methods to obtain real-time input variables; Transmit the real-time input variables to the input end of the sedimentation probability prediction model, and obtain the sedimentation probability prediction result from the output end.

7. The method according to claim 1, wherein According to the sedimentation probability prediction result, judge whether there is a sedimentation risk in the target channel, and issue a sedimentation probability warning when the risk meets the preset conditions, including: Set a sedimentation probability risk grading threshold adapted to the channel operation safety level, and establish a sedimentation risk discrimination criterion; Compare the sedimentation probability prediction result with the sedimentation risk discrimination criterion; When the sedimentation probability result exceeds the corresponding threshold of the discrimination criterion, it is determined that there is a sedimentation risk; Generate a warning instruction according to the sedimentation risk level, and send a sedimentation probability warning message to the channel operation management end through the warning interface.

8. A channel sedimentation probability warning system for water flow feature recognition, characterized in that A method for warning the sedimentation probability of a channel for water flow characteristic identification according to any one of claims 1 to 7, including: A characteristic index determination module, configured to obtain the characteristic parameters of the target channel, and input the characteristic parameters into a pre-constructed interactive knowledge graph to determine a key water flow characteristic index set for characterizing the water flow behavior of the channel; A coordinated feature group screening module, configured to use the key water flow characteristic index set as a feature space to perform replaceable random feature combinations, and screen the feature combination results based on historical channel operation data to determine K coordinated feature groups; A prediction model construction and integration module, configured to construct and train K baseline sedimentation prediction models with K coordinated feature groups as input variable groups respectively, and integrate the K baseline sedimentation prediction models to obtain a sedimentation probability prediction model; A water flow monitoring and sedimentation probability prediction module, configured to perform sensor layout and water flow monitoring based on the key water flow characteristic index set, obtain real-time water flow monitoring data, and input it into the sedimentation probability prediction model for predicting the sedimentation probability of the target channel after being processed by feature engineering; The sedimentation risk judgment and early warning module is used to judge whether there is a sedimentation risk in the target channel according to the sedimentation probability prediction result, and issue a sedimentation probability early warning when the risk meets the preset conditions.

Images