A multi-point ditch flow monitoring method

Abstract

The present invention discloses a multi-point ditch flow monitoring method, which specifically relates to the field of branch ditch flow monitoring and is used to solve the problem of complex branch ditch flow monitoring and optimization. The method comprises the following steps: constructing a branch ditch topological connection relationship matrix, combining flow and water level time series data to generate a flow characteristic curve and mark interference points, using dual-channel interactive analysis to generate an interference correlation matrix, extracting upstream and downstream flow offsets and disturbance propagation characteristics, judging the interference dominant mode by the degree of difference, and calculating the flow offset compensation matrix and the disturbance propagation correction matrix respectively; using a recursive optimization model to achieve segmented correction and reduce error residuals, combining spatiotemporal feature clustering to identify mutation sections and generate a key monitoring list for high-risk areas, extracting refined data based on a high-frequency acquisition strategy to analyze the causes of abnormalities, using a prediction model to generate dynamic correction results and update a flow balance data set, accurately correcting multi-mode interference, optimizing flow distribution, and improving monitoring accuracy and system stability.

Description

Technical Field

[0001] The present invention relates to the field of branch ditch flow monitoring, and more specifically, to a multi-point ditch flow monitoring method. Background Art

[0002] In farmland irrigation and ecological water networks, ditch systems usually present complex branching and interactive structures. Water flow is not limited to a single direction, but also exists in lateral series connection, branch diversion or parallel flow. Due to the needs of water resource management, multiple ditches may be interconnected through leakage, culverts or side flows, forming implicit water exchange. These complex flow characteristics make it impossible to simply regard the flow of a single ditch as an independent unit, but it is dynamically affected by the flow distribution, gate operation and water level changes between other ditches. In actual monitoring, in order to improve accuracy, it is often necessary to simultaneously monitor and comprehensively calculate the flow of these multi-branch ditches.

[0003] In multi-branch ditch flow monitoring, mutual interference between adjacent ditches may cause superposition of measurement data errors, which manifests as repeated flow measurement, hidden water loss or abnormal local measurement values. Specifically, first, the leakage and underflow paths between branch ditches are difficult to observe directly, resulting in the measurement values ​​of the monitoring points may contain unattributed side channel water. Second, flow interference between branches will cause part of the water flow to be repeatedly measured. Especially in the scenario where the branch canal merges into the main canal, the multi-point monitoring device may record the flow of the same water body separately, thereby causing the total flow to be falsely high. Third, due to the irregular distribution of ditches and relatively fixed monitoring points, the flow velocity and flow data of different branches are difficult to be fully synchronized on the time axis, which further amplifies the calculation error. The existence of these problems significantly increases the complexity of multi-point flow monitoring in data processing and comprehensive accounting, and puts higher requirements on the overall water balance analysis and regulation accuracy of the ditch. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a multi-point ditch flow monitoring method, which constructs a connection relationship matrix based on the branch ditch topology structure, combines the flow and water level time series data collected by the sensor, generates a branch flow characteristic curve and marks the interference point, uses dual-channel interactive analysis to generate an interference correlation matrix, extracts the upstream and downstream flow offset characteristics and disturbance propagation characteristics, and uses the difference to judge the dominant mode of the interference area. The flow offset compensation matrix and the disturbance propagation correction matrix are calculated for the perturbation-dominated area and the propagation-dominated area, respectively. The flow distribution is corrected in sections through a recursive optimization model to reduce the error residual, and the mutation section is identified by combining spatiotemporal feature clustering to generate a key monitoring list of high-risk areas. In addition, a high-frequency collection strategy is implemented in high-risk areas to extract refined flow data, analyze potential anomalies and dynamic impact mechanisms, and use a prediction model to generate dynamic correction results and update the flow balance data set, ultimately achieving multi-mode interference precision correction, multi-dimensional data fusion analysis, and global flow dynamic optimization, improving monitoring accuracy and system stability, so as to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A multi-point ditch flow monitoring method comprises the following steps:

[0007] S1: Construct a connection relationship matrix based on the branch-ditch topology structure to clarify the interaction paths and node positions between branches;

[0008] S2: Using the flow and water level time series collected by the sensor and combining it with the connection relationship matrix, the flow change characteristic curve of each branch is generated, the local fluctuation signal is extracted and the interference point is marked;

[0009] S3: For the marked interference points, the dual-channel interaction analysis method is used to calculate the correlation of the flow curves and generate an interference correlation matrix to identify the flow interaction intensity and range between branches;

[0010] S4: Based on the generated interference correlation matrix, the offset characteristics of upstream and downstream flow changes and the frequency domain characteristics of disturbance propagation between branches are extracted, and the difference between the two is used to determine the dominant mode of the interference area, and a correction matrix adapted to different modes is generated;

[0011] S5: Based on the determined interference pattern, a compensation matrix of flow deviation is calculated for the perturbation-dominated area, and a correction matrix of disturbance propagation is calculated for the propagation-dominated area;

[0012] S6: Input the correction matrix into the recursive optimization model, perform segmented correction on the marked interference area, gradually adjust the flow distribution of the interference area and reduce the overall error residual;

[0013] S7: Cluster the spatiotemporal characteristics of the optimized network-wide traffic distribution, identify mutation sections, and generate a key monitoring list of high-risk interference areas;

[0014] S8: Assign high-frequency collection strategies to high-risk areas in the key monitoring list;

[0015] S9: Use high-frequency collected data to establish a nonlinear prediction model for abnormal areas, generate dynamic correction results for high-risk areas, and output an updated data set for network-wide traffic balance.

[0016] In a preferred embodiment, the specific logic of step S1 is as follows:

[0017] Construct a connection relationship matrix based on the branch canal topology structure. Obtain the starting point, end point, confluence and diversion nodes, and upstream and downstream connection paths of the branch canal through geographic information system data and field survey results, and generate path topology data describing the interaction relationship between the branch canals; construct a branch connection relationship matrix based on the collected path information , the matrix elements Characterization Branch and The interaction strength between ,in is the distance between branches, is the scale factor, is the correlation coefficient of the flow time series.

[0018] In a preferred embodiment, the specific logic of step S2 is as follows:

[0019] The sensors installed at each branch section collect the flow and water level time series data in real time, and pre-process the collected time series data; use the connection relationship matrix to perform spatial weighted distribution on the flow data of each branch, specifically by calculating the Relative to its adjacent branches Traffic contribution , and superimpose the contribution of all related branches to generate the flow change characteristic curve of each branch ,in Representation and branching The set of all connected adjacent branches; Represents a branch In time The flow value at the moment; apply multi-scale wavelet transform to the generated flow change characteristic curve to decompose the signal into different frequency components; use the adaptive threshold method to dynamically adjust the anomaly detection standard and mark the fluctuation area that is significantly higher than the background noise as the interference point.

[0020] In a preferred embodiment, the specific logic of step S3 is as follows:

[0021] For the interference point, first, according to the time window where the interference point is located, from the flow change characteristic curve and its adjacent branches Extract the flow data of the corresponding time period to form a dual-channel signal pair , which is used to analyze the flow correlation between the interference point and the adjacent branch; then, the wavelet coherence analysis method is used to calculate the coherence spectrum of the signal pair at multiple frequency scales , the coherent spectrum value represents the branch and branches In frequency The traffic interaction intensity on is: ,in is the joint wavelet transform of the signal pair, and are their own wavelet transforms respectively; then, the coherence spectra of multiple frequency scales are converted into a single interaction value through frequency-weighted integral form , the formula is ,in is a frequency weight function, which is used to highlight the significant frequency band where the interference point is located; according to the set coherence strength threshold, all the branch pairs with significant interactions are screened out, and their interaction strength is recorded to determine the spatial correlation range of each interference point in the network; finally, the interaction strength results between all branches are organized into an interference correlation matrix .

[0022] In a preferred embodiment, the specific logic of step S4 is as follows:

[0023] Based on the structure of the interference correlation matrix and the interaction data between branches, the flow deviation characteristics of each interference area are first calculated; for each branch , by its flow characteristic curve and upstream and downstream branches Characteristic curve Flow characteristic curve By comparison, the deviation characteristics of upstream and downstream flows are extracted; the flow deviation index is defined as: ;in is the time window of the interference point; at the same time, the interference correlation matrix and wavelet analysis results are used to calculate the disturbance propagation characteristics of each interference area; the disturbance propagation index is defined as: ;in For branch and The disturbance propagation power spectral density is and is the frequency range of interference propagation; based on the relationship between the flow deviation index and the disturbance propagation index, the interference area is divided into two modes: if , it is determined to be a perturbation mode, indicating that the upstream and downstream flow matching deviation dominates the interference; if , determined as the propagation mode, indicating that the inter-branch disturbance propagation dominates the interference.

[0024] In a preferred embodiment, the specific logic of step S5 is as follows:

[0025] Calculation of flow offset compensation matrix for perturbation mode:

[0026] For the interference area determined as perturbation mode, flow mismatch is the main problem, which needs to be compensated by adjusting the upstream and downstream flow relationship; the flow offset compensation matrix is ​​defined as: ;in, Indicates branch and adjacent branches The flow compensation value, 2 is the dynamic compensation coefficient, which is adaptively adjusted according to the size of the flow deviation index, specifically: ,in is the threshold of the offset index, used to control the compensation intensity;

[0027] The disturbance propagation correction matrix of the propagation mode is calculated as:

[0028] For interference areas determined as propagation modes, the disturbance propagation path and intensity need to be optimized to reduce unnecessary propagation effects; the disturbance propagation correction matrix is ​​defined as: ;in, Indicates branch and adjacent branches The propagation correction value of is the dynamic correction coefficient, adjusted according to the disturbance propagation exponent: , through Dynamic adjustment ensures that the correction of high-intensity propagation areas has sufficient suppression effect.

[0029] In a preferred embodiment, the specific logic of step S6 is as follows:

[0030] The correction matrix, including the flow offset compensation matrix and the disturbance propagation correction matrix, is input into the recursive optimization model. The recursive optimization model aims to minimize the flow distribution error in the interference area and optimizes through the following logic:

[0031] S6.1, based on the global traffic distribution status and correction matrix, initialize the correction state of the interference area ; For each iteration, calculate the global error residual of the current state , the formula is: ;in, represents the set of all branches in the interference region, is the current iteration step number;

[0032] S6.2, in each iteration, according to the pattern determination result of the interference area, the application ratio of the flow offset compensation matrix and the disturbance propagation correction matrix is ​​dynamically adjusted; the specific fusion formula is: ;in and is the weight factor of the current iteration;

[0033] S6.3, according to the fused correction matrix , for the current state To update, the formula is: ;in is the step size factor;

[0034] S6.4, repeat S6.1-6.3, update the flow status , and recalculate the global error residual after each iteration; when the global error residual meets the preset convergence condition, the global error residual is less than the allowable error threshold, or when the maximum number of iterations is reached, the iteration is terminated;

[0035] S6.5, outputting the final iteration result as the optimized interference area traffic distribution.

[0036] In a preferred embodiment, step S7 includes the following contents:

[0037] For the optimized traffic distribution in the interference area, the following processing steps are performed to achieve spatiotemporal feature clustering and high-risk area identification:

[0038] The time series features of each branch are extracted from the optimized traffic distribution of the interference area, and the extracted features are standardized and calibrated as ; Using the connection relationship matrix and interference correlation matrix in the optimization results, calculate the spatial interaction characteristics of each branch, including the interaction strength and flow conductivity ; Combine temporal features and spatial features to form a feature vector , used for spatiotemporal feature clustering analysis; based on feature vector , use the density peak clustering algorithm to perform cluster analysis on all branches; for the spatiotemporal regions in the clustering results, calculate the anomaly index of each region , defined as: ;in, Indicates area The branch set within is the regional center feature vector; when the anomaly index exceeds the anomaly threshold, the corresponding area is determined to be a high-risk area, and the branches within it are marked as key monitoring branches; according to the high-risk area identification results, the key monitoring list of high-risk branches and their time windows is output.

[0039] In a preferred embodiment, step S8 includes the following contents:

[0040] According to the key monitoring list, the following processing logic is implemented for high-risk areas:

[0041] For each high-risk branch in the high-risk monitoring list and time window, dynamically adjust the sampling frequency and data collection density The specific adjustment is based on the abnormal index of the high-risk area and the interaction strength of the branch. The calculation formula is: ;in and are the basic sampling frequency and data acquisition density respectively.

[0042] In a preferred embodiment, step S9 includes the following contents:

[0043] For each high-risk branch , based on high frequency data collection ,in represents the change of branch flow over time, represents the change of branch water level over time, Indicates the change of branch flow velocity over time; the phase space of the time series is reconstructed by embedding the delayed coordinate method, and a dynamic prediction model of its flow anomaly characteristics is constructed; the formula is: ;in, For time delay, is the embedding dimension;

[0044] Based on the reconstructed phase space, the support vector regression model is used to predict the future flow state of each branch; the goal is to minimize the predicted value With the true value The error: ;in, is the regularization parameter, is the tolerance interval;

[0045] According to the deviation between the predicted value and the true value, a dynamic correction matrix is ​​constructed : ;in, is the interaction strength in the connection relationship matrix, is the adjustment coefficient, used to control the correction amplitude;

[0046] The dynamic correction matrix is ​​integrated into the optimized interference area traffic distribution, and the traffic status of each branch is updated as follows: ; Updated network-wide traffic balance dataset Contains dynamically corrected traffic status in high-risk areas.

[0047] The technical effects and advantages of a multi-point ditch flow monitoring method of the present invention are as follows:

[0048] The present invention constructs a connection relationship matrix based on the branch ditch topological structure, combines the flow and water level time series data collected by the sensor, generates a branch flow change characteristic curve and marks the interference points, uses a dual-channel interactive analysis method to generate an interference correlation matrix, further extracts the upstream and downstream flow offset characteristics and disturbance propagation characteristics, and judges the dominant mode of the interference area based on the difference between the two, thereby calculating the flow offset compensation matrix and the disturbance propagation correction matrix for the perturbation-dominated area and the propagation-dominated area, respectively, and corrects the marked interference area in sections through a recursive optimization model to reduce the error residual of the overall flow distribution, and combines spatiotemporal feature clustering to identify mutation sections, generates a key monitoring list of high-risk areas, allocates high-frequency acquisition strategies in high-risk areas, extracts refined flow data to analyze potential abnormal causes and dynamic influencing mechanisms, and finally uses a prediction model to generate dynamic correction results for high-risk areas and update the flow balance data set of the entire network, thereby realizing accurate identification and correction of multi-mode interference in complex branch ditch systems, efficient fusion and analysis of multi-dimensional data, and dynamic optimization of global flow balance, and significantly improving monitoring accuracy and system stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic flow chart of a multi-point ditch flow monitoring method of the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] Embodiment 1: Figure 1 The present invention provides a multi-point ditch flow monitoring method, comprising:

[0052] S1: Construct a connection relationship matrix based on the branch-ditch topology structure to clarify the interaction paths and node positions between branches;

[0053] S2: Using the flow and water level time series collected by the sensor and combining it with the connection relationship matrix, the flow change characteristic curve of each branch is generated, the local fluctuation signal is extracted and the interference point is marked;

[0054] S3: For the marked interference points, the dual-channel interaction analysis method is used to calculate the correlation of the flow curves and generate an interference correlation matrix to identify the flow interaction intensity and range between branches;

[0055] S4: Based on the generated interference correlation matrix, the offset characteristics of upstream and downstream flow changes and the frequency domain characteristics of disturbance propagation between branches are extracted, and the difference between the two is used to determine the dominant mode of the interference area, and a correction matrix adapted to different modes is generated;

[0056] S5: Based on the determined interference pattern, a compensation matrix of flow deviation is calculated for the perturbation-dominated area, and a correction matrix of disturbance propagation is calculated for the propagation-dominated area;

[0057] S6: Input the correction matrix into the recursive optimization model, perform segmented correction on the marked interference area, gradually adjust the flow distribution of the interference area and reduce the overall error residual;

[0058] S7: Cluster the spatiotemporal characteristics of the optimized network-wide traffic distribution, identify mutation sections, and generate a key monitoring list of high-risk interference areas;

[0059] S8: Assign high-frequency collection strategies to high-risk areas in the key monitoring list;

[0060] S9: Use high-frequency collected data to establish a nonlinear prediction model for abnormal areas, generate dynamic correction results for high-risk areas, and output an updated data set for network-wide traffic balance.

[0061] In branch ditch flow monitoring, the complex interaction paths and hidden water transmission between different branches pose significant challenges to the accurate attribution of flow data. In order to lay the foundation for subsequent data analysis and correction, it is necessary to build a topological model that can accurately describe the interaction relationship between branches and clarify the logical structure of each node and path, thereby providing a basic framework for data attribution for flow feature extraction and interference point marking.

[0062] The specific logic of step S1 is as follows:

[0063] Construct a connection relationship matrix based on the branch canal topology structure. Obtain the starting point, end point, confluence and diversion nodes, and upstream and downstream connection paths of the branch canal through geographic information system data and field survey results. Combined with dynamic characteristics such as gate opening, culvert location, and leakage channel, generate path topology data describing the interaction relationship between the branch canals. Construct a branch connection relationship matrix based on the collected path information. , the matrix elements Characterization Branch and The interaction intensity between them is calculated by using geometric distance, cross-sectional area and flow time series correlation, and is defined as ,in is the distance between branches, is the scale factor, is the correlation coefficient of the traffic time series; combining the connection relationship matrix and geographic coordinates, the traffic weight of each node is calculated , through the formula , calculates the traffic importance of a node in the network, where is the flow value of the adjacent node; finally, the connection relationship matrix and flow weight are dynamically corrected according to the historical monitoring data, and the elements and node weights of the connection relationship matrix are corrected to adapt to the dynamic characteristics changes caused by seasonal flow changes or topological structure adjustments, providing accurate topological relationships and flow attribution basis for subsequent steps.

[0064] After completing the construction of the connection relationship matrix based on the branch ditch topology, the interactive path of each node and its flow weight distribution are clarified, providing a reliable logical framework for subsequent steps. The connection relationship matrix and flow weight generated in this step are the direct input data for subsequent characteristic curve generation and interference point marking, ensuring the integrity and dynamic adaptability of flow data attribution, and solving the key problem of unclear initial data attribution under complex interactive paths.

[0065] After completing the construction of the connection relationship matrix based on the branch ditch topology structure (S1), the interaction paths and node locations between the branches have been clarified, providing a logical basis for the preliminary attribution of flow data. The next step is to use this connection relationship matrix, combined with the flow and water level time series data collected by the sensors installed on each branch section, to generate the flow change characteristic curve of each branch. This process requires the integration and processing of multi-source data to extract local fluctuation signals and mark possible interference points, laying the foundation for subsequent interference correlation analysis.

[0066] The specific logic of step S2 is as follows:

[0067] Step S2 collects flow and water level time series data from sensors installed at each branch section in real time, and ensures that all branch data are aligned under a unified time reference through high-precision time synchronization technology; preprocesses the collected time series data, including denoising, filling missing values ​​and standardization, to ensure data consistency; uses the connection relationship matrix constructed in step S1 to perform spatial weighted distribution of flow data of each branch, specifically by calculating the Relative to its adjacent branches Traffic contribution , and superimpose the contribution of all related branches to generate the flow change characteristic curve of each branch ,in Representation and branching The set of all connected adjacent branches; Represents a branch In time The flow value at the moment; apply multi-scale wavelet transform to the generated flow change characteristic curve, decompose the signal into different frequency components, and focus on extracting high-frequency local fluctuation signals to identify possible short-term flow disturbances or sudden anomalies; use the adaptive threshold method to dynamically adjust the anomaly detection standard and mark the fluctuation area that is significantly higher than the background noise as the interference point ; Combined with the water level time series, the authenticity of the marked interference points is verified through cross-validation technology to eliminate false alarms caused by natural water level fluctuations, and finally a list of interference points for each branch is generated.

[0068] By strictly using branches as unified names and combining the connection relationship matrix, step S2 successfully generated the flow change characteristic curve of each branch, and extracted and marked the potential interference points through wavelet transform and adaptive threshold method. The results of this step provide a reliable data basis for the subsequent construction of the interference association matrix (S3), while ensuring the consistency of symbols and names in the overall processing logic to avoid confusion and ambiguity.

[0069] In step S2, the flow change characteristic curve of each branch is generated and potential interference points are marked. However, the marking of these interference points is limited to the preliminary positioning of local flow fluctuations, and their interaction intensity and correlation range in space and time have not yet been clarified. In order to achieve in-depth quantification of interference points and their associated impacts, it is necessary to generate an interference correlation matrix based on these interference points using a dual-channel interaction analysis method in step S3 to accurately identify the interaction characteristics of traffic between branches.

[0070] The specific logic of step S3 is as follows:

[0071] For the interference point marked in step S2, first, according to the time window where the interference point is located, the flow change characteristic curve and its adjacent branches Extract the flow data of the corresponding time period to form a dual-channel signal pair , which is used to analyze the flow correlation between the interference point and the adjacent branch. Then, the wavelet coherence analysis method is used to calculate the coherence spectrum of the signal pair at multiple frequency scales. , the spectrum value represents the branch and branches In frequency The traffic interaction intensity on is: ,in is the joint wavelet transform of the signal pair, and Then, in order to extract the overall interaction characteristics of the interference point, the coherence spectrum of multiple frequency scales is converted into a single interaction value through the frequency-weighted integral form. , the formula is ,in is a frequency weight function, which is used to highlight the significant frequency band where the interference point is located. According to the set coherence strength threshold, all branch pairs with significant interactions are screened out. , and record their interaction strength to determine the spatial correlation range of each interference point in the network. Finally, the interaction strength results between all branches are organized into an interference correlation matrix , the matrix elements Represents a branch and branches The interaction strength under the interference point condition is set to The interference correlation matrix not only quantifies the interaction characteristics between each branch, but also provides direct data input for analyzing the dominant characteristics of the interference pattern in the subsequent steps.

[0072] Through dual-channel interaction analysis, step S3 generates an interference correlation matrix that accurately characterizes the interaction strength between branches. This matrix clarifies the multi-frequency interaction characteristics of the interference point from the frequency domain perspective, and screens out the branch pairs with significant interactions in combination with the spatial range, laying a rigorous logical foundation for the subsequent extraction of traffic offset characteristics and disturbance propagation characteristics (S4). This step is based on wavelet coherence analysis to ensure that subtle traffic correlation characteristics are captured under complex multi-scale interaction conditions while maintaining the consistency of the processing process.

[0073] In step S3, an interference correlation matrix is ​​generated based on the flow change characteristic curve and the timing window of the interference point through a dual-channel interactive analysis method. The matrix quantifies the flow interaction intensity and correlation range between branches. However, the specific characteristics of the interference area have not been further refined, and different areas may be dominated by upstream and downstream flow offsets or disturbance propagation between branches. In order to take targeted corrective measures in subsequent optimization, step S4 needs to extract the flow offset characteristics and disturbance propagation characteristics of the interference area, and judge the dominant mode of the interference area through the difference between the two, and generate a correction matrix adapted to different modes.

[0074] The specific logic of step S4 is as follows:

[0075] Based on the structure of the interference correlation matrix and the interaction data between branches, the flow deviation characteristics of each interference area are first calculated. , by its flow characteristic curve and upstream and downstream branches Characteristic curve Flow characteristic curve Compare and extract the deviation characteristics of upstream and downstream flows. The flow deviation index is defined as: ;in is the time window of the interference point. This index reflects the branch The degree of flow matching deviation between upstream and downstream.

[0076] At the same time, the disturbance propagation characteristics of each interference area are calculated using the interference correlation matrix and wavelet analysis results. The disturbance propagation index is defined as: ;in For branch and The disturbance propagation power spectral density is and is the frequency range of interference propagation. This index reflects the intensity and range of interference propagation in space.

[0077] Based on the relationship between the flow deviation index and the disturbance propagation index, the interference area is divided into two modes:

[0078] like , it is determined to be a perturbation mode, indicating that the upstream and downstream flow matching offset dominates the interference;

[0079] like , determined as the propagation mode, indicating that the inter-branch disturbance propagation dominates the interference.

[0080] The propagation mode refers to the characteristics of the interference area, which is mainly manifested as the lateral propagation of disturbances between branches. In this mode, the interference is not limited to a single branch, but spreads to other branches through a complex network structure, forming a cross-regional flow disturbance. Background technology points out that in a branch ditch network with high connectivity, local disturbances (such as short-term flow velocity fluctuations or sudden input changes) may be transmitted to other branches through interactive paths, especially at the intersection or diversion point of branch channels. The propagation effect of this disturbance is particularly significant. The propagation mode usually has the following characteristics: the interference is manifested as a change in the frequency of fluctuations, the propagation range is wide, and the interference intensity gradually decays with increasing distance, but the impact on the overall flow balance of the system is more complex and far-reaching.

[0081] By classifying the perturbation modes and propagation modes, targeted correction strategies can be formulated. For example, perturbations can be alleviated by adjusting the upstream and downstream flow matching, or the diffusion of propagation modes can be controlled by optimizing the disturbance propagation path, thus providing efficient and accurate interference correction means for the branch ditch flow monitoring system.

[0082] Through the analysis and calculation of step S4, the dominant mode of each interference area is clarified, and based on the size relationship between the flow deviation index and the disturbance propagation index, the interference area is divided into two types: perturbation mode and propagation mode. Corresponding correction matrices are generated for different modes, providing accurate input basis for recursive optimization. This step not only refines the interference characteristics through the extraction and judgment logic of two independent indexes, but also ensures the pertinence and dynamic adaptability of the subsequent correction method.

[0083] For the perturbation mode, the interference is mainly caused by the upstream and downstream flow offset; for the propagation mode, the interference is manifested as the propagation effect of the disturbance between branches. In step S5, it is necessary to generate adaptive correction matrices for different modes, including the flow offset compensation matrix of the perturbation mode and the disturbance propagation correction matrix of the propagation mode, and use them as direct inputs for subsequent optimization steps.

[0084] The specific logic of step S5 is as follows:

[0085] Calculation of flow offset compensation matrix for perturbation mode:

[0086] For interference areas determined to be perturbation modes, flow mismatch is the main problem and needs to be compensated by adjusting the upstream and downstream flow relationships.

[0087] The flow offset compensation matrix is ​​defined as: ;in, Indicates branch and adjacent branches The flow compensation value, 2 is the dynamic compensation coefficient, which is adaptively adjusted according to the size of the flow deviation index, specifically: ,in is the threshold of the offset index, which is used to control the compensation strength.

[0088] The disturbance propagation correction matrix of the propagation mode is calculated as:

[0089] For interference areas determined to be propagation modes, the disturbance propagation path and intensity need to be optimized to reduce unnecessary propagation impacts.

[0090] The disturbance propagation correction matrix is ​​defined as: ;in, Indicates branch and adjacent branches The propagation correction value of is the dynamic correction coefficient, adjusted according to the disturbance propagation exponent: , through Dynamic adjustment ensures that the correction of high-intensity propagation areas has sufficient suppression effect.

[0091] The flow offset compensation matrix of the perturbation mode and the disturbance propagation correction matrix of the propagation mode are calculated for the interference area respectively to obtain independent correction matrices for the two types of modes. Subsequently, according to the mode determination results of the interference area, the corresponding correction matrix is ​​directly output to the subsequent recursive optimization model to avoid the mismixing interference of the modes during the correction process.

[0092] In step S6, the correction matrix needs to be input into the recursive optimization model, and the flow data in the interference area needs to be corrected in sections in combination with the global flow distribution state, so as to gradually optimize the flow balance distribution in the interference area and reduce the overall error residual through iteration.

[0093] The specific logic of step S6 is as follows:

[0094] The correction matrix generated in step S5 (including the flow offset compensation matrix and the disturbance propagation correction matrix) is input into the recursive optimization model. The recursive optimization model aims to minimize the flow distribution error in the interference area and optimizes through the following logic:

[0095] S6.1, based on the global traffic distribution status and correction matrix, initialize the correction state of the interference area For each iteration, the global error residual of the current state is calculated , the formula is: ;in, represents the set of all branches in the interference region, is the current iteration step number.

[0096] Among them, the global flow distribution state refers to the overall distribution of the real-time flow of each branch in the branch ditch system in space and time, and is the core data representation that describes the dynamic operation characteristics of the entire ditch network. This state is composed of the flow characteristic curves of each branch and the connection relationship matrix, which comprehensively reflects the flow transfer between upstream and downstream, the interaction characteristics between branches, and the flow offset or disturbance diffusion in the local area. Through the global flow distribution state, the flow distribution of each branch and its relative balance can be systematically evaluated, providing a basic basis for the identification and correction of interference areas and the optimization of network-wide flow.

[0097] S6.2, in each iteration, according to the pattern determination result of the interference area, dynamically adjust the application ratio of the flow offset compensation matrix and the disturbance propagation correction matrix. The specific fusion formula is: ;in and is the weight factor of the current iteration, which is dynamically adjusted according to the changes in the pattern characteristics during the iteration process to ensure the adaptability of the correction matrix to the global optimization process.

[0098] S6.3, according to the fused correction matrix , for the current state To update, the formula is: ;in is the step size factor, which controls the adjustment amplitude of each update to avoid oscillation caused by overcorrection.

[0099] S6.4, repeat S6.1-6.3, update the flow status , and recalculate the global error residual after each iteration. When the global error residual meets the preset convergence condition, the global error residual is less than the allowable error threshold, or when the maximum number of iterations is reached, the iteration is terminated.

[0100] S6.5, output the final iteration result as the optimized interference area traffic distribution as part of the global traffic distribution, and record the local error and corrected state in the interference area to provide data support for subsequent analysis.

[0101] Through the recursive optimization process of step S6, the correction matrices of the perturbation mode and the propagation mode are successfully integrated into the global flow optimization process, which significantly reduces the error residual in the interference area and corrects the flow imbalance caused by upstream and downstream flow offset or disturbance propagation between branches. The recursive optimization model ensures the adaptability and flexibility of different modes by dynamically adjusting the application ratio of the correction matrix. At the same time, the global flow distribution is gradually optimized during the iteration process. The output correction results lay a solid foundation for the next step of spatiotemporal feature clustering analysis (S7).

[0102] In step S6, the interference area is corrected in sections using a recursive optimization model. By dynamically adjusting the flow offset compensation matrix and the disturbance propagation correction matrix, the global flow distribution state is optimized and the overall error residual is reduced. However, after the optimization of the flow distribution in the interference area, some areas may still have potential dynamic mutations, especially in highly complex branch intersection areas or high-frequency disturbance propagation paths. Step S7 requires the spatiotemporal feature clustering of the optimized global flow distribution, identifying mutation areas and generating a key monitoring list of high-risk interference areas to provide guidance for further precise analysis and subsequent adjustments.

[0103] Step S7 includes the following contents:

[0104] For the optimized traffic distribution in the interference area, the following processing steps are performed to achieve spatiotemporal feature clustering and high-risk area identification:

[0105] The time series features of each branch are extracted from the optimized traffic distribution in the interference area, including traffic change rate, fluctuation amplitude, periodic features, etc. In order to avoid the influence of feature dimension differences on clustering results, the extracted features are standardized and calibrated as .

[0106] Using the connection relationship matrix and interference correlation matrix in the optimization results, the spatial interaction characteristics of each branch, including the interaction strength, are calculated. and flow conductivity . Combine temporal features and spatial features to form a feature vector , used for spatiotemporal feature clustering analysis.

[0107] Based on the feature vector , the density peak clustering algorithm is used to perform cluster analysis on all branches. The algorithm calculates the local density and relative distance Determine the cluster center, the formula is: ;in By selecting points with high local density and large relative distance as cluster centers, the branch is divided into multiple spatiotemporal regions with similar characteristics.

[0108] For the spatiotemporal regions in the clustering results, calculate the anomaly index of each region , defined as: ;in, Indicates area The branch set within is the regional center feature vector. When the anomaly index exceeds the anomaly threshold, the corresponding area is judged as a high-risk area, and the branches within it are marked as key monitoring branches.

[0109] According to the high-risk area identification results, a key monitoring list of high-risk branches and their time windows is output, and the list is used as a direct input for the subsequent high-frequency acquisition strategy (S8).

[0110] Through the spatiotemporal feature clustering analysis in step S7, the optimized global traffic distribution state is successfully decomposed into multiple regions with similar spatiotemporal characteristics, and high-risk regions with significant mutations are identified. The anomaly index of the high-risk region quantifies the degree of characteristic deviation of the branches in the region, providing a reliable basis for accurate monitoring and interference warning. At the same time, the generated key monitoring list lays the foundation for subsequent refined collection and analysis (S8).

[0111] In step S7, through the spatiotemporal feature clustering analysis, the optimized global traffic distribution state is deeply analyzed, a key monitoring list is generated, and the branches and time windows of high-risk areas are clarified. However, the abnormal causes, dynamic characteristics and interference mechanisms of these high-risk areas have not yet been clarified, and higher-resolution real-time data acquisition and characteristic analysis are needed to provide reliable basic support for dynamic correction and subsequent optimization. In step S8, a high-frequency acquisition strategy will be assigned to the key monitoring list, and multi-dimensional refined data extraction and analysis will be performed on high-risk areas to further reveal the abnormal causes and dynamic mechanisms.

[0112] Step S8 includes the following contents:

[0113] According to the key monitoring list, the following processing logic is implemented for high-risk areas:

[0114] For each high-risk branch in the high-risk monitoring list and time window , dynamically adjust the sampling frequency and data collection density The specific adjustment is based on the abnormal index of the high-risk area and the interaction strength of the branch, and the calculation formula is: ;in and They are the basic sampling frequency and data collection density respectively. The higher the high risk index and interaction intensity, the higher the sampling frequency and data collection density.

[0115] In high-frequency acquisition mode, the real-time flow, water level, flow velocity and other multi-dimensional data of the branch are obtained from the sensor, and the collected data is synchronized and corrected in real time to ensure the spatiotemporal consistency between different branches and time windows. By extracting multi-dimensional characteristic data, the dynamic change of flow and spatial expansion characteristics of each high-risk branch are refined.

[0116] For high-frequency data collection, nonlinear time series analysis methods are used to extract dynamic characteristics from the data. Including:

[0117] Delayed coordinate reconstruction: embedding analysis of traffic time series to construct a high-dimensional dynamic phase space for identifying potential nonlinear features;

[0118] Dynamic complexity calculation: The dynamic complexity of high-frequency acquisition data is calculated using the multi-scale entropy method. The formula is: ;in Indicates high-risk branches The probability distribution of data under different states is used to characterize the diversity and abnormality of traffic dynamics;

[0119] Abnormal spectrum detection: Perform wavelet transform on traffic signals, extract high-frequency abnormal components, and identify the frequency distribution and intensity of interference.

[0120] Based on the extracted dynamic characteristics and combined with the branched spatial interaction network, the potential causes of abnormal areas are analyzed, including abnormal flow input, gate adjustment effects, imbalanced flow distribution, and interference characteristics caused by other physical conditions (such as topography or climate change).

[0121] Summarize the refined data and dynamic characteristics of each high-risk branch and generate an abnormal analysis report for the high-risk area. The report includes high-frequency acquisition results, dynamic complexity distribution, abnormal spectrum characteristics and potential cause analysis, providing a basis for the next step of dynamic correction (S9).

[0122] Through the high-frequency collection strategy, refined data extraction and dynamic characteristics analysis are carried out on the high-risk areas in the key monitoring list, and an abnormal analysis report containing traffic anomalies, dynamic complexity and spectrum characteristics is generated. However, the abnormal characteristics contained in these reports still need to be further quantified and modeled to predict the future traffic status of high-risk areas and dynamically correct their abnormal performance. Step S9 needs to combine high-frequency collection data and use nonlinear prediction models to dynamically correct high-risk areas, and finally output an updated full-network traffic balance data set to provide strong support for achieving system stability.

[0123] Step S9 includes the following contents:

[0124] For each high-risk branch , based on the high frequency acquisition data extracted in step S8 ,in represents the change of branch flow over time, represents the change of branch water level over time, Indicates the change of branch flow velocity over time; the method of embedding delayed coordinates is used to reconstruct the phase space of the time series and build a dynamic prediction model for its flow anomaly characteristics. The formula is: ;in, For time delay, To embed the dimension, we optimize it through the pseudo nearest neighbor algorithm and The choice of ensures that the reconstructed phase space can capture the nonlinear characteristics of high-risk areas.

[0125] Based on the reconstructed phase space, the support vector regression model is used to predict the future flow state of each branch. The goal is to minimize the predicted value With the true value The error: ;in, is the regularization parameter, is the tolerance interval. The model parameters are trained by historical data collected at high frequency. and bias terms to ensure that the model has a high prediction accuracy for future flow changes in abnormal areas.

[0126] According to the deviation between the predicted value and the true value, a dynamic correction matrix is ​​constructed : ;in, is the interaction strength in the connection relationship matrix, The dynamic correction matrix adjusts the flow of high-risk branches and their adjacent branches to eliminate the abnormal impact caused by the prediction deviation.

[0127] The dynamic correction matrix is ​​integrated into the optimized interference area traffic distribution, and the traffic status of each branch is updated as follows: ; Updated network-wide traffic balance dataset It contains the dynamically corrected flow status of high-risk areas, providing the system with the latest flow balance information.

[0128] Through the nonlinear prediction model and dynamic correction processing of step S9, the abnormal performance of high-risk areas was accurately predicted and corrected in real time, significantly improving the balance and stability of the traffic distribution of the entire network. The nonlinear prediction model effectively solves the problem of insufficient adaptability of traditional linear models to complex traffic dynamics by capturing the nonlinear characteristics in the traffic time series; the dynamic correction matrix achieves accurate adjustment of high-risk branches and their adjacent areas. The final output updated data set not only ensures the overall traffic balance of the system, but also provides a reliable basis for long-term traffic monitoring and regulation.

[0129] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0130] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0131] It should be noted that, in this article, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including a..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0132] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A multi-point ditch flow monitoring method, characterized in that: Includes steps: S1: Construct a connection relationship matrix based on the branch-ditch topology structure to clarify the interaction paths and node positions between branches; S2: Using the flow and water level time series collected by the sensor and combining it with the connection relationship matrix, the flow change characteristic curve of each branch is generated, the local fluctuation signal is extracted and the interference point is marked; S3: For the marked interference points, the dual-channel interaction analysis method is used to calculate the correlation of the flow curves and generate an interference correlation matrix to identify the flow interaction intensity and range between branches; S4: Based on the generated interference correlation matrix, the offset characteristics of upstream and downstream flow changes and the frequency domain characteristics of disturbance propagation between branches are extracted, and the difference between the two is used to determine the dominant mode of the interference area, and a correction matrix adapted to different modes is generated; S5: Based on the determined interference pattern, a compensation matrix of flow deviation is calculated for the perturbation-dominated area, and a correction matrix of disturbance propagation is calculated for the propagation-dominated area; S6: Input the correction matrix into the recursive optimization model, perform segmented correction on the marked interference area, gradually adjust the flow distribution of the interference area and reduce the overall error residual; S7: Cluster the spatiotemporal characteristics of the optimized network-wide traffic distribution, identify mutation sections, and generate a key monitoring list of high-risk interference areas; S8: Assign high-frequency collection strategies to high-risk areas in the key monitoring list; S9: Use high-frequency collected data to establish a nonlinear prediction model for abnormal areas, generate dynamic correction results for high-risk areas, and output an updated data set for network-wide traffic balance.

2. A multi-point ditch flow monitoring method according to claim 1, characterized in that: The specific logic of step S1 is as follows: Construct a connection relationship matrix based on the branch canal topology structure. Obtain the starting point, end point, confluence and diversion nodes, and upstream and downstream connection paths of the branch canal through geographic information system data and field survey results, and generate path topology data describing the interaction relationship between the branch canals; construct a branch connection relationship matrix based on the collected path information , the matrix elements Characterization Branch and The interaction strength between ,in is the distance between branches, is the scale factor, is the correlation coefficient of the flow time series.

3. A multi-point ditch flow monitoring method according to claim 2, characterized in that: The specific logic of step S2 is as follows: The sensors installed at each branch section collect the flow and water level time series data in real time, and pre-process the collected time series data; use the connection relationship matrix to perform spatial weighted distribution on the flow data of each branch, specifically by calculating the Relative to its adjacent branches Traffic contribution , and superimpose the contribution of all related branches to generate the flow change characteristic curve of each branch ,in Representation and branching The set of all connected adjacent branches; Represents a branch In time The flow value at the moment; apply multi-scale wavelet transform to the generated flow change characteristic curve to decompose the signal into different frequency components; use the adaptive threshold method to dynamically adjust the anomaly detection standard and mark the fluctuation area that is significantly higher than the background noise as the interference point.

4. A multi-point ditch flow monitoring method according to claim 3, characterized in that: The specific logic of step S3 is as follows: For the interference point, first, according to the time window where the interference point is located, from the flow change characteristic curve and its adjacent branches Extract the flow data of the corresponding time period from the , used to analyze the traffic correlation between the interference point and the adjacent branches; Then, the coherence spectrum of the signal pair at multiple frequency scales was calculated using the wavelet coherence analysis method. , the coherent spectrum value represents the branch and branches In frequency The traffic interaction intensity on is: ,in is the joint wavelet transform of the signal pair, and are their own wavelet transforms respectively; then, the coherence spectra of multiple frequency scales are converted into a single interaction value through frequency-weighted integral form , the formula is ,in is a frequency weight function, which is used to highlight the significant frequency band where the interference point is located; According to the set coherence strength threshold, all significantly interacting branch pairs are screened out and their interaction strength is recorded to determine the spatial correlation range of each interference point in the network; finally, the interaction strength results between all branches are organized into an interference correlation matrix .

5. A multi-point ditch flow monitoring method according to claim 4, characterized in that: The specific logic of step S4 is as follows: Based on the structure of the interference correlation matrix and the interaction data between branches, the flow deviation characteristics of each interference area are first calculated; for each branch , by its flow characteristic curve and upstream and downstream branches Characteristic curve Flow characteristic curve By comparison, the deviation characteristics of upstream and downstream flows are extracted; the flow deviation index is defined as: ;in is the time window of the interference point; at the same time, the interference correlation matrix and wavelet analysis results are used to calculate the disturbance propagation characteristics of each interference area; the disturbance propagation index is defined as: ;in For branch and The disturbance propagation power spectral density is and is the frequency range of interference propagation; based on the relationship between the flow deviation index and the disturbance propagation index, the interference area is divided into two modes: if , it is determined to be a perturbation mode, indicating that the upstream and downstream flow matching deviation dominates the interference; if , determined as the propagation mode, indicating that the inter-branch disturbance propagation dominates the interference.

6. A multi-point ditch flow monitoring method according to claim 5, characterized in that: The specific logic of step S5 is as follows: Calculation of flow offset compensation matrix for perturbation mode: For the interference area determined as perturbation mode, flow mismatch is the main problem, which needs to be compensated by adjusting the upstream and downstream flow relationship; the flow offset compensation matrix is ​​defined as: ;in, Indicates branch and adjacent branches The flow compensation value, 2 is the dynamic compensation coefficient, which is adaptively adjusted according to the size of the flow deviation index, specifically: ,in is the threshold of the offset index, used to control the compensation intensity; The disturbance propagation correction matrix of the propagation mode is calculated as: For interference areas determined as propagation modes, the disturbance propagation path and intensity need to be optimized to reduce unnecessary propagation effects; the disturbance propagation correction matrix is ​​defined as: ;in, Indicates branch and adjacent branches The propagation correction value of is the dynamic correction coefficient, adjusted according to the disturbance propagation exponent: , through dynamic adjustment.

7. A multi-point ditch flow monitoring method according to claim 6, characterized in that: The specific logic of step S6 is as follows: The correction matrix, including the flow offset compensation matrix and the disturbance propagation correction matrix, is input into the recursive optimization model. The recursive optimization model aims to minimize the flow distribution error in the interference area and optimizes through the following logic: S6.1, based on the global traffic distribution status and correction matrix, initialize the correction state of the interference area ; For each iteration, calculate the global error residual of the current state , the formula is: ;in, represents the set of all branches in the interference region, is the current iteration step number; S6.2, in each iteration, according to the pattern determination result of the interference area, the application ratio of the flow offset compensation matrix and the disturbance propagation correction matrix is ​​dynamically adjusted; the specific fusion formula is: ;in and is the weight factor of the current iteration; S6.3, according to the fused correction matrix , for the current state To update, the formula is: ;in is the step size factor; S6.4, repeat S6.1-6.3, update the flow status , and recalculate the global error residual after each iteration; when the global error residual meets the preset convergence condition, the global error residual is less than the allowable error threshold, or when the maximum number of iterations is reached, the iteration is terminated; S6.5, outputting the final iteration result as the optimized interference area traffic distribution.

8. A multi-point ditch flow monitoring method according to claim 7, characterized in that: Step S7 includes the following contents: For the optimized traffic distribution in the interference area, the following processing steps are performed to achieve spatiotemporal feature clustering and high-risk area identification: The time series features of each branch are extracted from the optimized traffic distribution of the interference area, and the extracted features are standardized and calibrated as ; Using the connection relationship matrix and interference correlation matrix in the optimization results, the spatial interaction characteristics of each branch, including the interaction strength, are calculated. and flow conductivity ; Combine temporal features and spatial features to form a feature vector , used for spatiotemporal feature clustering analysis; Based on the feature vector , use the density peak clustering algorithm to perform cluster analysis on all branches; for the spatiotemporal regions in the clustering results, calculate the anomaly index of each region , defined as: ;in, Indicates area The branch set within is the regional center feature vector; when the anomaly index exceeds the anomaly threshold, the corresponding area is determined to be a high-risk area, and the branches within it are marked as key monitoring branches; according to the high-risk area identification results, the key monitoring list of high-risk branches and their time windows is output.

9. A multi-point ditch flow monitoring method according to claim 8, characterized in that: Step S8 includes the following contents: According to the key monitoring list, the following processing logic is implemented for high-risk areas: For each high-risk branch in the high-risk monitoring list and time window, dynamically adjust the sampling frequency and data collection density ; The specific adjustment is based on the abnormal index of the high-risk area and the interaction strength of the branch. The calculation formula is: ;in and are the basic sampling frequency and data acquisition density respectively.

10. A multi-point ditch flow monitoring method according to claim 9, characterized in that: Step S9 includes the following contents: For each high-risk branch , based on high frequency data collection ,in represents the change of branch flow over time, represents the change of branch water level over time, Indicates the change of branch flow velocity over time; the phase space of the time series is reconstructed by embedding the delayed coordinate method, and a dynamic prediction model of its flow anomaly characteristics is constructed; the formula is: ;in, For time delay, is the embedding dimension; Based on the reconstructed phase space, the support vector regression model is used to predict the future flow state of each branch; the goal is to minimize the predicted value With the true value The error: ;in, is the regularization parameter, is the tolerance interval; According to the deviation between the predicted value and the true value, a dynamic correction matrix is ​​constructed : ;in, is the interaction strength in the connection relationship matrix, is the adjustment coefficient, used to control the correction amplitude; The dynamic correction matrix is ​​integrated into the optimized interference area traffic distribution, and the traffic status of each branch is updated as follows: ; Updated network-wide traffic balance dataset Contains dynamically corrected traffic status in high-risk areas.

Images