Method and system for real-time collection and intelligent fusion processing of polder area multi-site water conservancy data

Abstract

The invention relates to the technical field of intelligent water conservancy and automatic control, in particular to a polder area multi-station water conservancy data real-time collection and intelligent fusion processing method and system, and the method comprises the steps: a multi-dimensional collection system construction step: collecting time sequence state, fluid field and environment boundary data, and executing time base unified cleaning; an ideal model construction step: constructing an ideal water tank model based on the law of conservation of mass and generating an ideal water level change curve; a parameter injection simulation step: generating a simulation fault water level curve according to injection efficiency attenuation and medium leakage parameters; a double-track difference calculation step: respectively carrying out difference on actual measurement and simulation data and an ideal curve to generate real and theoretical residual vectors; a similarity judgment and control step: calculating sequence similarity to judge a real physical fault or sensor data noise, and generating a fusion control instruction; according to the invention, the technical problem of distinguishing the noise of the sensor from the real physical fault is effectively solved, and false alarm or missing alarm is avoided.

Description

Technical Field

[0001] This invention relates to the field of smart water conservancy and automated control technology, specifically to a method and system for real-time acquisition and intelligent fusion processing of water conservancy data from multiple stations in polder areas. Background Technology

[0002] In the existing polder water conservancy system, the system includes a monitoring center and multiple pumping stations and sluice gates distributed at various nodes. After the multiple pumping stations and sluice gates are connected to sensors and other control facilities through a communication network, they form a drainage and irrigation unit that can work together.

[0003] The monitoring system includes a water level gauge, a rain gauge, and a local controller. The water level gauge is connected to the local controller and compares the collected real-time water level data with a preset threshold. By determining whether the water level exceeds the limit, the system adjusts the start and stop status of the pump set so that the water conservancy system can drain water in areas with abnormally high water levels in a timely manner.

[0004] However, relying solely on direct comparison of sensor data makes it impossible to distinguish between sensor data noise caused by environmental factors such as wind and waves and real physical anomalies caused by equipment failures. This leads to false alarms or missed alarms in the system, resulting in inaccurate operation and scheduling decisions in the water conservancy system. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for real-time acquisition and intelligent fusion processing of multi-site water conservancy data in polder areas. This method effectively distinguishes between sensor data noise and actual physical faults, avoiding false alarms or missed alarms caused by environmental interference. It also identifies hidden faults that are difficult to detect using only electrical parameter monitoring and generates fusion control commands accordingly. This prevents equipment from frequently starting and stopping due to false fluctuations, extending the equipment's service life. Specifically, the technical solution of this invention is as follows:

[0006] A method for real-time acquisition and intelligent fusion processing of multi-site water conservancy data in polder areas includes the following steps:

[0007] A multi-dimensional data acquisition system for water conservancy in polder areas was constructed to collect time-series status data, fluid field data, and environmental boundary data, and time-base unified cleaning processing was performed on the collected data.

[0008] Based on the environmental boundary data, time-series state data and preset equipment rated parameters after cleaning, an ideal water tank model that follows the law of conservation of mass is constructed, and an ideal water level change curve is generated using the ideal water tank model.

[0009] Based on a pre-set fault knowledge base, efficiency decay parameters and medium leakage parameters are injected into the ideal water tank model to generate multiple sets of simulated fault water level curves.

[0010] The measured water level and the ideal water level change curve in the fluid field data are differentially calculated to generate a real residual vector. The theoretical residual vector is generated by differentially calculating the water level curves of multiple simulated faults and the ideal water level change curves.

[0011] The sequence similarity between the actual residual vector and the theoretical residual vector is calculated. Based on the comparison result between the sequence similarity and the preset similarity threshold, the current operating state of the water conservancy system is determined to be either a real physical fault or sensor data noise, and fusion control commands are generated accordingly.

[0012] Preferably, a multi-dimensional data acquisition system for polder water conservancy is constructed, collecting time-series status data, fluid field data, and environmental boundary data, including:

[0013] Collect pump start / stop status, gate opening, current, voltage and power as time-series status data;

[0014] Collect inland river water levels, external river water levels, and instantaneous flow rates as fluid field data;

[0015] Real-time rainfall, evaporation, and soil permeability are collected as environmental boundary data.

[0016] The time-base unification cleaning process includes using a time conversion algorithm to align the sampling frequency of all data to a unified time axis.

[0017] Preferably, based on the environmental boundary data after cleaning, time-series state data, and preset equipment rated parameters, an ideal water tank model following the law of conservation of mass is constructed, and an ideal water level change curve is generated using the ideal water tank model, including:

[0018] Establish the catchment area and runoff coefficient of the polder area, and calculate the inflow rate by combining real-time rainfall data;

[0019] Obtain the rated theoretical drainage capacity of the pump set under the current operating conditions;

[0020] A differential equation with time as the independent variable is established. The difference between the incoming flow rate and the rated theoretical drainage capacity is integrated to calculate the theoretical water level value sequence under the conditions of no equipment failure and no external water intrusion, thus forming an ideal water level change curve.

[0021] Preferably, based on a pre-set fault knowledge base, efficiency decay parameters and medium leakage parameters are injected into the ideal water tank model to generate multiple sets of simulated fault water level curves, including:

[0022] The preset efficiency decay factor is retrieved and applied to the rated theoretical drainage capacity of the pump set to simulate impeller winding or flow channel blockage conditions, generating the first type of simulated fault water level curve.

[0023] The preset gate leakage flow coefficient is retrieved and superimposed onto the input of the ideal water tank model to simulate the gate sealing failure condition and generate the second type of simulated fault water level curve.

[0024] Preferably, the measured water level and ideal water level change curves in the fluid field data are differentially calculated to generate a realistic residual vector, and multiple sets of simulated fault water level curves and ideal water level change curves are differentially calculated to generate a theoretical residual vector, including:

[0025] Subtract the corresponding value in the ideal water level change curve from the measured water level at the same time, and extract the mixed feature sequence containing real anomalies and environmental noise as the real residual vector.

[0026] The corresponding value in the ideal water level change curve is subtracted from the simulated fault water level curve at the same time, and the pure fault feature fingerprint sequence is extracted as the theoretical residual vector.

[0027] Preferably, the sequence similarity between the actual residual vector and the theoretical residual vector is calculated. Based on the comparison result of the sequence similarity with a preset similarity threshold, the current operating state of the water conservancy system is determined to be either a real physical fault or sensor data noise, including:

[0028] Calculate the cross-correlation coefficient between the actual residual vector and each set of theoretical residual vectors;

[0029] If the cross-correlation coefficient is greater than the preset matching threshold, it is determined that the actual residual vector matches the fault model to which the corresponding theoretical residual vector belongs, confirming that there is a real physical fault in the system, and identifying it as a mechanical efficiency fault or a hydraulic structure fault according to the matching model type.

[0030] If the cross-correlation coefficient is less than the preset matching threshold, and the fluctuation frequency or variance of the calculated real residual vector is greater than the preset noise judgment threshold, then it is determined to be non-physical sensor data noise, and the non-physical sensor data noise is filtered out.

[0031] Preferably, and based thereon generate fusion control instructions, including:

[0032] In response to the determination of a real physical fault, a hydrological state safety metric is calculated based on the difference between the current inland river water level and the preset warning water level. If the safety metric is lower than the preset safety threshold, a dispatching instruction is generated to force the pump group to start and the gate to close.

[0033] In response to the determination that the data is sensor noise, the current control strategy remains unchanged, and the corresponding measured water level is marked as a false fluctuation.

[0034] A multi-site water conservancy data real-time acquisition and intelligent fusion processing system for polder areas, including:

[0035] The data acquisition and cleaning module is used to build a multi-dimensional data acquisition system for polder water conservancy, collect time-series status data, fluid field data and environmental boundary data, and perform time-base unified cleaning processing on the collected data;

[0036] The ideal model building module is used to construct an ideal water tank model that follows the law of conservation of mass based on the cleaned environmental boundary data, time-series state data and preset equipment rated parameters, and to generate an ideal water level change curve using the ideal water tank model.

[0037] The parameter injection simulation module is used to inject efficiency decay parameters and medium leakage parameters into the ideal water tank model based on a preset fault knowledge base, and generate multiple sets of simulated fault water level curves.

[0038] The dual-track differential extraction module is used to perform differential calculations on the measured water level and ideal water level change curves in the fluid field data to generate a real residual vector, and to perform differential calculations on multiple sets of simulated fault water level curves and ideal water level change curves to generate a theoretical residual vector.

[0039] The coupled decision and control module is used to calculate the sequence similarity between the actual residual vector and the theoretical residual vector. Based on the comparison result of the sequence similarity and the preset similarity threshold, it determines whether the current operating state of the water conservancy system is a real physical fault or sensor data noise, and generates fusion control commands accordingly.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] 1. This invention constructs an ideal water tank model that follows the law of conservation of mass and generates an ideal water level change curve. It uses dual-track differential calculation to generate real residual vectors and theoretical residual vectors. By calculating the sequence similarity between the two, the system can accurately distinguish between sensor data noise caused by environmental factors such as wind and waves and real physical faults. This effectively solves the problem of false alarms or missed alarms caused by the inability to identify environmental interference in existing technologies, and significantly improves the accuracy of system fault diagnosis.

[0042] 2. This invention generates fusion control commands based on fault determination results, realizing differentiated scheduling strategies; for cases of real physical faults and low safety metrics, the system generates commands to force the pump group to start to ensure safety; for sensor data noise, the current control strategy remains unchanged and false fluctuations are marked; this mechanism avoids frequent invalid start-stop of equipment due to data fluctuations, and extends the service life of equipment while ensuring the hydrological safety of the polder area.

[0043] 3. This invention constructs a multi-dimensional acquisition system covering temporal state, fluid field and environmental boundary data, and aligns data with different sampling frequencies to a unified time axis through time base unification and cleaning processing. This not only eliminates the spatiotemporal deviation of multi-source heterogeneous data and establishes a unified calculation benchmark, but also ensures the completeness and temporal synchronization of the physical model input variables, laying a solid data foundation for subsequent high-precision differential calculation and state determination.

[0044] 4. This invention introduces a parameter injection simulation mechanism, which injects efficiency decay and medium leakage parameters into the ideal model based on the fault knowledge base, generating multiple sets of simulated water level curves and theoretical residual vectors that reflect different fault modes. This synthetic analysis method transforms abstract anomaly detection into a specific feature fingerprint matching problem, enabling the system to quantitatively identify hidden problems such as mechanical efficiency faults or hydraulic structure faults, thereby improving the level of intelligent operation and maintenance of water conservancy facilities. Attached Figure Description

[0045] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0046] Figure 1 This is a flowchart of the method of the present invention;

[0047] Figure 2 This is a structural diagram of the system of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0049] Example 1:

[0050] Please see Figure 1 The method for real-time acquisition and intelligent fusion processing of water conservancy data from multiple stations in the polder area includes the following steps: constructing a multi-dimensional acquisition system for water conservancy data in the polder area, acquiring time-series status data, fluid field data and environmental boundary data, and performing time-base unified cleaning processing on the acquired data;

[0051] Based on the cleaned environmental boundary data, time-series state data, and preset equipment rated parameters, an ideal water tank model following the law of conservation of mass is constructed, and an ideal water level change curve is generated using the ideal water tank model. Based on the preset fault knowledge base, efficiency decay parameters and medium leakage parameters are injected into the ideal water tank model to generate multiple sets of simulated fault water level curves.

[0052] The measured water level and the ideal water level change curve in the fluid field data are differentially calculated to generate a real residual vector. The theoretical residual vector is generated by differentially calculating the water level curves of multiple simulated faults and the ideal water level change curves.

[0053] The sequence similarity between the actual residual vector and the theoretical residual vector is calculated. Based on the comparison result between the sequence similarity and the preset similarity threshold, the current operating state of the water conservancy system is determined to be either a real physical fault or sensor data noise, and fusion control commands are generated accordingly.

[0054] This embodiment elaborates on the specific execution logic of the above-mentioned method for real-time acquisition and intelligent fusion processing of water conservancy data from multiple stations in the polder area. This method aims to solve the technical pain point in the prior art that it is difficult to distinguish sensor data noise, such as liquid level fluctuations caused by wind and waves, from real physical faults, such as mechanical efficiency faults or hydraulic structure faults, which leads to false alarms or missed alarms.

[0055] The core logic of this method lies in introducing synthetic analysis, which transforms the anomaly detection problem into a morphological matching problem of model residuals by constructing an ideal water tank model that follows the physical conservation laws.

[0056] The system executes the steps of building a multi-dimensional acquisition system. Through the sensor network deployed at various key nodes in the polder area, it acquires low-level data reflecting the system's operating status in real time and performs time-base unification and cleaning processing. It uses interpolation or resampling algorithms to map data of different frequencies to the same time axis to eliminate phase differences between data sources.

[0057] The system constructs an ideal water tank model and generates an ideal water level change curve. This ideal water tank model is based on the hydraulic balance equation, assuming that the system is in an ideal state with no faults, no leaks, and no noise, and calculates the ideal water level change curve as the zero point reference.

[0058] The system performs parameter injection simulation based on a fault knowledge base, actively introducing efficiency decay parameters and medium leakage parameters into the ideal water tank model to generate multiple sets of simulated fault water level curves reflecting different fault modes. On this basis, the system performs dual-track differential calculation to generate a real residual vector containing a mixture of real anomalies and environmental noise, and a theoretical residual vector representing a pure fault feature fingerprint. The system calculates sequence similarity and performs state determination. If the similarity is high, it is determined to be a real physical fault; if the similarity is low and the fluctuation frequency is high, it is determined to be sensor data noise, and fusion control commands are generated accordingly.

[0059] This embodiment effectively isolates environmental noise by establishing an ideal water tank model benchmark and performing dual-track differential analysis, because the random fluctuation pattern of noise cannot match the fault pattern generated based on physical mechanisms. At the same time, this method can identify hidden faults. For example, when the pump current is normal but the water level drops slower than the ideal water tank model, the system can accurately identify mechanical efficiency faults, avoiding blind spots in monitoring only electrical parameters.

[0060] Example 2:

[0061] A multi-dimensional data acquisition system for polder water conservancy was constructed, collecting time-series status data, fluid field data, and environmental boundary data. This included: collecting pump start-up and shutdown status, gate opening, current, voltage, and power as time-series status data; collecting inland river water level, outer river water level, and instantaneous flow rate as fluid field data; and collecting real-time rainfall, evaporation, and soil permeability as environmental boundary data. The time-base unification and cleaning process involved using a time conversion algorithm to align the sampling frequency of all data to a unified time axis.

[0062] This embodiment further specifies the steps for constructing a multi-dimensional data acquisition system for polder water conservancy in Embodiment 1, detailing the specific composition and cleaning process of the multi-dimensional acquisition system. The system acquires time-series status data from the equipment side, where the pump start-stop status is used as a Boolean switch quantity to determine whether the drainage item in the model is activated; the gate opening is used as a percentage value to calculate the flow area; and electrical parameters such as current, voltage, and power are acquired simultaneously to assist in verifying whether the equipment is powered on, preventing it from being mistakenly judged as having zero efficiency when not powered on.

[0063] Meanwhile, the system collects fluid field data from the water body side, including the inland river water level as the core control variable, the external river water level that determines the head pressure of self-drainage or backflow, and the instantaneous flow rate used to calibrate the actual operating point of the pump set; in addition, the system simultaneously collects environmental boundary data from the meteorological and geological side, including real-time rainfall as the main input source of the model, evaporation as a secondary loss term, and soil permeability that affects the runoff convergence rate.

[0064] To address the issue of inconsistent sampling frequencies among different sensors—for example, rain gauges use a tipping-bucket trigger mechanism while water level gauges use a timed polling mechanism—this embodiment employs an adaptive linear interpolation algorithm as the time conversion algorithm. Specifically, the sensor frequency with the highest sampling frequency in the acquisition system is selected, for example, the electrical parameter acquisition frequency is 1Hz, as the reference frequency for a unified time axis. For low-frequency data, such as rainfall or river water levels, the slope is calculated using the values ​​of two adjacent sampling points and their timestamps. Intermediate interpolation points are then generated according to the reference frequency. The specific calculation follows the formula: ;

[0065] in, Indicates a unified timeline The physical quantity values ​​after time alignment; The target time envelope in the original data The previous and next valid sampling timestamps, and the program enforces verification. If timestamps are duplicated due to sensor retransmission, duplicates will be automatically removed to avoid calculation errors where the denominator is zero. To correspond to the physical quantity values, all data is aligned to a unified time axis;

[0066] Addressing the issue of missing successor timestamps for the latest sampling point in a real-time system To address boundary conditions that prevent interpolation, this embodiment employs an adaptive extrapolation strategy: if the current system time... Exceeding the latest sampling point Based on and The slope is extrapolated linearly, and its calculation formula is as follows: ;

[0067] in, The rate of change of data, calculated based on the two most recent valid sampling points, is given by the following formula: ;

[0068] like Then a mandatory order To avoid calculation errors; or to employ a zero-order hold strategy when data fluctuates significantly, forcing the output. This ensures that at any given time... All systems have continuous data input, meeting the time continuity requirements of real-time control. However, to prevent the cumulative error caused by linear extrapolation from becoming infinitely divergent due to long-term sensor offline operation, this embodiment introduces a data validity time window (TTL) constraint in the adaptive extrapolation strategy: if the current system time... With the latest sampling point If the difference exceeds the preset safety threshold, such as 30 minutes, the system will stop extrapolation calculation, directly invalidate the data quality label at that moment, and trigger a data missing alarm to prevent fictitious data from misleading subsequent control decisions.

[0069] This embodiment constructs a three-in-one data acquisition system encompassing equipment, water body, and environment, ensuring the completeness of input variables for the physical model and avoiding non-convergence of model boundary conditions due to missing key parameters, such as the water level of the outer river. Time base unification processing eliminates data time deviations, ensuring the accuracy of real-time alignment during subsequent difference calculations.

[0070] Example 3:

[0071] Based on the cleaned environmental boundary data, time-series state data, and preset equipment rated parameters, an ideal water tank model following the law of conservation of mass is constructed, and an ideal water level change curve is generated using the ideal water tank model, including: establishing the catchment area and runoff coefficient of the polder area, calculating the inflow rate in combination with real-time rainfall; and obtaining the rated theoretical drainage capacity of the pump set under the current operating conditions.

[0072] A differential equation with time as the independent variable is established. The difference between the incoming flow rate and the rated theoretical drainage capacity is integrated to calculate the theoretical water level value sequence under the conditions of no equipment failure and no external water intrusion, thus forming an ideal water level change curve.

[0073] This embodiment is a further specification of the ideal water tank model construction steps in Embodiment 1, aiming to solve the problem of simulation distortion caused by the fuzzy definition of physical parameters in the general model; in order to quantify the theoretical changes in the water level of the polder, this embodiment constructs a dynamic equilibrium model based on differential equations;

[0074] The system establishes the catchment area of ​​the polder. With comprehensive runoff coefficient The inflow rate is calculated using real-time rainfall intensity; the specific inference formula is as follows: ;

[0075] in, Real-time rainfall intensity, unit , For catchment area, in units The formula includes complete unit conversions for output. The system obtains the rated theoretical drainage capacity of the pump set under the current operating conditions. The calculation is performed by calling the pre-stored QH performance curve in real time: ;

[0076] in, The coefficients of the quadratic polynomial, obtained by fitting the pump set's QH characteristic curve using the least squares method, correspond to the coefficients of the quadratic, linear, and constant terms, respectively, and are used to determine the head. Mapped to theoretical drainage capacity;

[0077] To avoid the fault tracing defects caused by ideal benchmark models relying on measured data that may contain faults, this embodiment explicitly specifies the head. The current water level of the outer river Compared with the ideal inland river water level calculated at the previous moment The difference, in Time can be initialized to This design is crucial: if measured water levels are used... The model calculates the head when a system leak causes an abnormal drop in water level. It will increase, assuming the water level of the outer river is higher, thus changing the theoretical discharge volume and affecting the ideal curve. This leads to a deviation from the fault-free trajectory and the tracking of measured anomalies, resulting in a significant reduction or even disappearance of the residuals;

[0078] Using the ideal water level calculated recursively as input ensures that the ideal model is a theoretical reference system that is completely independent of the actual operating state. Regardless of whether there is leakage or efficiency decline in the actual system, the ideal curve always reflects the expected state of no faults, thus ensuring that the residuals obtained by subsequent differential calculations can accurately quantify the degree of faults.

[0079] To ensure that the model strictly adheres to the law of conservation of mass and to correct the calculation bias caused by directly treating the entire catchment area as the bottom area for water level rise in traditional models, the system introduces the concept of effective storage water surface area (Asurface) and establishes the following differential equation: ;

[0080] in, for Ideal inland river water level at any given time, subscript This represents the ideal design value, expressed in meters (m). The time step for numerical integration is typically 1 second, consistent with the sampling interval. The effective water storage area within the polder area, in units of This parameter is not equal to the total catchment area. The calculation yields the following formula: ;

[0081] in, The regional water surface ratio, determined through remote sensing mapping, typically ranges from 0.05 to 0.15. The specific method for obtaining this ratio is as follows: acquire high-resolution multispectral satellite imagery of the polder area from the most recent quarter, and calculate the normalized difference water index. Water body pixels are extracted by setting a threshold, and the total area of ​​water body pixels is counted. Total coverage area of ​​the polder area The ratio as This parameter is updated quarterly to ensure that the model reflects the difference in water storage capacity between dry and wet seasons.

[0082] If used incorrectly Using it as the denominator will cause the calculated water level fluctuation range to be compressed by 10-20 times, resulting in serious physical distortion; For the system in Total output flow rate at any given time (including pump discharge and gate self-discharge), unit Regarding the issue of the undefined variable in the formula, this embodiment clarifies its calculation logic: It consists of the drainage capacity of the pump unit and the drainage capacity of the gate gravity flow, and its calculation formula is as follows: ;

[0083] in, For the pump set in The start / stop status at any given time, Boolean value: 1 = on, 0 = off; For the gate at The opening status indicator at any time, a Boolean value: 1 if the gate opening degree is greater than 0, otherwise 0; The gravity flow component is calculated based on the gate opening and water level difference using the broad-crested weir flow formula.

[0084] This ensures that the differential equations accurately reflect the physical drainage process of the system. Environmental loss flow mainly consists of surface evaporation and riverbed seepage, and the calculation formula is as follows: ;

[0085] in, This represents the real-time water surface evaporation rate. The values ​​represent riverbed permeability, all in mm / h. To effectively regulate water surface area; the denominator of the formula is used for unit conversion;

[0086] The system uses the Runge-Kutta method or the Euler method to numerically integrate the above differential equations, with initial conditions... Anchored to the measured water level after low-pass filtering, an ideal water level change curve with physical consistency is generated.

[0087] Example 4:

[0088] Based on a pre-set fault knowledge base, efficiency decay parameters and medium leakage parameters are injected into the ideal water tank model to generate multiple sets of simulated fault water level curves. These include: retrieving a pre-set efficiency decay factor and applying it to the rated theoretical discharge capacity of the pump set to simulate impeller winding or flow channel blockage conditions, generating the first type of simulated fault water level curve; and retrieving a pre-set gate leakage flow coefficient and superimposing it onto the input items of the ideal water tank model to simulate gate seal failure conditions, generating the second type of simulated fault water level curve.

[0089] This embodiment further specifies the parameter injection simulation steps in Embodiment 1, detailing the specific implementation mechanism of the parameter injection simulation. To simulate the water level behavior during a fault, this embodiment designs two typical fault injection modes and uses a parameter scanning method to generate a fault sample library covering different severity levels. The system executes the efficiency decay simulation in Mode 1, retrieving a preset discrete set of efficiency decay factors. ;

[0090] Iterate through each value in the collection subscript Represents the set of efficiency decay factors The first in Each element is applied to the rated theoretical drainage capacity of the pump unit; for the sake of consistency of terminology, the element defined in Example 3 is used here. This represents the rated drainage capacity under the current operating conditions, and each iteration is based on the simulated fault water level from the previous moment. By dynamically recalculating the QH curve to reflect the drift of the operating point, corresponding differential equation correction terms are generated, resulting in a series of... The first type of simulated fault water level curve; specifically, the injection calculation formula for the efficiency decay parameter is as follows: ;

[0091] in, express At the time of Actual water discharge under various efficiency degradation conditions; Based on the simulated fault water level at the previous moment The calculated head corresponds to the rated discharge capacity, rather than based on the current measured water level, in order to ensure accurate simulation of the operating point drift under fault conditions; This is the efficiency decay parameter;

[0092] The system performs a media leakage simulation in mode two, retrieving the preset gate leakage flow coefficient. And the preset set of leakage gap areas: ;

[0093] Traversal Each value in subscript Represents the set of pre-defined leakage gap areas The first in Each element, combined with the gate's cross-sectional area and the functional relationship between gravitational acceleration and water level difference, calculates the leakage flow rate. This flow rate is then used as input and superimposed onto an ideal water tank model to simulate hydraulic structure failure conditions, generating a series of... The second type of simulated fault water level curve; the injection calculation formula for the medium leakage parameters is as follows: ;

[0094] in, For the calculated first Superposition of directional leakage flow under various gap conditions This is the medium leakage flow coefficient, typically taken as 0.60 to 0.62. Let gravitational acceleration be approximately [value missing]. , The water level of the outer river. The ideal inland river water level calculated in Example 3; The direction sign is calculated using the following formula: ;

[0095] If the water level of the outer river is higher than that of the inner river, the leakage value is positive, indicating inflow; conversely, if the water level is lower, the leakage value is negative, indicating outflow.

[0096] Ideal inland river water level is used for this location. The reason for performing the calculation is specifically explained in this embodiment: In order to generate a pure fault feature fingerprint, the simulation model must be stripped of the wind and wave noise and random fluctuations contained in the actual sensor data; using a smooth ideal water level as a benchmark to calculate the leakage driving force can ensure that the generated theoretical residual vector contains only fault features, thereby effectively distinguishing high-frequency sensor noise through frequency domain differences in subsequent steps.

[0097] To complete the curve generation step, the system will use the above calculations... Sets and The sets are substituted into the differential equation solver of Example 3, and the pump group operating status parameters collected in Example 2 are forcibly introduced. Gating is applied to the drainage items, while ensuring the consistency of the time dependence of all flow variables with the area parameter, generating corresponding time-series curves: the first type of simulated fault water level curve cluster. The equation is solved iteratively, and its calculation formula is as follows: ;

[0098] Second type of simulated fault water level curve cluster The equation is solved iteratively, and its calculation formula is as follows: ;

[0099] The pump unit drainage volume subtracted here It needs to be based on the second type of simulated water level at the previous moment. Real-time calculations are performed to accurately reflect the reverse impact of water level changes caused by leakage on the pump head and drainage capacity, ensuring the closed-loop nature of the physical process.

[0100] This step explicitly maps discrete fault parameters into a continuous water level trajectory matrix that can be differentially compared with the measured water level, providing a fine reference coordinate system for subsequent pattern matching.

[0101] Example 5:

[0102] The measured water level and ideal water level change curves in the fluid field data are differentially calculated to generate a realistic residual vector. Furthermore, multiple sets of simulated fault water level curves and ideal water level change curves are differentially calculated to generate theoretical residual vectors, including:

[0103] Subtract the corresponding value in the ideal water level change curve from the measured water level at the same time to extract a mixed feature sequence containing real anomalies and environmental noise, which is used as the real residual vector; subtract the corresponding value in the ideal water level change curve from the simulated fault water level curve at the same time to extract a pure fault feature fingerprint sequence, which is used as the theoretical residual vector.

[0104] This embodiment is a further specification of the dual-track differential extraction step in Embodiment 1, detailing the specific calculation process of dual-track differential extraction. The core of this step is to extract the residual as a feature vector. The system performs the extraction of the real residual vector by subtracting the measured water level sequence from the ideal water level sequence point by point within a time window to generate the real residual vector. This vector physically represents the water volume deviation that is not explained by the model, which may be caused by faults or noise.

[0105] The system extracts the theoretical residual vector by subtracting the simulated fault water level curve from the ideal water level sequence point by point within the same time period to generate the theoretical residual vector. This vector physically represents the theoretical deviation trajectory caused purely by a specific fault. It is smooth and has a deterministic trend, such as exhibiting linear divergence or exponential divergence.

[0106] This embodiment, by subtracting the ideal water level, is equivalent to filtering out the DC component and normal trend in signal processing, retaining only the abnormal component; this differential processing greatly amplifies the signal characteristics of minor faults, making subtle leaks that were originally covered by tides or rainstorms clearly visible in the residual plot.

[0107] Example 6:

[0108] Calculate the sequence similarity between the actual residual vector and the theoretical residual vector. Based on the comparison result of the sequence similarity and the preset similarity threshold, determine whether the current operating state of the water conservancy system is a real physical fault or sensor data noise. This includes: calculating the cross-correlation coefficient between the actual residual vector and each set of theoretical residual vectors.

[0109] If the cross-correlation coefficient is greater than the preset matching threshold, it is determined that the actual residual vector matches the fault model to which the corresponding theoretical residual vector belongs, confirming that there is a real physical fault in the system, and identifying it as a mechanical efficiency fault or a hydraulic structure fault according to the matching model type.

[0110] If the cross-correlation coefficient is less than the preset matching threshold, and the fluctuation frequency or variance of the calculated real residual vector is greater than the preset noise judgment threshold, then it is determined to be non-physical sensor data noise, and the non-physical sensor data noise is filtered out.

[0111] This embodiment further specifies the sequence similarity calculation and judgment logic steps in Embodiment 5, focusing on supplementing the specific source and adaptive setting method of the threshold parameter; the system calculates the cross-correlation coefficient between the actual residual vector and each set of theoretical residual vectors. The formula uses the standard Pearson correlation coefficient calculation method and sets a zero-prevention mechanism for the denominator; if the maximum cross-correlation coefficient... If the value is greater than the preset matching threshold, such as 0.85, which is determined based on the upper quartile of the statistical distribution of the historical fault sample library, it is judged as a real physical fault. The key point is that if the correlation is low, the system needs to distinguish between normal deviation and sensor noise.

[0112] This embodiment calculates the variance of the actual residual vector. With fluctuation frequency To avoid missed or false judgments due to arbitrary setting of a preset noise threshold, this system uses a dynamic background noise calibration method to obtain this threshold: the system periodically, such as at dawn each day, selects a period of quiet time when there is no rainfall and the pumps are stationary, and calculates the variance of the measured water level during this period. If no quiet period meeting the conditions is detected within the preset period, the previously valid noise threshold or the preset default empirical threshold will be used to ensure continuous operation of the system under extreme weather conditions; and the noise variance threshold will be set as follows: ;

[0113] recommend To cover 3 The interval, and the frequency threshold is set to twice the system hydraulic response cutoff frequency, such as... ;like or If the signal is not clear, it is determined to be sensor data noise, such as wind, waves, or circuit interference, and a signal filtering operation is performed, marking the data at that moment as invalid and filling it with the ideal value; otherwise, it is determined to be a model error within the system's allowable range, and no alarm is triggered.

[0114] Example 7:

[0115] Based on this, a fusion control command is generated, including: in response to the determination of a real physical fault, calculating a hydrological state safety metric value based on the difference between the current inland river water level and the preset warning water level; if the safety metric value is lower than the preset safety threshold, generating a scheduling command to forcibly start the pump group and close the gate.

[0116] In response to the determination that the data is sensor noise, the current control strategy remains unchanged, and the corresponding measured water level is marked as a false fluctuation.

[0117] This embodiment further specifies the steps for generating fusion control commands in Embodiment 6, detailing the logic for generating fusion control commands. The system executes differentiated control strategies based on the judgment results. In cases of real physical faults, when the system confirms the existence of a hydraulic structure fault, such as backflow of external water or mechanical efficiency failure, the system calculates a hydrological state safety metric. The calculation formula is as follows: ;

[0118] in, This is the preset upper limit of the warning water level, such as 4.5m. This refers to the current measured inland river water level; in response to a safety metric value falling below a preset safety threshold, for example, a threshold value is set. That is, satisfying The conditions covered water levels approaching the warning line. Critical areas, and areas where water levels have exceeded limits. The overflow zone indicates that the water level has approached or exceeded the warning line. The system generates a forced dispatch command to force the standby pump group to start and sends a command to close or check the relevant gates. At the same time, it sends the highest level alarm to the central control room.

[0119] To prevent frequent start-stop operations (oscillations) of the equipment under critical conditions, this embodiment incorporates hysteresis control logic when generating forced commands and sets a reset threshold. ,For example Only when the calculated safety metric returns to Only when the above conditions are met will the system rescind the forced start command and revert to the normal automatic scheduling mode;

[0120] In the case of sensor data noise, when the noise is determined to be interference from ship waves, the system maintains the current control strategy, ignores the water level fluctuation during that period, maintains the original pump group operation status, and does not perform unnecessary start-up and shutdown operations.

[0121] At the same time, the data for this period is marked as spurious fluctuations in the historical database to prevent them from contaminating subsequent hydrological analysis models;

[0122] This embodiment realizes a closed loop from monitoring to control; when a fault is confirmed and the safety level is low, the system can break through conventional logic to force intervention and ensure the safety of the polder area; when noise is identified, the system maintains stability, avoiding frequent start-stop of the pump group due to sensor jitter, extending equipment life and reducing energy consumption.

[0123] Example 8:

[0124] Please see Figure 2The system for real-time acquisition and intelligent fusion processing of water conservancy data from multiple stations in the polder area includes: a data acquisition and cleaning module, which is used to construct a multi-dimensional acquisition system for water conservancy data in the polder area, acquire time-series status data, fluid field data and environmental boundary data, and perform time-base unified cleaning processing on the acquired data;

[0125] The ideal model building module is used to construct an ideal water tank model that follows the law of conservation of mass based on the cleaned environmental boundary data, time-series state data and preset equipment rated parameters, and to generate an ideal water level change curve using the ideal water tank model.

[0126] The parameter injection simulation module is used to inject efficiency decay parameters and medium leakage parameters into the ideal water tank model based on a preset fault knowledge base, and generate multiple sets of simulated fault water level curves.

[0127] The dual-track differential extraction module is used to perform differential calculations on the measured water level and ideal water level change curves in the fluid field data to generate a real residual vector, and to perform differential calculations on multiple sets of simulated fault water level curves and ideal water level change curves to generate a theoretical residual vector.

[0128] The coupled decision and control module is used to calculate the sequence similarity between the actual residual vector and the theoretical residual vector. Based on the comparison result of the sequence similarity and the preset similarity threshold, it determines whether the current operating state of the water conservancy system is a real physical fault or sensor data noise, and generates fusion control commands accordingly.

[0129] This embodiment discloses a real-time acquisition and intelligent fusion processing system for multi-site water conservancy data in a polder area. This system is the hardware and software implementation carrier of the method in Embodiment 1. The system includes the following core modules: the data acquisition and cleaning module is connected to the water level gauge, rain gauge, and PLC controller in hardware, and time synchronization and interpolation algorithms are deployed in software to unify the time base of the data.

[0130] The ideal model building module has a built-in calculation engine containing geographical information of the polder area, such as area and runoff coefficient, and equipment parameters, such as pump curves, for real-time solving of differential equations and outputting ideal water level change curves; the parameter injection simulation module stores a fault knowledge base containing empirical parameters such as efficiency decay factor and leakage coefficient, runs multiple simulation threads in parallel, and outputs simulated fault water level curves.

[0131] Based on this, the dual-track differential extraction module, as a vector operation unit, is responsible for performing the subtraction operation between the measured water level and the ideal water level, as well as the subtraction operation between the simulated water level and the ideal water level; the coupled decision and control module is equipped with a correlation analysis algorithm and a decision logic tree, which receives the residual vector and outputs control commands to the field PLC actuator;

[0132] This embodiment achieves decoupling of data flow and control flow through modular design; each module works in concert to map the real-time state of the physical world to the digital space for calculation and verification, and then feeds the decision results back to the physical world, thus forming a set of intelligent water conservancy control terminals with self-diagnosis and self-adaptation capabilities.

[0133] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for real-time acquisition and intelligent fusion processing of multi-site water conservancy data in polder areas, characterized in that, Includes the following steps: A multi-dimensional data acquisition system for water conservancy in polder areas was constructed to collect time-series status data, fluid field data, and environmental boundary data, and time-base unified cleaning processing was performed on the collected data. Based on the environmental boundary data, time-series state data and preset equipment rated parameters after cleaning, an ideal water tank model that follows the law of conservation of mass is constructed, and an ideal water level change curve is generated using the ideal water tank model. Based on a pre-set fault knowledge base, efficiency decay parameters and medium leakage parameters are injected into the ideal water tank model to generate multiple sets of simulated fault water level curves. The measured water level and the ideal water level change curve in the fluid field data are differentially calculated to generate a real residual vector. The theoretical residual vector is generated by differentially calculating the water level curves of multiple simulated faults and the ideal water level change curves. The sequence similarity between the actual residual vector and the theoretical residual vector is calculated. Based on the comparison result between the sequence similarity and the preset similarity threshold, the current operating state of the water conservancy system is determined to be either a real physical fault or sensor data noise, and fusion control commands are generated accordingly.

2. The method for real-time acquisition and intelligent fusion processing of multi-site water conservancy data in polder areas according to claim 1, characterized in that, The aforementioned multi-dimensional data acquisition system for polder water conservancy includes the collection of time-series status data, fluid field data, and environmental boundary data, including: Collect pump start / stop status, gate opening, current, voltage and power as time-series status data; Collect inland river water levels, external river water levels, and instantaneous flow rates as fluid field data; Real-time rainfall, evaporation, and soil permeability are collected as environmental boundary data. The time-base unification cleaning process includes using a time conversion algorithm to align the sampling frequency of all data to a unified time axis.

3. The method for real-time acquisition and intelligent fusion processing of multi-site water conservancy data in polder areas according to claim 1, characterized in that, Based on the cleaned environmental boundary data, time-series state data, and preset equipment rated parameters, an ideal water tank model following the law of conservation of mass is constructed, and an ideal water level change curve is generated using the ideal water tank model, including: Establish the catchment area and runoff coefficient of the polder area, and calculate the inflow rate by combining real-time rainfall data; Obtain the rated theoretical drainage capacity of the pump set under the current operating conditions; A differential equation with time as the independent variable is established. The difference between the incoming flow rate and the rated theoretical drainage capacity is integrated to calculate the theoretical water level value sequence under the conditions of no equipment failure and no external water intrusion, thus forming an ideal water level change curve.

4. The method for real-time acquisition and intelligent fusion processing of multi-site water conservancy data in polder areas according to claim 1, characterized in that, Based on a pre-set fault knowledge base, efficiency decay parameters and medium leakage parameters are injected into the ideal water tank model to generate multiple sets of simulated fault water level curves, including: The preset efficiency decay factor is retrieved and applied to the rated theoretical drainage capacity of the pump set to simulate impeller winding or flow channel blockage conditions, generating the first type of simulated fault water level curve. The preset gate leakage flow coefficient is retrieved and superimposed onto the input of the ideal water tank model to simulate the gate sealing failure condition and generate the second type of simulated fault water level curve.

5. The method for real-time acquisition and intelligent fusion processing of multi-site water conservancy data in polder areas according to claim 1, characterized in that, The method involves performing differential calculations on the measured water level and ideal water level change curves in the fluid field data to generate a realistic residual vector, and performing differential calculations on multiple sets of simulated fault water level curves and ideal water level change curves to generate a theoretical residual vector, including: Subtract the corresponding value in the ideal water level change curve from the measured water level at the same time, and extract the mixed feature sequence containing real anomalies and environmental noise as the real residual vector. The corresponding value in the ideal water level change curve is subtracted from the simulated fault water level curve at the same time, and the pure fault feature fingerprint sequence is extracted as the theoretical residual vector.

6. The method for real-time acquisition and intelligent fusion processing of multi-site water conservancy data in polder areas according to claim 5, characterized in that, The calculation of the sequence similarity between the actual residual vector and the theoretical residual vector, and the determination of whether the current operating state of the water conservancy system is a real physical fault or sensor data noise based on the comparison result of the sequence similarity with a preset similarity threshold, includes: Calculate the cross-correlation coefficient between the actual residual vector and each set of theoretical residual vectors; If the cross-correlation coefficient is greater than the preset matching threshold, it is determined that the actual residual vector matches the fault model to which the corresponding theoretical residual vector belongs, confirming that there is a real physical fault in the system, and identifying it as a mechanical efficiency fault or a hydraulic structure fault according to the matching model type. If the cross-correlation coefficient is less than the preset matching threshold, and the fluctuation frequency or variance of the calculated real residual vector is greater than the preset noise judgment threshold, then it is determined to be non-physical sensor data noise, and the non-physical sensor data noise is filtered out.

7. The method for real-time acquisition and intelligent fusion processing of multi-site water conservancy data in polder areas according to claim 6, characterized in that, The process of generating fusion control commands accordingly includes: In response to the determination of a real physical fault, a hydrological state safety metric is calculated based on the difference between the current inland river water level and the preset warning water level. If the safety metric is lower than the preset safety threshold, a dispatching instruction is generated to force the pump group to start and the gate to close. In response to the determination that the data is sensor noise, the current control strategy remains unchanged, and the corresponding measured water level is marked as a false fluctuation.

8. A real-time acquisition and intelligent fusion processing system for multi-site water conservancy data in polder areas, applied to the real-time acquisition and intelligent fusion processing method for multi-site water conservancy data in polder areas as described in any one of claims 1-7, characterized in that, include: The data acquisition and cleaning module is used to build a multi-dimensional data acquisition system for polder water conservancy, collect time-series status data, fluid field data and environmental boundary data, and perform time-base unified cleaning processing on the collected data; The ideal model building module is used to construct an ideal water tank model that follows the law of conservation of mass based on the cleaned environmental boundary data, time-series state data and preset equipment rated parameters, and to generate an ideal water level change curve using the ideal water tank model. The parameter injection simulation module is used to inject efficiency decay parameters and medium leakage parameters into the ideal water tank model based on a preset fault knowledge base, and generate multiple sets of simulated fault water level curves. The dual-track differential extraction module is used to perform differential calculations on the measured water level and ideal water level change curves in the fluid field data to generate a real residual vector, and to perform differential calculations on multiple sets of simulated fault water level curves and ideal water level change curves to generate a theoretical residual vector. The coupled decision and control module is used to calculate the sequence similarity between the actual residual vector and the theoretical residual vector. Based on the comparison result of the sequence similarity and the preset similarity threshold, it determines whether the current operating state of the water conservancy system is a real physical fault or sensor data noise, and generates fusion control commands accordingly.

Images