A water conservancy project intelligent operation and maintenance management method and system

Abstract

The application relates to the technical field of water conservancy engineering operation and maintenance, and particularly discloses a water conservancy engineering intelligent operation and maintenance management method and system, which comprises the following steps: collecting multi-source operation data of water conservancy engineering and forming an operation and maintenance data set; constructing an operation and maintenance state unit according to the spatial position, water flow direction and adjacent relationship of water conservancy facilities, and establishing a digital twin model for each operation and maintenance state unit; determining to-be-corrected parameters and physical constraint intervals of the digital twin model, and performing parameter optimization correction based on a particle swarm algorithm; calculating particle swarm diversity indexes, parameter physical consistency indexes and cross-condition prediction errors in the optimization process, and identifying premature convergence risks according to the indexes; and when the premature convergence risks exist, performing disturbance reconstruction and multi-swarm split search. The application can avoid prediction failure of the digital twin model caused by local optimal parameters, and improve the reliability of water conservancy engineering risk identification, task disposal and operation and maintenance archiving.

Description

Technical Field

[0001] This invention relates to the field of water conservancy project operation and maintenance technology, specifically to a smart operation and maintenance management method and system for water conservancy projects. Background Technology

[0002] Water conservancy projects are crucial infrastructure for ensuring flood control and drainage, agricultural irrigation, urban and rural water supply, water resource allocation, and ecological water replenishment. Reservoirs, dams, gates, pumping stations, river cross-sections, water conveyance pipelines, culverts, and drainage facilities are susceptible to operational risks due to various factors during long-term operation, including rainfall changes, water level fluctuations, siltation, equipment aging, frequent opening and closing, foundation settlement, and human intervention. These risks include abnormal gate opening and closing, increased pumping station vibration, abnormal dam seepage pressure, reduced river drainage capacity, pipeline blockage, and abnormal local displacement. With the development of IoT monitoring, video inspection, and digital twin technologies, existing water conservancy project operation and maintenance systems can now collect data on water level, rainfall, flow rate, gate opening, pumping station status, and structural safety. Furthermore, digital twin models can be used to simulate and predict water level changes, flow response, equipment operating status, and dam seepage conditions. In actual operation, river channel roughness can change due to siltation, revetment damage, weed growth, or cross-sectional changes; gate flow coefficients can shift due to gate wear, partial blockage, incomplete opening and closing, or changes in flow conditions; pump station efficiency parameters can decrease due to impeller wear, pipeline blockage, motor performance degradation, or changes in water intake conditions; dam permeability coefficients and seepage hysteresis coefficients can change due to seepage channel development, local crack expansion, changes in dam material condition, or water level fluctuations; and pipe and canal resistance coefficients can change due to sediment deposition, foreign object blockage, changes in inner wall roughness, or local damage. Once these parameters deviate from the actual state, although the digital twin model can still obtain a small fitting error through parameter compensation within a short time window, its prediction results will lack stable physical reliability.

[0003] Existing digital twin parameter calibration typically uses the fitting error of measured water level, flow rate, seepage pressure, displacement, or equipment operation data as the optimization objective, and employs swarm intelligence optimization methods such as particle swarm optimization (PSO) for parameter optimization. PSO is characterized by its simplicity, fast search speed, and suitability for multi-parameter optimization; however, it is prone to insufficient population diversity under complex engineering constraints. During iteration, particles may prematurely cluster near a local extremum, leading to premature convergence. This local optimum may exhibit a small numerical fitting error at the current moment, but its corresponding combination of physical parameters often deviates from the true physical state.

[0004] For example, the model might mask actual upstream siltation through abnormally high channel roughness, actual gate opening deviations through understated gate flow coefficients, impeller wear or pipeline blockages through incorrect pump station efficiency parameters, and actual seepage anomalies through distorted dam permeability coefficients. While these parameters may make the simulated curves fit historical data within the current time window, they do not reflect the actual operating mechanisms of water conservancy facilities. Once operating conditions change, such as heavy rainfall, a sudden increase in upstream inflow, gate scheduling changes, pump station start-up and shutdown changes, diurnal temperature variations, or downstream water level fluctuations, the digital twin model built on locally optimal parameters will quickly fail, leading to significant deviations in predicted water levels, flow rates, seepage pressure, displacement, or equipment status.

[0005] This deviation directly impacts risk level assessment, task prioritization, and maintenance dispatch decisions, leading to false alarms, missed alarms, or delayed responses. Existing intelligent maintenance platforms primarily focus on data collection, anomaly alarms, and task management, lacking mechanisms to identify and suppress the problem of "small current fitting error but distorted physical parameters" during digital twin parameter calibration. They also lack a complete solution for further applying the results of premature convergence suppression to upstream and downstream risk prediction and closed-loop maintenance verification. Therefore, identifying and suppressing premature convergence of particle swarms during the self-calibration of digital twin model parameters in water conservancy projects, preventing the model from falling into local optima with small fitting errors but incorrect physical parameters, and achieving closed-loop verification of upstream and downstream risk prediction, maintenance task dispatch, and response effectiveness based on the calibrated model, has become a pressing technical problem to be solved in the intelligent maintenance management of water conservancy projects. Summary of the Invention

[0006] This invention provides a smart operation and maintenance management method and system for water conservancy projects, aiming to solve the problems in related technologies such as the tendency of digital twin model parameter correction to fall into local optima, small current fitting error but distorted physical parameters, prediction failure after changes in operating conditions, and lack of closed-loop verification in operation and maintenance task handling.

[0007] A smart operation and maintenance management method for water conservancy projects includes the following steps: S1. Collect multi-source operation data of water conservancy projects and aggregate them according to facility location, equipment type and collection time to form operation and maintenance dataset; S2. Construct multiple operation and maintenance status units based on the spatial location, water flow direction, and adjacent relationships of water conservancy facilities, and establish digital twin models for each operation and maintenance status unit; S3. Determine the parameters to be corrected and the physical constraint range of the digital twin model, and optimize and correct the parameters to be corrected based on the particle swarm optimization algorithm; S4. During the optimization process of the particle swarm optimization algorithm, calculate the particle swarm diversity index, the parameter physical consistency index, and the cross-condition prediction error, and identify the risk of premature convergence based on the above indicators. S5. When the risk of premature convergence is identified, perturbation reconstruction and multi-subgroup split search are performed on the particle swarm to obtain a digital twin model after parameter correction. S6. Use the digital twin model after parameter correction to predict the operating status of each operation and maintenance status unit and its upstream and downstream related units, and generate dynamic risk values ​​based on the prediction results. S7. Generate operation and maintenance tasks and dispatching schemes based on the dynamic risk values; S8. Receive the on-site handling results and perform closed-loop verification based on sensor data before and after handling, digital twin prediction results, and inspection image data. Based on the closed-loop verification results, execute task archiving or secondary handling.

[0008] Preferably, the multi-source operational data includes water level data, flow rate data, gate opening data, hoist current data, pump station operation data, motor current data, motor temperature rise data, vibration data, seepage pressure data, displacement data, rainfall data, video inspection image data, manual inspection records, dispatch instruction data, and historical maintenance records.

[0009] Preferably, the operation and maintenance status unit is divided according to reservoir, dam, gate, pumping station, river cross section, water conveyance pipeline, culvert or drainage facility; the digital twin model includes water level and flow response model, gate flow model, pumping station operation model, dam seepage model or pipeline resistance model; the parameter to be corrected includes river roughness. Gate flow coefficient Pump station efficiency parameters Dam permeability coefficient Pipeline resistance coefficient Equipment attenuation coefficient or seepage hysteresis coefficient The parameters to be corrected constitute the particle parameter vector. The particle parameter vector satisfy: in, Indicates the roughness of the river channel. Indicates the gate flow coefficient. This indicates the efficiency parameters of the pumping station. Indicates the permeability coefficient of the dam. Indicates the pipe and channel resistance coefficient. Indicates the equipment attenuation coefficient. This represents the seepage hysteresis coefficient.

[0010] Preferably, the fitness function of the particle swarm optimization algorithm From the measured fitting error Parameter physical constraint penalty item and continuous error in upstream and downstream water volume Together, the fitness function satisfy: in, This represents the fitness function value. This represents the measured fitting error between the predicted output of the digital twin model and the actual operating data. This represents the physical constraint penalty term for parameters that deviate from the physical constraint range or physical correlation relationship of the parameter to be corrected. This indicates the continuous error in water volume between the current operation and maintenance status unit and its upstream and downstream related units. , , They represent the measured fitting errors, respectively. Parameter physical constraint penalty item and continuous error in upstream and downstream water volume The weighting coefficients.

[0011] Preferably, the particle swarm diversity index Based on particle parameter vector With the particle swarm center parameter vector The particle swarm diversity index is obtained by calculating the average distance between them. satisfy: in, Indicators representing particle swarm diversity Indicates the number of particles. Indicates the first The parameter vector of each particle. Represents the particle swarm center parameter vector. The upper limit parameter vector representing the physical constraint interval. A vector of parameters representing the lower bound of the physical constraint interval. Indicates the first The parameter vector of each particle With the particle swarm center parameter vector The distance between them It represents the scale range of the physical constraint interval.

[0012] Preferably, the physical consistency index of the parameters satisfy: in, This represents the physical consistency index of the parameters. This represents the physical constraint penalty term for the parameters. This indicates continuous error in upstream and downstream water volume. This indicates the trend consistency error; the cross-condition prediction error satisfy: in, This indicates the prediction error across operating conditions. Indicates the number of test case verification sets. Indicates the first The first working condition verification set Similar to actual test running data, The digital twin model represents the first The first output of the working condition verification set Predictive running data, Indicates the first The allowable fluctuation range of the class of operational data; the operational condition verification set includes the rainfall operational condition verification set, the gate scheduling operational condition verification set, the pump station start-up and shutdown operational condition verification set, or the water level rise and fall operational condition verification set.

[0013] Preferably, the particle swarm is considered to have a risk of premature convergence when the following criteria are met: in, Indicators representing particle swarm diversity This indicates a preset diversity threshold. This represents the measured fitting error. This indicates the preset fitting error threshold. This represents the physical consistency index of the parameters. This indicates a preset consistency threshold. This indicates the prediction error across operating conditions. This indicates the preset cross-operating condition error threshold.

[0014] Preferably, the dynamic risk value Based on the predicted deviation value Impact value of upstream and downstream linkages Trend evolution value Importance level of facilities and historical maintenance is in full swing The calculated dynamic risk value satisfy: in, Indicates dynamic risk value. This represents the prediction deviation value. This indicates the impact value related to upstream and downstream connections. Indicates the trend evolution value. Indicates the importance level of the facility. This indicates that the historical maintenance period is in full swing. , , , , These represent the prediction deviation values. Impact value of upstream and downstream linkages Trend evolution value Importance level of facilities and historical maintenance is in full swing The weighting coefficients.

[0015] Preferably, the closed-loop verification includes calculating the closed-loop verification value. The closed-loop verification value satisfy: in, This represents the closed-loop verification value. Indicates the degree of recovery from prediction bias. Indicates the degree of recovery of upstream and downstream connections. Indicates the degree of hazard removal in the inspection images. Indicates the degree of confirmation by manual retesting. , , , These represent the degree of recovery from prediction bias. Degree of recovery of upstream and downstream connections , degree of hazard removal in inspection images Confirmation level by manual retesting The weighting coefficients; when ,and , When the closed-loop review results are deemed to meet the defect elimination conditions, then... This represents the closed-loop review threshold. This indicates the predicted deviation value after treatment. This indicates the allowable prediction bias threshold. This indicates the impact value on upstream and downstream relationships after the treatment. This indicates the threshold that allows upstream and downstream connections to influence the data.

[0016] A smart operation and maintenance management system for water conservancy projects includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a smart operation and maintenance management method for water conservancy projects.

[0017] By adopting the above technical solution, the beneficial effects of the present invention are as follows: By utilizing a digital twin model with corrected parameters, the operational status of each maintenance unit and its upstream and downstream related units is predicted. Dynamic risk values ​​are generated by combining prediction bias, upstream and downstream impacts, trend evolution, facility importance level, and historical maintenance positive values. This allows risk assessment to reflect both the current degree of anomaly and the trend of its propagation, spread, or amplification along the water flow direction. Maintenance tasks and dispatch plans are generated based on these dynamic risk values. After on-site handling, a closed-loop review is performed based on sensor data, digital twin prediction results, and inspection image data. This ensures that potential hazards have truly been eliminated, avoiding the residual hazards caused by relying solely on manually entered completion status information. Attached Figure Description

[0018] Figure 1 This is a flowchart of the method of the present invention; Figure 2 A diagram showing the relationship between the operation and maintenance status unit and the digital twin model; Figure 3 Flowchart for identifying and suppressing premature convergence; Figure 4 This is a schematic diagram of the convergence identification curve for premature convergence. Detailed Implementation

[0019] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0020] like Figures 1-4 As shown, a smart operation and maintenance management method for water conservancy projects includes the following specific steps S1 to S8.

[0021] S1. Collect multi-source operational data of water conservancy projects and aggregate them according to facility location, equipment type, and collection time to form an operation and maintenance dataset. The operational status of water conservancy projects is jointly determined by water conditions, engineering conditions, rainfall, equipment status, structural safety status, dispatch instructions, and on-site images. If the system uses only single sensor data for judgment, it is easy to misjudge normal dispatch responses as abnormalities, and it is also easy to miss hidden risks caused by mismatch between upstream and downstream water flow, equipment performance degradation, or structural seepage changes. Therefore, this invention first establishes a data acquisition system with clear data sources and unifies data from different sources under the same facility object and the same time axis.

[0022] Specifically, water level data comes from water level gauges, radar water level gauges, or pressure water level sensors; flow rate data comes from flow meters, cross-sectional velocity meters, or flow rates calculated from cross-sectional water level-flow rate curves; gate opening data comes from gate opening sensors, hoist encoders, or gate PLC control systems; hoist current data comes from hoist electrical control cabinets, current transformers, or motor protectors; pump station operation data comes from the pump station SCADA system, pump unit control cabinet, and unit operation monitoring system; motor current, motor temperature rise, and vibration data come from smart meters, temperature sensors, and vibration sensors, respectively. Dynamic sensors; seepage pressure data comes from piezometers or pore water pressure gauges; displacement data comes from GNSS displacement monitoring devices, crack gauges, settlement gauges, or tilt sensors; rainfall data comes from rain gauges, meteorological stations, or water conservancy remote sensing rainfall terminals; video inspection image data comes from fixed cameras, drone inspection equipment, mobile inspection terminals, or water conservancy video monitoring platforms; manual inspection records come from mobile operation and maintenance terminals; dispatch instruction data comes from water conservancy dispatch systems, gate control platforms, or pump station dispatch platforms; historical maintenance records come from operation and maintenance work order systems, equipment ledger systems, or historical maintenance archives.

[0023] Furthermore, the system at each data acquisition moment Forming the original runtime data vector It satisfies: in, express The original runtime data vector at time step, Indicates water level data. Represents traffic data, This indicates the gate opening data. This indicates the current data of the gate opener. This indicates the start / stop or operating status data of the pumping station. This indicates the motor current data. This indicates the motor temperature rise data. Representing vibration data, This represents the osmotic pressure data. Represents displacement data. Represents rainfall data. This represents the inspection image data. Represents scheduling instruction data, This indicates manual inspection records or historical maintenance records.

[0024] Furthermore, the system is based on facility number. Equipment type Geographic coordinates and collection time For the original running data vector The dataset is aggregated to obtain the operation and maintenance dataset. The operation and maintenance dataset satisfy: in, Indicates the water conservancy facility number, Indicates the type of water conservancy facility. Indicates the geographical coordinates or station number of water conservancy facilities. Indicates the collection time. This represents the original running data vector at the corresponding time point. For data with different sampling frequencies, the system uses a uniform time step. Time alignment is performed; for data that was not collected, the system records a missing identifier and uses the missing identifier as the basis for confidence correction in subsequent calculations, rather than directly replacing the original data with the interpolation result.

[0025] S2. Multiple operation and maintenance (O&M) status units are constructed based on the spatial location, flow direction, and adjacency relationships of water conservancy facilities, and digital twin models are established for each O&M status unit. Water conservancy project risks exhibit significant spatial transmission characteristics. When an anomaly occurs in a gate, pumping station, river cross-section, or pipeline, its impact propagates along the flow direction, affecting the water level, flow rate, and equipment operating status of adjacent facilities. This invention divides water conservancy projects into multiple O&M status units and establishes a corresponding digital twin model in each O&M status unit, ensuring that subsequent parameter correction, risk prediction, and task assignment all target facility units with clearly defined physical boundaries.

[0026] like Figure 2 As shown, multiple operation and maintenance (O&M) status units establish data input and model feedback relationships around the digital twin model. Reservoir units, gate units, river cross-section units, pipeline units, pumping station units, and dam units are each connected to the digital twin model as independent O&M status units. Solid arrows between each unit and the digital twin model represent the uploading of operational data, the calling of model parameters, and the feedback of prediction results. Upstream and downstream arrows between units represent the transmission relationship of water level, flow rate, and scheduling response in the direction of water flow. Dashed arrows represent the spatial association relationship between adjacent units. When constructing O&M status units, the system first determines the upstream and downstream associated unit sets based on the water flow direction. Then, it determines the adjacent associated unit sets based on spatial location, dam adjacency relationships, pipeline connection relationships, and facility management boundaries. This allows the digital twin model to simultaneously receive unit's own operational data, upstream and downstream transmitted data, and data on the impact of adjacent facilities.

[0027] Specifically, no. Each maintenance status unit Represented as: in, Indicates the first Each operation and maintenance status unit This indicates the facility number of the operation and maintenance status unit. Indicates the facility type. Indicates geographic coordinates, river station numbers, or levee section locations. Represents the set of upstream related units. Represents the set of downstream related units. Represents a set of adjacent and related units. This represents the digital twin model corresponding to this operation and maintenance status unit. This represents the operation and maintenance dataset belonging to this operation and maintenance status unit.

[0028] Furthermore, a gate-type operation and maintenance status unit establishes a gate overcurrent model, the basic expression of which is: in, Indicates the gate flow rate. Indicates the gate flow coefficient. Indicates the gate width. Indicates the gate opening degree. Represents gravitational acceleration. This represents the head difference between the upstream and downstream sides of the gate. The data sources for the gate flow model include upstream water level gauges, downstream water level gauges, gate opening sensors, gate actuator encoders, gate actuator current acquisition devices, and the gate dispatching command system.

[0029] Furthermore, a water level-discharge response model is established for the river cross-section-type operation and maintenance status unit, and its basic expression is: in, Indicates the flow rate at the river cross-section. Indicates the roughness of the river channel. Indicates the cross-sectional area of ​​the water passage. Indicates the hydraulic radius. This represents the water surface gradient or energy gradient. Data sources for the water level-discharge response model include river level gauges, cross-sectional measurement data, flow velocity meters, historical water level-discharge relationship curves, river topographic measurement data, and hydrological station data.

[0030] Furthermore, the pump station operation and maintenance status unit establishes a pump station operation model, the basic expression of which is: in, Indicates the pump station's outlet flow rate. This indicates the efficiency parameters of the pumping station. Indicates the motor input power. Indicates the density of water. Represents gravitational acceleration. This indicates the pump station head. Data sources for the pump station operation model include the pump station SCADA system, smart meters, motor protectors, inlet tank level gauges, outlet tank level gauges, flow meters, vibration sensors, and temperature sensors.

[0031] Furthermore, a seepage model for dams is established within the dam operation and maintenance status unit, and its basic expression is as follows: in, This indicates the model's predicted seepage pressure or seepage response value. Indicates the permeability coefficient of the dam. Indicates the water level in front of the dam or dike. This indicates the water level on the back side or the drainage reference water level. Indicates the equivalent length of the seepage path. This represents the seepage hysteresis coefficient. This indicates the rate of change of water level in front of the dam or dike. Data sources for the dam seepage model include piezometers, pore water pressure gauges, water level gauges, displacement monitoring devices, crack gauges, rainfall stations, dam design data, and geological survey data.

[0032] Furthermore, a pipeline resistance model is established for the pipeline operation and maintenance status unit, and its basic expression is as follows: in, Indicates the head loss in the pipeline. Indicates the pipe and channel resistance coefficient. Indicates the length of the pipe / channel. This indicates the average flow velocity in the pipes and channels. Indicates the equivalent diameter or hydraulic diameter of the pipe or channel. This represents the acceleration due to gravity. Data sources for the pipeline resistance model include inlet and outlet water level gauges, pipeline flow meters, pipeline design data, dredging records, and pipeline inspection image data.

[0033] S3. Determine the parameters to be corrected and the physical constraint range for the digital twin model. The predictive reliability of the digital twin model depends on whether the parameters to be corrected conform to the actual operating mechanism of the hydraulic facilities. Riverbed siltation leads to changes in riverbed roughness; gate wear or opening deviations lead to changes in gate flow coefficients; impeller wear or pipeline blockages lead to changes in pump station efficiency parameters; alterations to dam seepage channels lead to changes in permeability and seepage hysteresis coefficients; and siltation or local damage to pipes and canals leads to changes in resistance coefficients. Therefore, this invention determines the corresponding parameters to be corrected based on the facility type and sets physical constraint ranges.

[0034] Specifically, the parameters to be corrected constitute the particle parameter vector. For different types of operational state units, the system starts from the particle parameter vector. The parameters corresponding to this unit are selected as the actual optimization objects; the gate unit is selected. and Pump station unit selection and Dam unit selection and River channel cross-sectional unit selection Selection of pipeline units .

[0035] Furthermore, physical constraint intervals satisfy: in, Indicates the physical constraint interval. This represents the lower limit parameter vector for each parameter to be corrected. This represents the upper limit parameter vector for each parameter to be corrected. This indicates the inbound traffic entering the current operation and maintenance status unit. This indicates the outgoing traffic leaving the current operation and maintenance status unit. This indicates the current operation and maintenance status unit at the time step. Changes in water storage within the area This indicates the allowable continuous error in water volume. This represents the physical constraint penalty term for the parameters. This indicates the upper limit of the allowed physical constraint error. and The data sources include water conservancy facility design data, equipment factory parameters, historical calibration parameters, engineering specification control values, historical normal operation data, and the continuity relationship of water flow between adjacent units.

[0036] Furthermore, in the gate unit, The constraints are determined by the gate type, gate width, historical flow rate determination results, and upstream and downstream water level differences; within the river channel cross-sectional unit, The constraints are determined by the riverbank revetment material, riverbed roughness survey results, cross-sectional morphology, and dredging records; within the pumping station unit, The constraints are determined by the unit's design efficiency curve, operating years, maintenance records, and historical current-flow relationships; within the dam unit, and The constraints are determined by geological survey data, dam design documents, historical seepage pressure response, and water level fluctuations; within the pipe-channel unit, The constraints are determined by the pipeline design data, pipe diameter, pipe material, historical dredging records, and the difference between the inlet and outlet water levels.

[0037] S3 includes: S31, optimizing and correcting the parameters to be corrected based on the particle swarm optimization algorithm, so that the predicted output of the digital twin model matches the measured operating data in the operation and maintenance dataset. Each particle represents a set of parameter vectors to be corrected, and the particle swarm updates the particle velocity and position according to the individual optimal position and the global optimal position during the iteration process.

[0038] Specifically, no. The particle in the first The position at the next iteration is represented as Speed ​​is expressed as The particle swarm is updated as follows: in, Indicates the first The particle in the first Speed ​​at the next iteration Indicates inertia weight, Represents individual learning factors. Represents the group learning factor. and Represents a random number between 0 and 1. Indicates the first The particle in the first The optimal position of the individual obtained before the next iteration. Indicates the particle swarm in the th The globally optimal position obtained before the next iteration. Indicates the first The particle in the first The position at the next iteration.

[0039] Furthermore, the measured fitting error satisfy: in, This indicates the number of data types involved in the fitting process. Indicates the first Similar to actual test running data, The output of the digital twin model represents the first... Predictive running data, Indicates the first The allowable fluctuation range of the class's operating data. For gate units, This includes the flow rate through the sluice gate, changes in upstream water level, and changes in downstream water level; for pumping station units, This includes water flow rate, motor current, vibration, and temperature rise; for dam units, This includes the results of image recognition of seepage pressure, displacement, and seepage flow; for pipe and channel units, This includes the difference between inlet and outlet water levels and the flow rate.

[0040] Furthermore, the parametric physical constraint penalty term satisfy: in, Indicates the first One parameter to be corrected, Indicates the first The upper limit value of the parameter to be corrected. Indicates the first The lower limit value of each parameter to be corrected. This indicates a physical association penalty. Used to characterize the degree of inconsistency between parameter combinations and the physical mechanism of a facility, such as pump station efficiency parameters. Increase when there is a mismatch between motor current, head, and outlet flow rate. Gate flow coefficient Increase when there is a mismatch between the gate opening, the difference in water head between upstream and downstream, and the flow rate through the gate. .

[0041] Furthermore, the continuous error in upstream and downstream water volume satisfy: in, This indicates the inbound traffic entering the current operation and maintenance status unit. This indicates the outgoing traffic leaving the current operation and maintenance status unit. This indicates the current operation and maintenance status unit at the time step. Changes in water storage within the area This indicates the reference flow rate. and The data comes from upstream flow meters, downstream flow meters, gate flow models, pump station outlet flow meters, or water level-flow curves. It is calculated based on the current unit water level changes and reservoir capacity curves, river cross-sectional area or pipeline volume relationship.

[0042] S4. During the optimization process of the particle swarm optimization algorithm, calculate the particle swarm diversity index. Physical consistency index of parameters and cross-condition prediction error And based on particle swarm diversity index Physical consistency index of parameters and cross-condition prediction error Identify the risk of premature convergence. The danger of premature convergence in particle swarm optimization lies not in the model's inability to fit the current data, but in the possibility of obtaining a locally optimal solution that fits well at the moment but has incorrect physical parameters. This locally optimal parameter may make the simulated curve fit the measured curve within the current time window, but it will fail when the operating conditions change, such as rainfall, scheduling, pump station start-up and shutdown, or water level fluctuations.

[0043] like Figure 4 As shown, the premature convergence identification curves are plotted with the number of iterations on the x-axis and the values ​​of each identification index on the y-axis. Curve Efit represents the measured fitting error, curve D represents the particle swarm diversity index, and curve Ec represents the cross-condition prediction error. Efit,th, Dth, and Ec,th represent the preset fitting error threshold, preset diversity threshold, and preset cross-condition error threshold, respectively. During normal particle swarm optimization, Efit decreases with the number of iterations, D maintains a sufficient search range, and Ec decreases as the model's generalization ability improves. When iterating to the vicinity of Tec, if Efit is lower than Efit,th, D is lower than Dth, but Ec is still higher than Ec,th, it indicates that although the particle swarm has achieved a smaller current fitting error, the particles have concentrated in a local parameter region, and this parameter combination cannot stably pass the cross-condition verification. Figure 4 The shaded area in the diagram represents the premature convergence risk zone. When the system enters this zone, it triggers perturbation reconstruction and multi-subgroup split search to avoid directly outputting the current local optimum parameters.

[0044] Furthermore, trend consistency error satisfy: in, Indicates trend consistency error. Indicates the first The change in similar measured operational data within a continuous time window. The digital twin model predicts the first... Changes in class runtime data Indicates the first Reference changes in class runtime data.

[0045] The data sources for the validation sets under different operating conditions include historical heavy rainfall data, historical gate scheduling data, pump station start-up and shutdown switching data, rapid water level rise and fall data, low-flow operation data during the dry season, and recently manually confirmed typical operating samples. When the particle swarm diversity index is lower than the preset diversity threshold, the measured fitting error is lower than the preset fitting error threshold, and the parameter physical consistency index is lower than the preset consistency threshold, or the cross-operating condition prediction error is higher than the preset cross-operating condition error threshold, the system determines that the particle swarm has a risk of premature convergence.

[0046] S5. When premature convergence risk is identified, perturbation reconstruction and multi-subgroup split search are performed on the particle swarm to allow the particles to re-enter the parameter search region that satisfies the physical constraints of hydraulic engineering. After identifying the premature convergence risk, if the search continues according to the original particle swarm iteration method, the particles will continue to move around local extrema in a small range, making it difficult to escape the already formed erroneous parameter region. This invention, through perturbation reconstruction and multi-subgroup split search, enables the particle swarm to regain search diversity, while avoiding the loss of already obtained effective search information due to completely random restarts.

[0047] Specifically, the reconstructed particle positions after perturbation satisfy: in, Indicates the first The parameter positions of the reconstructed particles This represents a reference parameter vector selected from elite particles or historical normal parameter distributions. This represents the residual directional perturbation coefficient. This represents the direction vector of the upstream and downstream residuals. This represents the random disturbance coefficient. This represents a random parameter vector generated within the physical constraint interval. This indicates that the parameters are restricted to... and The truncation function between.

[0048] Furthermore, the upstream and downstream residual direction vectors The determination is based on the predicted residuals from the current digital twin model. For the gate unit, when the upstream water level continues to rise while the downstream flow rate is lower than the predicted value... To reduce the gate flow coefficient Or increase the direction of parameters affecting the pre-gate obstruction; for pumping station units, when the motor current increases while the outflow rate is lower than the predicted value, Pointing to reduce pump station efficiency parameters Or increase the equipment attenuation coefficient The parameter direction; for a dam unit, when the water level recedes but the seepage pressure does not decrease. To adjust the permeability coefficient of the dam and seepage hysteresis coefficient The parameter direction; for a pipe and channel unit, when the inlet and outlet water level difference increases while the flow rate decreases... Increasing the pipe and channel resistance coefficient The direction of the parameters.

[0049] Furthermore, the multi-subgroup splitting search involves dividing the reconstructed particle swarm into a state-fitting subgroup, a physically consistent subgroup, and a cross-condition verification subgroup. The state-fitting subgroup is used to reduce... As the primary objective, physically consistent subgroups are used to improve With this as the primary objective, a cross-condition validation subgroup is used to reduce... The primary objective is to exchange locally optimal particles after each subgroup reaches a preset number of iterations, thereby ensuring that the particle swarm maintains its current data fitting ability, the accuracy of physical parameters, and the stability of cross-condition predictions.

[0050] S6. Using the parameter-corrected digital twin model, predict the operating status of each operation and maintenance status unit and its upstream and downstream related units, and generate dynamic risk values ​​based on the prediction results. The digital twin model after premature convergence suppression has a more reliable physical parameter basis and can be used to predict the operating status of different operation and maintenance status units and their upstream and downstream related units.

[0051] Specifically, the system will correct the parameter vector Substituting into the digital twin model, the predicted operating state is obtained. And compare it with the measured operating status within the next time window or a future set time window. Compare the prediction deviation values. satisfy: in, This represents the prediction deviation value. Indicates the first The values ​​of the actual measured running data at the predicted time. This represents the output of the digital twin model at the prediction time. Predictive running data, Indicates the first The allowable fluctuation range of class-based running data.

[0052] Furthermore, the impact value of upstream and downstream linkages satisfy: in, This indicates the impact value related to upstream and downstream connections. Indicates the output flow of the upstream associated unit. This indicates the current operation and maintenance status unit input traffic. This indicates the output traffic of the current operation and maintenance status unit. This indicates the traffic received by the downstream associated unit. This indicates the measured water level upstream. Indicates the predicted water level upstream. This indicates the measured water level downstream. Indicates the predicted downstream water level. Indicates reference flow rate. Indicates the allowable range of water level fluctuations. , , , This represents the corresponding weighting coefficient.

[0053] Furthermore, the trend evolution value satisfy: in, Indicates the trend evolution value. Indicates the prediction deviation value The rate of change over time Indicates the duration of the anomaly. Indicates the direction factor of abnormal fluctuations. , , This represents the corresponding weighting coefficient. When the prediction deviation continues to increase, the duration of the anomaly lengthens, and the direction of the anomaly fluctuations develops in an unfavorable direction, the trend evolution value... Increase.

[0054] Furthermore, historical maintenance is currently underway. satisfy: in, This indicates that the historical maintenance period is in full swing. This indicates the number of times the same type of anomaly occurs within a preset period. Indicates the number of recurrences after treatment. This indicates that the handling of historical issues was efficient. , , This represents the corresponding weighting coefficient. , and The data comes from the operation and maintenance work order system, maintenance records, closed-loop review conclusions, and historical inspection files.

[0055] S7. Generate operation and maintenance tasks and dispatch plans based on dynamic risk values. When the system is in the first risk zone, it generates an observation task; when When the system is in the second risk zone, it generates an inspection task; when When the system is in the third risk zone, it generates a maintenance task; when When the system is in the fourth risk zone, it generates an early warning task or an emergency response task. The risk classification threshold is determined jointly by historical hazard records, facility importance level, management unit's handling procedures, and expert experience.

[0056] Furthermore, order priority value satisfy: in, Indicates the priority value for order dispatch. Indicates dynamic risk value. Indicates the urgency of the task. This indicates the skill matching degree of the operation and maintenance personnel. Indicates the road traffic matching degree. This indicates the current workload of the operations and maintenance personnel. , , , , This represents the corresponding weighting coefficient. The system selects the personnel to perform the tasks based on the dispatch priority value and generates the inspection route.

[0057] S8. Receive on-site handling results and perform closed-loop verification of the handling effect based on sensor data before and after handling, digital twin prediction results, and inspection image data. Maintenance personnel upload on-site handling results via mobile maintenance terminals. On-site handling results include on-site photos, videos, text records, handling time, personnel involved, materials used, repaired parts, explanation of the cause of the anomaly, and retest data. After receiving the on-site handling results, the system automatically retrieves the sensor data before and after handling, digital twin prediction results, and inspection image data for that maintenance status unit and calculates the closed-loop verification value. .

[0058] Furthermore, the degree of recovery of prediction bias satisfy: Recovery level of upstream and downstream linkages satisfy: in, This indicates the predicted deviation value before treatment. This indicates the predicted deviation value after treatment. This indicates the upstream and downstream related impact value before the disposal. This indicates the upstream and downstream related impact values ​​after the action. It also represents the degree of hazard removal in the inspection images. Determined by image recognition results or manual annotation results, the data originates from fixed camera, drone images, or photos from mobile maintenance terminals. The degree of confirmation through manual retesting is also considered. It is determined by the on-site retest data, retest photos, and retest records of the maintenance personnel.

[0059] For gate clearing tasks, closed-loop verification data sources include gate opening degree, hoist current, upstream water level, downstream water level, flow rate, and images before and after the treatment; for pump station maintenance tasks, closed-loop verification data sources include motor current, motor temperature rise, vibration, inlet and outlet water levels, outflow rate, and pump station images before and after the treatment; for dam seepage treatment tasks, closed-loop verification data sources include seepage pressure, displacement, water level, rainfall, wet marks images, and manual inspection records before and after the treatment; for pipeline dredging tasks, closed-loop verification data sources include inlet water level, outlet water level, flow rate, pipeline resistance coefficient, and pipeline inspection images before and after the treatment.

[0060] In Scenario 1, the gate opening data of a river gate operation and maintenance status unit did not reach the fault alarm threshold, but the system detected that the upstream water level continued to rise, the downstream flow rate did not increase synchronously with the gate opening, and the rainfall data continued to increase. The system first used the gate flow model to determine the gate flow coefficient. Perform parameter correction. During particle swarm optimization, a certain combination of parameters... Smaller, but the corresponding parameter combination Deviating from the historical normal range, and under historical heavy rainfall conditions The magnitude of the error is relatively large. Based on this, the system determines that the particle swarm optimization has a risk of premature convergence and performs perturbation reconstruction and multi-subgroup splitting search. After parameter correction, the digital twin model predicts that the gate's flow capacity is lower than normal. The system generates a maintenance task and assigns it to personnel with hoist maintenance capabilities. The maintenance personnel find floating debris accumulated in front of the gate and clear it. The system then calculates the closed-loop verification value based on the gate opening after the clearing, hoist current, upstream water level, downstream flow rate, and images of the area in front of the gate. After confirming that the water level-flow relationship has been restored, the task is archived.

[0061] In Scenario 2, the vibration value of a drainage pumping station did not exceed the fixed alarm threshold, but it showed an upward trend for three consecutive days. The motor current increased slightly, and the motor temperature rose above the historical average under similar operating conditions. The system uses the pumping station operation model to analyze the pumping station efficiency parameters. and equipment attenuation coefficient Corrections should be made. If the particle swarm concentrates too early at a higher concentration... Nearby, although it can fit the current outflow rate, during pump station start-up and shutdown switching conditions... A significant increase in efficiency indicates that the system may be masking impeller wear or pipe blockage. The system performs a multi-subgroup split search, where the physically consistent subgroup corrects the parameter search direction based on the engineering relationship between current, flow rate, and efficiency, and the cross-condition verification subgroup verifies the prediction error based on pump station start-up and shutdown conditions. After correction, the system determines that the pump station efficiency has decreased and generates a maintenance task. After maintenance, the system compares the vibration, current, temperature rise, effluent flow rate, and model prediction deviations before and after the intervention. Once the operational status is confirmed to have recovered, the task is archived.

[0062] In Scenario 3, the seepage pressure at a monitoring section of a dam failed to decrease synchronously after the rainfall ended and the water level receded, resulting in localized wet marks in the inspection images. The system uses a dam seepage model to determine the dam's permeability coefficient. and seepage hysteresis coefficient Correction is performed. If a local optimum parameter is too large... This explains the current seepage pressure hysteresis phenomenon, but this parameter is relevant under historical conditions of rapid water level decline. The system determined that the seepage risk was too high and initiated a perturbation reconstruction. The corrected model predicted an abnormal seepage risk in the embankment section, and the system generated a high-risk inspection task based on the importance level of the embankment facilities. After on-site treatment, the system continued to read seepage pressure, displacement, and image data; if the seepage pressure decreased and the area of ​​the wet marks in the image shrank, the treatment was deemed effective; if the seepage pressure continued to rise or the wet marks expanded, the treatment was deemed ineffective, and a secondary treatment task was generated.

[0063] In Scenario 4, the water level difference between the inlet and outlet of a water conveyance pipeline gradually increases, and the actual flow rate is lower than the predicted value by the digital twin model. The system's resistance coefficient for the pipeline... Perform parameter calibration and, in conjunction with the continuity of upstream and downstream water flow, determine whether there is a risk of local blockage or leakage. If the particle swarm passes through an abnormally enlarged... The system fits the current flow rate decrease, but this parameter is significantly inconsistent with the normal resistance range after historical dredging. The system identifies premature convergence risk and performs perturbation reconstruction. After recalibration, the system determines that the pipeline has a blockage risk and generates an inspection task. On-site inspection reveals localized siltation, which is then cleared. The system performs a closed-loop verification based on the inlet and outlet water level difference before and after clearing, the degree of flow recovery, changes in pipeline resistance coefficient, and pipeline inspection images. Once the water flow capacity is confirmed to have been restored, the task is archived.

[0064] Based on the same inventive concept, this invention also provides a smart operation and maintenance management system for water conservancy projects, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the aforementioned smart operation and maintenance management method for water conservancy projects. The system includes a data acquisition module, a unit construction module, a digital twin modeling module, a parameter correction module, a premature convergence identification module, a convergence suppression module, a correlation risk prediction module, a task generation module, an intelligent dispatch module, a mobile operation and maintenance terminal, a closed-loop verification module, and an archive update module. The data acquisition module is used to collect data on water level, flow rate, gate opening, hoist current, pump station operation, motor current, motor temperature rise, vibration, seepage pressure, displacement, rainfall, dispatch instructions, and inspection image data. The unit construction module is used to construct multiple operation and maintenance status units based on the spatial location, flow direction, and adjacency relationships of the water conservancy facilities. The digital twin modeling module is used to establish water level and flow response models, gate flow models, pump station operation models, dam seepage models, or pipe and canal resistance models for each operation and maintenance status unit. The parameter correction module optimizes and corrects the parameters to be corrected in the digital twin model based on the particle swarm optimization algorithm. The premature convergence identification module calculates particle swarm diversity indices, parameter physical consistency indices, and cross-condition prediction errors, and identifies premature convergence risks. The convergence suppression module performs perturbation reconstruction and multi-subgroup splitting search when premature convergence risks are identified. The associated risk prediction module uses the parameter-corrected digital twin model to predict the operating status of each maintenance unit and its upstream and downstream associated units, and generates dynamic risk values. The task generation module generates observation tasks, inspection tasks, maintenance tasks, early warning tasks, review tasks, or deep maintenance tasks based on the dynamic risk values. The intelligent dispatch module generates dispatch plans and inspection routes based on task location, task urgency, maintenance personnel status, and road conditions. The mobile maintenance terminal receives tasks and uploads on-site handling results. The closed-loop review module reviews the handling effect based on sensor data before and after handling, digital twin prediction results, and inspection image data. The archive update module is used to update historical operation and maintenance records when the disposal is effective, and to trigger a secondary disposal task when the disposal is ineffective.

[0065] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for intelligent operation and maintenance management of water conservancy projects, characterized in that, Includes the following steps: S1. Collect multi-source operation data of water conservancy projects and aggregate them according to facility location, equipment type and collection time to form operation and maintenance dataset; S2. Construct multiple operation and maintenance status units based on the spatial location, water flow direction, and adjacent relationships of water conservancy facilities, and establish digital twin models for each operation and maintenance status unit; S3. Determine the parameters to be corrected and the physical constraint range of the digital twin model, and perform optimization correction on the parameters to be corrected based on the particle swarm optimization algorithm; S4. During the optimization process of the particle swarm optimization algorithm, calculate the particle swarm diversity index, the parameter physical consistency index, and the cross-condition prediction error, and identify the risk of premature convergence based on the above indicators. S5. When the risk of premature convergence is identified, perturbation reconstruction and multi-subgroup split search are performed on the particle swarm to obtain a digital twin model after parameter correction. S6. Use the digital twin model after parameter correction to predict the operating status of each operation and maintenance status unit and its upstream and downstream related units, and generate dynamic risk values ​​based on the prediction results. S7. Generate operation and maintenance tasks and dispatching schemes based on the dynamic risk values; S8. Receive the on-site handling results and perform closed-loop verification based on sensor data before and after handling, digital twin prediction results, and inspection image data. Based on the closed-loop verification results, execute task archiving or secondary handling.

2. The intelligent operation and maintenance management method for water conservancy projects according to claim 1, characterized in that, The multi-source operational data includes water level data, flow rate data, gate opening data, hoist current data, pump station operation data, motor current data, motor temperature rise data, vibration data, seepage pressure data, displacement data, rainfall data, video inspection image data, manual inspection records, dispatching instruction data, and historical maintenance records.

3. The intelligent operation and maintenance management method for water conservancy projects according to claim 1, characterized in that, The operation and maintenance status units are divided according to reservoirs, dams, gates, pumping stations, river cross-sections, water conveyance pipelines, culverts, or drainage facilities; the digital twin models include water level and flow response models, gate flow models, pumping station operation models, dam seepage models, or pipeline resistance models; the parameters to be corrected include river roughness. Gate flow coefficient Pump station efficiency parameters Dam permeability coefficient Pipeline resistance coefficient Equipment attenuation coefficient or seepage hysteresis coefficient The parameters to be corrected constitute the particle parameter vector. The particle parameter vector satisfy: in, Indicates the roughness of the river channel. Indicates the gate flow coefficient. This indicates the efficiency parameters of the pumping station. Indicates the permeability coefficient of the dam. Indicates the pipe and channel resistance coefficient. Indicates the equipment attenuation coefficient. This represents the seepage hysteresis coefficient.

4. The intelligent operation and maintenance management method for water conservancy projects according to claim 1, characterized in that, The fitness function of the particle swarm optimization algorithm From the measured fitting error Parameter physical constraint penalty item and continuous error in upstream and downstream water volume Together, the fitness function satisfy: in, This represents the fitness function value. This represents the measured fitting error between the predicted output of the digital twin model and the actual operating data. This represents the physical constraint penalty term for parameters that deviate from the physical constraint range or physical correlation relationship of the parameter to be corrected. This indicates the continuous error in water volume between the current operation and maintenance status unit and its upstream and downstream related units. , , They represent the measured fitting errors, respectively. Parameter physical constraint penalty item and continuous error in upstream and downstream water volume The weighting coefficients.

5. The intelligent operation and maintenance management method for water conservancy projects according to claim 1, characterized in that, The particle swarm diversity index Based on particle parameter vector With the particle swarm center parameter vector The particle swarm diversity index is obtained by calculating the average distance between them. satisfy: in, Indicators representing particle swarm diversity Indicates the number of particles. Indicates the first The parameter vector of each particle. Represents the particle swarm center parameter vector. The upper limit parameter vector representing the physical constraint interval. The lower bound parameter vector representing the physical constraint interval. Indicates the first The parameter vector of each particle With the particle swarm center parameter vector The distance between them It represents the scale range of the physical constraint interval.

6. The intelligent operation and maintenance management method for water conservancy projects according to claim 1, characterized in that, The physical consistency index of the parameters satisfy: in, This represents the physical consistency index of the parameters. This represents the physical constraint penalty term for the parameters. This indicates continuous error in upstream and downstream water volume. This indicates the trend consistency error; the cross-condition prediction error satisfy: in, This indicates the prediction error across operating conditions. Indicates the number of test case verification sets. Indicates the first The first working condition verification set Similar to actual test running data, The digital twin model represents the first The first output of the working condition verification set Predictive running data, Indicates the first The allowable fluctuation range of the class of operational data; the operational condition verification set includes the rainfall operational condition verification set, the gate scheduling operational condition verification set, the pump station start-up and shutdown operational condition verification set, or the water level rise and fall operational condition verification set.

7. The intelligent operation and maintenance management method for water conservancy projects according to claim 1, characterized in that, The particle swarm is considered to have a risk of premature convergence when the following criteria are met: in, Indicators representing particle swarm diversity This indicates a preset diversity threshold. This represents the measured fitting error. This indicates the preset fitting error threshold. This represents the physical consistency index of the parameters. This indicates a preset consistency threshold. This indicates the prediction error across operating conditions. This indicates the preset cross-operating condition error threshold.

8. The intelligent operation and maintenance management method for water conservancy projects according to claim 1, characterized in that, The dynamic risk value Based on the predicted deviation value Impact value of upstream and downstream linkages Trend evolution value Importance level of facilities and historical maintenance is in full swing The calculated dynamic risk value satisfy: in, Indicates dynamic risk value. This represents the prediction deviation value. This indicates the impact value related to upstream and downstream connections. Indicates the trend evolution value. Indicates the importance level of the facility. This indicates that the historical maintenance period is in full swing. , , , , These represent the prediction deviation values. Impact value of upstream and downstream linkages Trend evolution value Importance level of facilities and historical maintenance is in full swing The weighting coefficients.

9. The intelligent operation and maintenance management method for water conservancy projects according to claim 1, characterized in that, The closed-loop verification includes calculating the closed-loop verification value. The closed-loop verification value satisfy: in, This represents the closed-loop verification value. Indicates the degree of recovery from prediction bias. Indicates the degree of recovery of upstream and downstream connections. Indicates the degree of hazard removal in the inspection images. Indicates the degree of confirmation by manual retesting. , , , These represent the degree of recovery from prediction bias. Degree of recovery of upstream and downstream connections , degree of hazard removal in inspection images Confirmation level by manual retesting The weighting coefficients; when ,and , When the closed-loop review results are deemed to meet the defect elimination conditions, then... This represents the closed-loop review threshold. This indicates the predicted deviation value after treatment. This indicates the allowable prediction bias threshold. This indicates the impact value on upstream and downstream relationships after the treatment. This indicates the threshold that allows upstream and downstream connections to influence the data.

10. A smart operation and maintenance management system for water conservancy projects, characterized in that, It includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement a smart operation and maintenance management method for water conservancy projects according to any one of claims 1-9.

Images