Real-time Early Warning System for Water Conservancy Risks with Multimodal Spatiotemporal Fusion

Abstract

The present invention relates to the technical field of water conservancy project risk early warning, specifically a real-time early warning system for water conservancy risks with multi-modal spatio-temporal fusion. The system includes: a multi-modal spatio-temporal data fusion module, which performs spatio-temporal alignment and feature extraction calculations based on the sensor data streams of reservoir groups, remote sensing images, and weather forecasts to obtain the fused spatio-temporal situation values. In the present invention, by fusing the data of the sensor data streams of reservoir groups, remote sensing images, and weather forecasts, spatio-temporal alignment and feature extraction calculations are performed to construct the fused spatio-temporal situation values, improving the collaborative analysis accuracy of multi-source heterogeneous data and the situation awareness ability, enhancing the data utilization rate and the accuracy of risk analysis; when establishing the potential scheduling operation set, the gate opening and flood discharge flow adjustment options under different time windows are defined to strengthen the flexibility and adaptability of the scheduling plan, making the scheduling operation more in line with the actual hydrological condition changes.

Description

Technical Field

[0001] The present invention relates to the technical field of water conservancy project risk early warning, and particularly to a real-time early warning system for water conservancy risks with multimodal spatio-temporal fusion. Background Art

[0002] The technical field of water conservancy project risk early warning mainly involves a series of theories and methods for pre-monitoring, identifying and analyzing various possible risk events in water conservancy project facilities and related basin systems, and predicting the development trend and severity of risks to issue warning information in advance.

[0003] The existing technologies have problems of single data and static analysis mode when dealing with water conservancy project risks. In the actual operation process, the risk assessment of water conservancy projects usually only relies on the data monitored by local sensors, lacking in-depth correlation analysis among multi-source data, and it is difficult to capture the complex interaction relationship between hydrological conditions and the operation state of engineering facilities. At the same time, the dynamic changes of the risk situation under real-time conditions are ignored, resulting in lag and deviation in risk early warning. For example, it is difficult to timely identify the sudden risks brought by dam displacement or sharp increase in incoming flow, thus delaying the best opportunity for emergency response. Therefore, improvements are needed. Summary of the Invention

[0004] The purpose of the present invention is to solve the deficiencies existing in the prior art, and a real-time early warning system for water conservancy risks with multimodal spatio-temporal fusion is proposed.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions: The real-time early warning system for water conservancy risks with multimodal spatio-temporal fusion includes:

[0006] A multimodal spatio-temporal data fusion module, which performs spatio-temporal alignment and feature extraction calculations based on the sensor data streams of reservoir groups, remote sensing images, and weather forecasts, obtains a fused spatio-temporal situation value, defines the gate opening and flood discharge flow adjustment options under different windows based on the fused spatio-temporal situation value and preset scheduling rules, and establishes a set of potential scheduling operations;

[0007] A water conservancy risk quantitative assessment module, which correlates the water level and incoming flow to calculate the hydrodynamic response based on the set of potential scheduling operations, calculates the influence degree of each operation option, obtains the scenario hydraulic influence degree, evaluates the dam-break probability index based on the scenario hydraulic influence degree, and constructs a quantitative risk index library;

[0008] A resource and risk budget allocation module, which matches the water supply guarantee rate and power generation benefit index to calculate the balanced performance of each scenario based on the quantitative risk index library, obtains a risk-benefit balance index, screens and ranks the scenarios based on the risk-benefit balance index, and generates a risk constraint efficiency ranking list;

[0009] The dispatching plan early warning release module selects scenarios based on the risk-constrained efficiency ranking list according to the priority, and generates real-time risk warning signals according to the dam-break probability index of the scenarios.

[0010] Preferably, the steps for obtaining the fused spatio-temporal situation value are as follows:

[0011] Based on the water level height, the inflow velocity, and the flood discharge rate, perform linear interpolation and outlier removal processing respectively according to the time stamp, align the water level height, the inflow velocity, and the flood discharge rate to the same time series, combine the surface reflectance map and the terrain slope map in the same time window of the remote sensing image, and synchronize the precipitation intensity grid sequence within the next three hours in the weather forecast to generate the sensor data stream of the reservoir group, the remote sensing image map, and the weather forecast grid sequence after time alignment;

[0012] According to the sensor data stream of the reservoir group, the remote sensing image map, and the weather forecast grid sequence after time alignment, map the water level height, the inflow velocity, the flood discharge rate, and the precipitation intensity at each moment to the three-dimensional geographic grid according to the regional grid number, and generate a multi-modal spatio-temporal feature vector set under the geographic grid in combination with the surface reflectance and the terrain slope at the corresponding position;

[0013] Based on the multi-modal spatio-temporal feature vector set under the geographic grid, calculate the fused spatio-temporal situation value.

[0014] Preferably, the steps for obtaining the set of potential dispatching operations are as follows:

[0015] Based on the fused spatio-temporal situation value, call the gate water level threshold, the flood discharge warning line flow rate, and the safety water level interval in the preset dispatching rules, perform parameter threshold comparison item by item, map the fused spatio-temporal situation value to the dispatching conditions in sequence, and generate a set of spatio-temporal situation matching parameters that meet the gate dispatching conditions;

[0016] Based on the set of spatio-temporal situation matching parameters that meet the gate dispatching conditions, call the gate opening adjustment level and the flood discharge flow rate adjustment level defined for different time windows in the preset dispatching rules, perform gate opening level selection item by item, and generate a set of potential dispatching operations.

[0017] Preferably, the steps for obtaining the hydraulic influence degree of the scenario are as follows:

[0018] Based on the set of potential dispatching operations, call the gate opening change amount, the flood discharge flow rate change amount, the gate opening duration, the water level monitoring sequence, the inflow discharge monitoring sequence, and the operation trigger time in each dispatching operation, establish a parameter correspondence table according to the dispatching operation number respectively, and generate a set of water level and inflow discharge dispatching response combinations;

[0019] According to the set of water level and reservoir inflow scheduling response combinations, conduct hydrodynamic simulations item by item to obtain the maximum water level rise amplitude, instantaneous slope of reservoir inflow, duration of flow disturbance, flood discharge response delay, and water level lagging fall time corresponding to each scheduling operation, and generate a set of hydrodynamic response parameters for scheduling operations;

[0020] Based on the set of hydrodynamic response parameters for scheduling operations, calculate the scenario hydraulic impact degree, and the calculation formula is:

[0021] ;

[0022] wherein, is the scenario hydraulic impact degree of the th scheduling operation, is the maximum water level rise amplitude corresponding to the th scheduling operation, is the instantaneous slope of reservoir inflow corresponding to the th scheduling operation, is the duration of flow disturbance corresponding to the th scheduling operation, is the flood discharge response delay corresponding to the th scheduling operation, is the water level lagging fall time corresponding to the th scheduling operation, is the starting time of scheduling corresponding to the th scheduling operation.

[0023] Preferably, the steps for obtaining the quantitative risk index library are as follows:

[0024] Based on the scenario hydraulic impact degree, extract the scenario hydraulic impact degree values corresponding to the numbers of each scheduling operation, and simultaneously retrieve the real-time displacement of the dam body, the rising height of the phreatic line around the dam body, the peak downstream flow velocity, the rising amplitude of the seepage pressure at the dam foundation, and the growth of the structural crack length corresponding to the numbers, and uniformly organize the corresponding relationship of the number index to generate a set of basic risk parameters for scheduling operations;

[0025] According to the set of basic risk parameters for scheduling operations, calculate the dam-break probability index corresponding to the scheduling operation, and the calculation formula is:

[0026] ;

[0027] wherein, is the dam-break probability index of the th scheduling operation, is the real-time displacement of the dam body corresponding to the th scheduling operation, is the rising height of the phreatic line around the dam body corresponding to the th scheduling operation, is the The peak downstream flow velocity corresponding to the th rise in seepage pressure at the dam foundation corresponding to the th increase in the length of structural cracks corresponding to the th scenario hydraulic impact degree of the

[0028] Based on the dam-break probability index, each scheduling operation is bound one by one with the dam-break probability index according to the operation number, and an index structure is established in the order of the numbers to generate a quantitative risk index library.

[0029] Preferably, the steps for obtaining the risk-benefit balance index are as follows:

[0030] Based on the quantitative risk index library, extract the dam-break probability index corresponding to each scheduling operation number Synchronously retrieve the regional water supply achievement, the water supply plan value in the corresponding time period, the power generation output value, the installed capacity, and the power generation duration recorded in the scheduling operation to generate a water supply and power generation basic parameter set;

[0031] Based on the water supply and power generation basic parameter set, calculate the risk-benefit balance index corresponding to the scheduling operation. The calculation formula is:

[0032] ;

[0033] where is the risk-benefit balance index corresponding to the th scheduling operation, is the water supply guarantee rate corresponding to the th scheduling operation, is the power generation output value corresponding to the th scheduling operation, is the dam-break probability index corresponding to the

[0034] Preferably, the steps for obtaining the risk constraint efficiency ranking list are as follows:

[0035] Based on the risk-benefit balance index of each scheduling operation, using the risk-benefit balance index as the sorting basis, compare the numerical values item by item and perform a descending order sorting to form a scenario risk-benefit sorting sequence;

[0036] Based on the scenario risk-benefit ranking sequence, call the corresponding scenario numbers in the ranking order, and perform item-by-item index matching for the serial numbers. Using the risk ranking order as the index, establish a risk constraint effectiveness ranking list.

[0037] Preferably, the step of obtaining the real-time risk warning signal is as follows:

[0038] Call the scheduling operation number with the earliest ranking number in the risk constraint effectiveness ranking list, extract information item by item, and generate the scenario plan with the highest priority.

[0039] According to the scenario plan with the highest priority, obtain the numerical value of the dam-break probability index, determine the real-time risk level threshold interval into which the numerical value of the dam-break probability index falls, and obtain the real-time risk warning signal.

[0040] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0041] In the present invention, by integrating the data streams of reservoir group sensors, remote sensing images, and meteorological forecasts, performing spatio-temporal alignment and feature extraction calculations, and constructing a fused spatio-temporal situation value, the collaborative analysis accuracy of multi-source heterogeneous data and the situation awareness ability are improved, and the data utilization rate and the accuracy of risk analysis are enhanced; when establishing the potential scheduling operation set, define the gate opening and flood discharge flow adjustment options under different time windows, strengthen the flexibility and adaptability of the scheduling plan, and make the scheduling operation more in line with the actual changes in hydrological conditions; for the calculation of the scenario hydraulic influence degree and the dam-break probability index and the determination of the risk-benefit balance index, realize a refined and quantitative risk assessment process, reduce the subjectivity of risk judgment, and improve the decision-making reliability; in the generation link of the real-time risk warning signal, through the comprehensive evaluation of the risk ranking sequence and the real-time risk level, match the risk warning signal, improve the timeliness and accuracy of risk warning, and enhance the initiative of emergency response. Brief Description of the Drawings

[0042] Figure 1 is the system flow chart of the present invention. Detailed Embodiment

[0043] In order to make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0044] Please refer to Figure 1 , the present invention provides a technical solution: a multi-modal spatio-temporal fusion real-time water conservancy risk warning system includes:

[0045] The multi-modal spatio-temporal data fusion module performs spatio-temporal alignment and feature extraction calculations based on the sensor data streams of the reservoir group, remote sensing images, and weather forecasts to obtain the fused spatio-temporal situation value. Based on the fused spatio-temporal situation value and the preset scheduling rules, it defines the gate opening and flood discharge flow adjustment options under different windows and establishes a potential scheduling operation set;

[0046] The water conservancy risk quantification and assessment module calculates the influence degree of each operation option based on the potential scheduling operation set, associates the water level and the incoming flow to calculate the hydrodynamic response, and obtains the scenario hydraulic influence degree. Based on the scenario hydraulic influence degree, it evaluates the dam break probability index and constructs a quantitative risk index library;

[0047] The resource and risk budget allocation module calculates the balanced performance of each scenario by matching the water supply guarantee rate and the power generation benefit index based on the quantitative risk index library, obtains the risk-benefit balance index. Based on the risk-benefit balance index, it screens and sorts the scenarios to generate a risk constraint efficiency ranking list;

[0048] The scheduling plan early warning and release module selects scenarios according to the priority based on the risk constraint efficiency ranking list and generates real-time risk warning signals according to the dam break probability index of the scenarios.

[0049] The steps for obtaining the fused spatio-temporal situation value are as follows:

[0050] Based on the water level height, incoming flow velocity, and flood discharge rate, linear interpolation and outlier removal are performed respectively according to the time stamps, and the water level height, incoming flow velocity, and flood discharge rate are aligned to the same time series. Combining the surface reflectance map and terrain slope map in the same time window of the remote sensing image, and synchronizing the precipitation intensity grid sequence within the next three hours in the weather forecast, a time-aligned sensor data stream of the reservoir group, remote sensing image map, and weather forecast grid sequence are generated;

[0051] According to the time-aligned sensor data stream of the reservoir group, remote sensing image map, and weather forecast grid sequence, the water level height, incoming flow velocity, flood discharge rate, and precipitation intensity at each moment are mapped to a three-dimensional geographical grid according to the regional grid number, and a multi-modal spatio-temporal feature vector set under the geographical grid is generated by combining the corresponding surface reflectance and terrain slope;

[0052] Based on the multi-modal spatio-temporal feature vector set under the geographical grid, the fused spatio-temporal situation value is calculated, and the calculation formula is:

[0053] ;

[0054] Among them, is the fused spatio-temporal situation value, is the water level height under the unified time series, is the incoming flow velocity under the unified time series, To unify the flood discharge rate under a unified time series, is the surface reflectivity value in the grid of the corresponding area, is the terrain slope value in the grid of the corresponding area, is the precipitation intensity value within three hours in the grid of the corresponding area.

[0055] Specifically, based on the data sources of water level height, inflow velocity, and flood discharge rate, first read the records of water level sensors and velocity sensors within the specified monitoring range, perform sequential comparison through timestamps and establish corresponding indexes. Immediately afterwards, compare whether the water level values of each record in the monitoring data are within the range of 0 meters to 60 meters, whether the inflow velocity values are within the range of 0 meters per second to 3 meters per second, and whether the flood discharge rate values are within the range of 0 meters per second to 4 meters per second. Any records exceeding this range are marked and analyzed separately. During the analysis process, for records exceeding the range, linear interpolation is used with the data of two adjacent legal moments before and after to fill the exceeded part or remove obviously abnormal records. Then, organize the data sequence after interpolation and removal, and then sort it in ascending order according to the monitoring timestamps. Align all water level data, inflow velocity data, and flood discharge rate data on the same time axis. If duplicate data entries with the same timestamp are found during the alignment process, only keep the first one and remove the subsequent duplicate entries. Then, use the surface reflectivity map and terrain slope map within the same time window given by the remote sensing image acquisition part, perform index matching between these maps and the aligned sensor time series, and check the corresponding time periods and geographical locations one by one through grid numbers. Subsequently, load the meteorological forecast grid data of the precipitation intensity within the next three hours and establish a mapping relationship with the previously constructed time series and map data. Each time period has its forecast precipitation intensity value for use. Check whether there are vacant positions among them. If there are no vacant positions, continue to associate the precipitation intensity. If there are vacant positions, supplement the data by referring to the adjacent moment values and comparing the rainfall distribution curve of the monitoring period. Finally, merge all the above multi-source data into a complete and continuous record structure. After summarization, a time-aligned sensor data stream of the reservoir group, remote sensing image maps, and meteorological forecast grid sequence can be obtained.

[0056] According to the sensor data stream of the reservoir group, the remote sensing image atlas, and the meteorological forecast grid sequence after time alignment, first determine the timestamp corresponding to each record and compare it with the pre-numbered set of regional grids. Compare the geographical coordinates of the water level monitoring points with the pixel coordinates of the grids in the remote sensing image. If the positioning deviation is less than 1 pixel range, it is determined to be the same regional grid. If the positioning deviation is greater than 1 pixel range, the record and the corresponding grid are temporarily placed in the pending state and the coordinate mapping relationship is verified one by one. At the same time, check the incoming flow rate and flood discharge rate parameters one by one. Compare again through the collected flow rate range (0 m / s to 3 m / s) and flood discharge rate range (0 m / s to 4 m / s). If it is found that there are data points outside the above range, trace back to the original sensor readings and match the average value of the flow rate or flood discharge rate at the adjacent moment. Then, load the precipitation intensity value within three hours into the same grid number, and map the surface reflectivity and terrain slope to the same three-dimensional geographical grid. Each mapped record contains fields such as timestamp, water level height, incoming flow rate, flood discharge rate, precipitation intensity, surface reflectivity, and terrain slope. Subsequently, assemble these fields to form a multi-modal spatio-temporal feature vector set, ensuring that each field is complete and in one-to-one correspondence with the grid number at each timestamp. After all data combinations, generate a multi-modal spatio-temporal feature vector set under the geographical grid.

[0057] Formula: , The advantage of the formula is that it can incorporate multiple spatio-temporal features of different dimensions in the same expression, including water level height, incoming flow rate, flood discharge rate, surface reflectivity, terrain slope, and precipitation intensity in the next three hours. By comprehensively calculating these parameters in a superimposed and non-linear combination manner, the dynamic situation of the reservoir can be characterized in multiple aspects and a more refined spatio-temporal risk level can be presented, thereby providing a more comprehensive quantitative reference for real-time early warning or scheduling in subsequent steps.

[0058] The acquisition steps of are as follows: This parameter represents the water level height under the unified time series and is directly read from the data collected by the water level sensor. During the data collection process, the water level value is captured every 10 minutes and compared with the backup float measurement information at the same moment. If the difference exceeds 0.3 meters, it is checked and the sensor status is re-compared. The obtained valid water level height is merged into the main record and assigned to .

[0059] The acquisition steps of are as follows: This parameter represents the incoming flow rate under the unified time series and is obtained from the flow rate monitoring equipment installed at the entrance of the reservoir area. The collection interval is 10 minutes each time. First, compare the value with the upper limit of the flow rate threshold of 3 m / s. If it exceeds 3 m / s, check each record one by one and refer to the average flow rate at the adjacent moment to determine whether to correct the incoming flow rate. After correction, assign the final flow rate data to , for example, during a continuous monitoring period, the inflow velocity of 1.2 m / s can be obtained according to the records of the flow velocity monitoring device. After verification, it is confirmed as 1.2 m / s.

[0060] The acquisition steps of are as follows: This parameter represents the flood discharge rate under a unified time series, which is obtained from the on-line flow velocity monitor at the flood discharge control pipeline. The monitor outputs real-time values every 10 minutes and compares them with the flood discharge warning threshold of 2 m / s. If it exceeds 2 m / s, the average value of the adjacent two monitoring period values is read to verify whether there is a sudden anomaly, so as to finally obtain an accurate , for example, in a certain record, the monitored value within 10 minutes is 1.0 m / s. After comparison with the previous and subsequent periods, there is no anomaly, so it is determined that the value at this moment is equal to 1.0 m / s.

[0061] The acquisition steps of are as follows: This parameter represents the surface reflectance value in the corresponding regional grid, which can be obtained by calibrating the brightness value of the remote sensing image pixels of the remote sensing image and performing atmospheric correction. First, use the calibration file of the sensor to read the gain and offset , and substitute the brightness value of each pixel of the remote sensing image into the following calibration formula: , for example, in a certain image, the gain of the corresponding band is , the offset is , when the digital number of a certain pixel is , the result of the operation is: . Subsequently, atmospheric correction is performed. For the effects of aerosol scattering and water vapor absorption, an atmospheric correction coefficient is set. Here, is taken as , then the reflectance after deducting the atmospheric effect can be expressed as: . In this example, , so it is recorded as the surface reflectance in the field of the corresponding regional grid.

[0062] The acquisition steps of are as follows: This parameter represents the terrain slope value in the corresponding regional grid, which is calculated by combining the elevation data in the remote sensing image with the digital terrain model provided by the surveying and mapping department. The specific method is to collect the elevation difference and the ratio of the horizontal distance to obtain the local slope, and then normalize the slopes of all grid points. The normalization formula can be written as , where is the set of slopes of all grid points, is the original slope value of a certain grid, is the normalized slope value. Finally, the normalized slope is assigned to , for example, if the initial slope obtained in a certain grid is 0.15 and it remains 0.15 after normalization, then it is recorded as 0.15.

[0063] The acquisition steps of are as follows: This parameter represents the precipitation intensity value within three hours in the corresponding regional grid, which is extracted from the meteorological forecast grid sequence and calculated cumulatively based on the hourly precipitation data provided by the meteorological department. For example, the upper limit of the precipitation intensity within three hours can reach 100 millimeters. At this time, it is necessary to first add up the hourly precipitation values and then compare with the pre-set alarm value of 80 millimeters. If it exceeds 80 millimeters, it is necessary to cross-check the meteorological radar map and ensure the accuracy of the value. Then, the total amount for three hours is merged and recorded in the of the corresponding grid. For example, if the total precipitation for three hours in a certain grid is statistically 12 millimeters, after verification, it is directly assigned to

[0064] Calculation process:

[0065] First, calculate , where , , to obtain , , and its absolute value is still 0.16551;

[0066] Substitute into , and calculate . Add 0.16551 to get 2772.45551, and then take the square root, which is approximately 52.665;

[0067] Substitute the denominator into , so , , then take the cube root, which is approximately 1.2599. Divide the numerator 52.665 by 1.2599, which is approximately 41.81;

[0068] Another part in , , , first calculate , and then substitute the denominator , and the result is . Add the two parts together: 41.81 + 10.0696 = 51.8796, and finally square the whole to get .

[0069] The result shows that when is approximately 2690.36, it represents a relatively high combined value of the multi-modal spatio-temporal situation under the current moment and spatial grid. Combining information such as water level, flow velocity, and precipitation, certain risk signs can be detected in the reservoir area, and more detailed scheduling or detection is required in the subsequent process.

[0070] The steps to obtain the potential scheduling operation set are as follows:

[0071] Based on the fused spatio-temporal situation value, call the gate water level threshold, flood discharge warning line flow, and safety water level interval in the preset scheduling rules, perform parameter threshold comparison item by item, and map the fused spatio-temporal situation value to the scheduling conditions in turn to generate a set of spatio-temporal situation matching parameters that meet the gate scheduling conditions;

[0072] Based on the set of spatio-temporal situation matching parameters that meet the gate scheduling conditions, call the gate opening adjustment level and flood discharge flow adjustment level defined for different time windows in the preset scheduling rules, perform gate opening level selection item by item, and generate a potential scheduling operation set.

[0073] Specifically, based on the fused spatio-temporal situation value, specific values of the gate water level threshold, flood discharge warning line flow rate, and safe water level range are first disassembled and extracted from the annual reservoir operation records. The setting of the gate water level threshold is determined by referring to the highest water level curve of the reservoir in the past decade and combining the structural bearing limit provided by the design party. For example, the gate water level threshold is set at 55 meters because it is statistically obtained from historical data that most spillways will experience overload when the water level exceeds 55 meters. The flood discharge warning line flow rate is comprehensively obtained by querying the maximum bearing capacity of the downstream area during previous gate openings and combining the flow monitoring data of the plain river channel. For example, this warning line is set at 500 cubic meters per second because multiple monitoring data show that when the measured downstream flow rate exceeds 500 cubic meters per second, the river channel is prone to signs of levee overtopping. The safe water level range is set between 30 meters and 50 meters in combination with the structural characteristics of the reservoir and the downstream incoming water situation to identify a relatively stable water level range. After obtaining these reference values, the water level information and possible flow rate changes included in the currently obtained fused spatio-temporal situation value are compared one by one. If it is found that the current water level exceeds the pre-set gate water level threshold, the record is marked and the flood discharge warning line flow rate is compared to see if it also exceeds the specified standard. If both are in an over-limit state, it is determined that the water level and flow rate conditions represented by the current situation value do not meet the safety requirements. Subsequently, it is continued to check whether the water level falls within the safe water level range or is lower than the lower limit value of 30 meters of this range. If it is detected that the water level continues to be lower than 30 meters, it indicates that the water level in the reservoir area is not sufficient to trigger the flood discharge condition. At this time, a note will be retained in the record and this piece of information will be set to a lower priority. Finally, all time nodes and corresponding flow rate records that meet the gate scheduling conditions are classified and summarized, and are associated and matched with grid number indexes, so as to generate a set of spatio-temporal situation matching parameters that meet the gate scheduling conditions.

[0074] Based on the set of spatio-temporal situation matching parameters that meet the gate scheduling conditions, different time windows are first segmented. The window division refers to the flow rate and water level changes in different time periods sorted according to historical scheduling experience. For example, a smaller three-hour window is set during the night because the inflow velocity often stabilizes at about 1 meter per second at this time, while a shorter two-hour window is used during the daytime to fully view sudden increases in flow. Then, matching parameters are retrieved within each segmented time period, and compared with the pre-defined gate opening adjustment level table and flood discharge flow adjustment level table. The settings of these two adjustment levels refer to the safety margins of the reservoir outflow capacity and the mechanical performance of the gates. For example, the gate opening is divided into ten levels from 0% to 100% and corresponds to the coded values of the scheduling opening from 0 degrees to 10 degrees respectively. The flood discharge flow is divided into several levels from 0 cubic meters per second to a maximum of 1000 cubic meters per second to correspond to different gate opening output intensities. Each data record in each time period will be compared one by one with these two tables to obtain the appropriate gate opening level and the corresponding flood discharge flow level. If the water level is higher than the aforementioned gate water level threshold of 55 meters and the flow approaches the warning line of 500 cubic meters per second within this time period, a larger opening level and a relatively higher flood discharge flow level will be selected. If the current record shows that the water level is slightly higher than the lower limit of 30 meters and the flow stabilizes around 200 cubic meters per second, a medium opening and the corresponding flow can be selected. In this way, after traversing all time periods in turn, the gate opening level selection can be carried out item by item and the scheduling records of each time period can be summarized. Finally, a potential scheduling operation set can be integrated from all the compared time periods.

[0075] The steps for obtaining the scenario hydraulic impact degree are as follows:

[0076] Based on the potential scheduling operation set, call the gate opening change amount, flood discharge flow change amount, gate opening duration, water level monitoring sequence, inflow flow monitoring sequence, and operation trigger time in each scheduling operation, and establish a parameter correspondence table according to the scheduling operation number respectively to generate a water level and inflow flow scheduling response combination set;

[0077] According to the water level and inflow flow scheduling response combination set, conduct hydrodynamic simulations item by item to obtain the maximum water level rise amplitude, instantaneous slope of the inflow flow, flow disturbance duration, flood discharge response delay, and water level hysteresis fall time corresponding to each scheduling operation, and generate a scheduling operation hydrodynamic response parameter set;

[0078] Based on the scheduling operation hydrodynamic response parameter set, calculate the scenario hydraulic impact degree. The calculation formula is:

[0079] ;

[0080] Among them, is the scenario hydraulic impact degree of the th scheduling operation, is the maximum rising amplitude of the water level corresponding to the th scheduling operation, is the instantaneous slope of the incoming flow corresponding to the th scheduling operation, is the duration of the flow disturbance corresponding to the th scheduling operation, is the flood discharge response delay corresponding to the th scheduling operation, is the time for the water level to lag and fall back corresponding to the th scheduling operation, is the starting time of the scheduling corresponding to the th scheduling operation.

[0081] Specifically, based on the potential scheduling operation set, first retrieve parameter contents such as the change amount of the gate opening, the change amount of the flood discharge flow, the duration of the gate opening, the water level monitoring sequence, the incoming flow monitoring sequence, and the operation trigger time item by item from a scheduling operation list. When disassembling and extracting these parameters, first read the change amount of the gate opening, and clarify through consulting the mechanical characteristic records of the gate that the numerical range of the change amount of the gate opening is between 0% and 100%. If it is found that the change amount of the gate opening in this record exceeds 100%, mark it. Subsequently, establish a corresponding relationship between the change amount of the flood discharge flow and the change amount of the gate opening, and compare the upper and lower limits of the flood discharge flow. If it exceeds the preset upper limit value of 1000 cubic meters per second, check whether there are duplicate registrations or abnormal registrations in the record. After determining that the value is available, then read the duration of the gate opening and compare it with the water level data in the same time period in the water level monitoring sequence. For the water level monitoring sequence, it is required that its value is within the range of 0 meters to 60 meters. If there are water level records not in this range, conduct a check. After the comparison is completed, include the incoming flow monitoring sequence in the comparison range. Its range is usually controlled between 0 cubic meters per second and 500 cubic meters per second. Record whether the incoming flow exceeds 500 cubic meters per second during the observation period. If it is found that there is such a situation, note it in the record. Finally, read the operation trigger time and perform time series alignment in combination with the relevant parameter information listed previously. The alignment method uses the operation trigger time as the index to match the corresponding time periods of the change amount of the gate opening, the change amount of the flood discharge flow, the water level monitoring sequence, and the incoming flow monitoring sequence. Check whether each record after the match is coherent and complete. If a parameter is missing, the next step of data cannot be generated. If all are available, then splice and assemble these related records, summarize them separately by scheduling operation number in the time series dimension and the parameter dimension, sort out and uniformly store the corresponding tables of each scheduling operation that has completed alignment and has content integrity. Through this method, a set of water level and incoming flow scheduling response combinations can be obtained.

[0082] According to the combined response set of water level and inflow flow scheduling, the judgment basis of the maximum water level rise and the instantaneous slope of inflow flow is loaded one by one from a compiled hydraulic simulation reference standard, and the water level increasing rate in the monitoring data is compared to see whether it is in the range of 0 meters to 1 meter per hour. If the water level rising rate is greater than 1 meter per hour in any period of time in the data, the period is marked and the subsequent records are tracked and checked. Combined with the change of the instantaneous slope of the flow, it is compared to see whether it exceeds the reference value of 0.2 cubic meters per second per minute. If it is detected that it exceeds this value, the period is recorded as the disturbance moment, and then the flow disturbance duration is calculated, that is, after any record is marked as the disturbance moment Continue to check backward in sequence until the time difference when the flow rate drops back to the original stable range of 0 cubic meters per second to 500 cubic meters per second is found. Similarly, the flood discharge response delay can be determined based on the difference between the gate opening time and the flow increase time measured at the downstream monitoring section. Compare whether this delay meets the parameter range of 0 minutes to 30 minutes set in the previous safety assessment. If a delay that is too large is found, mark it again. The water level lag drop time is obtained by tracking the time it takes for the water level to drop from the highest point after the gate is closed to the stable water level before opening the gate. These five indicators are organized in a unified manner and combined with each scheduling operation number to obtain the scheduling operation hydrodynamic response parameter set.

[0083] formula:

[0084] The benefit of the formula is that it can integrate multiple factors involved in the scheduling operation, such as the maximum water level rise, the instantaneous slope of the inflow, the duration of the flow disturbance, the flood discharge response delay, the water level lag time and the scheduling start time, into a unified measurement scale. It can amplify or attenuate the impact of the scheduling process through nonlinear superposition and logarithmic operations, and reflect the comprehensive balance of various key variables as a whole. This makes it more targeted and complete when evaluating the impact of the scheduling operation on the actual hydraulic conditions.

[0085] The steps to obtain are: This parameter represents the The maximum rise in water level corresponding to the scheduling operation is obtained by comparing the water level monitoring sequence before and after the scheduling operation. After each operation is completed, it is determined by retrieving the peak water level in the measured data and subtracting the corresponding water level before the operation starts. Its range is between 0 meters and 5 meters. If it exceeds 5 meters, additional verification is performed in the record. In the specific measurement process, the water level record before the scheduling is first archived, and then the highest water point after the operation is recorded. Finally, the difference between the two is made. If the actual measurement error exceeds 0.05 meters, it is corrected after comparison with the adjacent time period and finally recorded. For example, in a dispatching operation, the water level was 49.2 meters before the operation began, and the highest water level monitored after the operation was about 52.3 meters. =(52.3 - 49.2)=3.1 m.

[0086] The acquisition steps are as follows: This parameter represents the instantaneous slope of the inflow rate and is specifically used to depict the steepness of the change in the inflow rate at the current operation. Its acquisition consists of the inflow rate records within 30 minutes before and after the operation trigger moment. First, calculate the maximum increment and the corresponding time difference of this 30 - minute monitoring sequence, and then divide the two to obtain the flow rate change rate value, which usually ranges from 0 cubic meters per second per minute to 0.5 cubic meters per second per minute. If it is detected that the value exceeds 0.5 cubic meters per second per minute, it will be marked and reviewed in the record. To refine this parameter, it is necessary to trace back 15 minutes before and 15 minutes after the operation trigger moment, and summarize the flow rate data per minute. Specifically, it can be described by where is the maximum flow rate increment within this interval, is the number of minutes used to generate this increment. Finally, is recorded into . For example, in Operation No. 2, the average inflow rate in the 15 minutes before the start is 145 cubic meters per second, and the peak inflow rate in the 15 minutes after the operation starts is 170 cubic meters per second. Divide (170 - 145) by 10 minutes to get 2.5 cubic meters per second per minute, that is =2.5 cubic meters per second per minute.

[0087] The acquisition steps are as follows: This parameter represents the duration of flow disturbance. Its value is calculated by tracking the duration that the inflow rate or the flood discharge rate stays within the obvious fluctuation interval. When the monitoring system detects that the flow rate reaches or exceeds a statistical vibration determination threshold once, the timing starts until the flow rate returns to the pre - operation level and then the timing stops. The threshold is selected as 110% of the average pre - operation flow rate. Once the flow rate exceeds 110% of this average value, it is considered to enter the disturbance. If it is always lower than this threshold, then is 0, and the value usually does not exceed 300 minutes. For example, for Operation No. 3, the average pre - operation inflow rate is 180 cubic meters per second, and 110% of it is 198 cubic meters per second. When the system observes that the flow rate exceeds 198 cubic meters per second after the operation trigger and lasts for 40 minutes before falling back to around 180 cubic meters per second, then is 40 minutes.

[0088] The acquisition steps are as follows: This parameter is used to record the flood discharge response delay, which is the time difference between the moment when the gate is opened and the moment when a significant increment in the flow rate appears at the downstream monitoring section. During the operation, first read the moment when the gate is opened, and then continuously track the flow rate per minute at the downstream monitoring section. When the flow rate exceeds 120% of the average value before the gate is opened, determine this moment as the end point of the delay. Use (end moment - opening moment) to obtain , for example, the 4th operation is started at 13:20, and the downstream monitoring section finds that the flow rate reaches 120% of the average value before the start at 13:50, then this operation = 30 minutes.

[0089] The acquisition steps are as follows: This parameter represents the water level lagging and falling time, which is the duration of tracking the water level from the peak to the average water level before closing the gate after the gate is closed. First, record the gate closing time, and then find the time node when the water level returns close (i.e., within a deviation of ±0.1 m) to the average water level before closing the gate in the subsequent monitoring sequence. The difference between this node and the gate closing time is the , for example, in a gate closing operation, the average water level before closing the gate is 51.5 m, the gate closing time is 14:10, and the water level drops and stabilizes at 51.6 m at 14:35, then = 25 minutes.

[0090] The acquisition steps are as follows: This parameter refers to the starting time of the scheduling, which is marked in minutes as the interval from 0:00 on the same day to the triggering of the scheduling. For example, if it is triggered at 2:00 in the early morning, it is recorded as 120 minutes, and if it is triggered at 12:00 at noon, it is recorded as 720 minutes. To make the parameter more comparable, it will also be interactively compared with a day-night cycle characteristic table to check whether the scheduling starts at night or during the day. When summarizing, all scheduling operations are arranged in ascending order. For example, the 5th operation is triggered at 7:30 in the morning on the same day, and the corresponding time is 450 minutes, then = 450.

[0091] Calculation process:

[0092] First, calculate , when = 2.5, , , then If = 3.1, then ;

[0093] Multiply it by , if = 40, then , so the numerator part ;

[0094] The denominator part , if = 30, = 25, , , so the denominator , and the numerator 78.58 divided by the denominator 3.236 is approximately equal to 24.29;

[0095] Recalculate , if = 450, = 3.1, ;

[0096] Add 24.29 and 15.59 to get 39.88. Taking the absolute value of it is still 39.88, and finally substitute it into the outermost layer ;

[0097] From this, we can obtain .

[0098] This result indicates that when all parameters are configured and calculated according to the above measured values, the scenario hydraulic impact degree corresponding to the th scheduling operation is approximately 6.32, indicating that the scheduling operation presents a certain degree of impact amplitude in the current hydraulic environment. If the value increases to a higher level, it can indicate that the impact brought by this operation on water level or flow rate is more significant, while when is lower than a relatively small interval, it represents that the impact degree is limited, which is a reference basis for subsequent further decision-making.

[0099] The steps to obtain the quantitative risk index library are as follows:

[0100] Based on the scenario hydraulic impact degree, extract the scenario hydraulic impact degree values corresponding to the numbers of each scheduling operation, and simultaneously retrieve the real-time displacement of the dam body, the rise height of the phreatic line around the dam body, the peak downstream flow velocity, the increase amplitude of the seepage pressure at the dam foundation, and the growth of the structural crack length corresponding to the numbers. Uniformly organize the corresponding relationship of the number index to generate the basic parameter set of the scheduling operation risk;

[0101] According to the basic parameter set of the scheduling operation risk, calculate the dam-break probability index corresponding to the scheduling operation. The calculation formula is:

[0102] ;

[0103] Among them, is the dam-break probability index of the th scheduling operation, is the real-time displacement of the dam body corresponding to the th scheduling operation, is the rise height of the phreatic line around the dam body corresponding to the th scheduling operation, is the peak downstream flow velocity corresponding to the th scheduling operation, is the increase amplitude of the seepage pressure at the dam foundation corresponding to the th scheduling operation, is the growth value of the structural crack length corresponding to the th scheduling operation, The scenario hydraulic impact degree of the

[0104] Based on the dam-break probability index, each scheduling operation is bound to the dam-break probability index one by one according to the operation number, and an index structure is established in the order of the numbers to generate a quantitative risk index library.

[0105] Specifically, based on the corresponding relationship between the scenario hydraulic impact degree value and the scheduling operation number, first disassemble the number listed in the scheduling operation list and its associated scenario hydraulic impact degree value, and then retrieve the parameters such as the real-time displacement of the dam body, the rising height of the phreatic line around the dam body, the peak downstream flow velocity, the rising amplitude of the seepage pressure at the dam foundation, and the growth of the structural crack length one by one. The sources of these parameters include the dam structure monitoring records, the seepage observation records around the dam, the acquisition results of the downstream flow velocity monitoring equipment, and the measurement results of the dam body cracks during regular inspections. After retrieval, the values under each number are matched in the time axis and spatial position to ensure that the measurement period of the real-time displacement of the dam body coincides with the detection time period of the rising height of the phreatic line, and at the same time, check the overlap degree between the specific moment when the peak downstream flow velocity appears and the operation trigger interval. If it is found that the monitoring data of a certain number scheduling operation jumps between the front and back two measurement points, it is necessary to compare with other continuous observation nodes to judge the rationality of the value. If the abnormality is confirmed, record the relevant entries and manually check whether there is data deviation. For the rising amplitude of the seepage pressure at the dam foundation, compare the acquisition readings of the front and back two seepage pressure measurement points to obtain the rising amplitude. If the rising amplitude significantly exceeds the preset interval of 0 MPa to 0.5 MPa, it is necessary to reconfirm the acquisition process and equipment calibration. For the growth of the structural crack length, refer to the initial crack starting length and the final measurement length marked by the crack detection personnel during daily inspections to calculate the increment. If the increment is higher than the preset upper limit of 2 meters, record it as an emergency state and recheck it. After completing all data verification, index the number, the scenario hydraulic impact degree value, and the above-mentioned parameters together in the order of the number to ensure that each scheduling operation has a matching hydraulic impact degree, real-time displacement of the dam body, rising height of the phreatic line, peak downstream flow velocity, rising amplitude of the seepage pressure at the dam foundation, and crack length growth value. In this way, a relatively complete basic parameter set of the scheduling operation risk can be formed.

[0106] Formula: , The advantage of the formula is that by simultaneously introducing the real-time displacement of the dam body, the rising height of the phreatic line around the dam body, the peak downstream flow velocity, the rising amplitude of the seepage pressure at the dam foundation, the growth of the structural crack length, and the scenario hydraulic impact degree, non-linear coupling processing is carried out on various risk factors. The combination between the logarithm and the exponent is used to amplify the small parameter fluctuations to a certain extent, while the large parameters are suppressed within an appropriate range. In this way, the impact degree of the scheduling operation on the dam body safety can be more comprehensively described, which is convenient for quantitatively expressing the dam-break risk in the subsequent evaluation.

[0107] The acquisition steps are as follows: This parameter represents the real-time displacement of the dam body during the th scheduling operation. The monitoring means include arranging multiple displacement sensors on the dam body. These sensors record the change in the dam body displacement every 10 minutes, and perform a difference with the previous reading to obtain the displacement increment, and then perform a rolling accumulation within the same day to grasp whether there is a concentrated change in the dam body displacement in a short period of time. If the monitoring period spans multiple hours, it is merged according to the segmented records to form the total displacement corresponding to the scheduling operation period. When acquiring, it is necessary to exclude measuring point failures and data interference, such as instantaneous reading offsets caused by human movement of the measuring device or external strong impacts. Specifically, upper and lower limit checks can be performed on each record. For example, control it within the range of 0 meters to 0.05 meters. If it exceeds 0.05 meters, it means that the displacement is too large and the sensor status needs to be rechecked. Finally, take the verified reading as , for example, during the duration of the 1st scheduling operation, if the cumulative reading of the displacement sensor reaches 0.02 meters during this period, then is determined to be 0.02 meters. The method for establishing this range is to check the monitoring records of the dam body in the past two years, collect all displacement data during 200 consecutive operating days, calculate the difference between the maximum and minimum values to obtain the distribution interval, and then calculate the mean and standard deviation. The interval of 0 meters to 0.05 meters is defined as the relatively common displacement floating range. If a displacement amount greater than this range appears, it is necessary to immediately conduct a deep investigation.

[0108] The acquisition steps are as follows: This parameter represents the rising height of the phreatic line around the dam body. The monitoring method is to arrange seepage pipes or observation wells on the dam shoulders and around the dam body, regularly read the phreatic line height, and obtain a rising value after comparing it with the baseline. In actual use, the phreatic line is monitored at an interval of once per hour. Subtract the measured value of the phreatic line in the current hour from the phreatic line reference value before the start of the previous scheduling operation, and summarize the rising values. If the rise exceeds 1.0 meter, it needs to be recorded and other observation wells need to be rechecked for comparison. For example, during the duration of the 2nd scheduling operation, the phreatic line rises from the reference value of 44.2 meters to 44.8 meters, with a rise of 0.6 meters, then = 0.6 meters.

[0109] The acquisition steps are as follows: This parameter refers to the peak downstream flow velocity, which is obtained by continuously observing the flow velocity probes on the downstream monitoring cross-section. During each scheduling operation, select the peak point among them and record it as , whose significance lies in observing the impact of gate operation on the downstream water flow velocity. When reading the data, first check whether the peak value of this flow velocity exceeds the range of 0 m / s to 5 m / s. If it exceeds 5 m / s, it is necessary to further review the probe status and the river channel flow. For example, during a certain flood control operation, if the downstream flow velocity probe shows that the peak value reaches 2.8 m / s, it is recorded as = 2.8 m / s.

[0110] The acquisition steps of are as follows: This parameter refers to the rising amplitude of the seepage pressure at the dam foundation, which is obtained by reading the pressure change relative to that before the start of the previous operation from the dam foundation monitoring wells or piezometers. During the execution of the operation, the seepage pressure value is usually read every 20 minutes and compared with the reference value. If the rising value is higher than 0.5 MPa, it is marked as abnormal. For example, during the No. 3 operation, the seepage pressure increases from 0.8 MPa to 1.1 MPa, with a rise of 0.3 MPa, then = 0.3 MPa.

[0111] The acquisition steps of are as follows: This parameter corresponds to the growth value of the structural crack length. During daily inspections, the staff use crack monitors or visual recognition means to record the starting coordinates and ending coordinates of the cracks. If the length of the same crack increases after the operation, this increment is recorded as , and the range is generally between 0 m and 2 m. Repeated measurements are required after each inspection. If it is found that multiple cracks have all increased, calculate the increments separately and then accumulate them. For example, after the No. 4 operation, during the inspection, it is found that a crack originally 2.3 m long has grown to 2.45 m, with a growth of 0.15 m, then = 0.15 m.

[0112] The acquisition steps of are as follows: This parameter represents the scenario hydraulic influence degree, which has been calculated in the previous text.

[0113] Calculation process:

[0114] First, process the numerator part . If = 0.02 m, = 0.6 m, = 2.8 m / s, = 0.3 MPa, then first calculate = 2.5, add 1 to get 3.5, , divide by 1.871 to get 0.3315, and after squaring, it is approximately 0.1100;

[0115] Then calculate . If = 0.15 m, = 6.32, ; ;

[0116] Adding the two gives ;

[0117] Substituting into the outermost layer and , ,

[0118] Therefore , that is, the dam-break probability index is 0.8723.

[0119] This result indicates that when the dam displacement is 0.02 m, the uplift height of the phreatic line is 0.6 m, the peak downstream flow velocity is 2.8 m / s, the increase in seepage pressure at the dam foundation is 0.3 MPa, the increase in the length of structural cracks is 0.15 m, and the scenario hydraulic impact degree is 6.32, the dam-break probability index corresponding to this dispatching operation reaches 0.8723, indicating a relatively high risk during this dispatching process. If this value further increases, it indicates a lower safety level of the dam under the dispatching operation. When it is lower than 0.2 or 0.3, the risk can be considered to be in a relatively controllable state, and in subsequent evaluations, the dam-break probability indices of other operations can be compared and sorted, and corresponding control measures can be taken.

[0120] Based on the dam-break probability index, first check the number of each dispatching operation one by one against the calculated dam-break probability index value, and retrieve the records of these dispatching operation numbers in chronological order. If it is found that some dispatching operations occur within a very short time interval and the dam-break probability indices all exceed the warning value of 0.8 established through engineering practice, it is necessary to cross-verify with the monitoring logs around the dam to check whether the input of monitoring data is accurate and whether there are abnormal operation trigger events. If confirmed, these numbered operations can be marked as key attention objects. Subsequently, compare those operations with dam-break probability indices falling within the range of 0.2 to 0.8, and determine whether their safety status is within an acceptable range by comparing parameters such as seepage pressure changes, uplift amounts of the phreatic line, and newly added structural cracks in the dam body. Check whether they belong to the same batch of dispatching in the current period. If they belong to the same batch, sort these numbers in descending order of the dam-break probability value and record them in the index structure. For operations with probability values lower than 0.2, classify them as low-risk operations and directly attach corresponding real-time dam displacements and downstream flow velocities, etc. information. List the final integrated data item by item according to the numbers to establish an index file of the quantitative risk index library for subsequent risk identification. After such a number binding and retrieval process, the dam-break probability indices of different operations can be clearly presented and provided to relevant management personnel together with their respective monitoring data for further decision-making.

[0121] The steps to obtain the risk-benefit balance index are as follows:

[0122] Extract the operation number of each scheduling operation based on the quantified risk index library Extract the corresponding dam-break probability index, and synchronously retrieve the regional water supply achievement, water supply plan value for the corresponding time period, power generation output value, installed capacity, and power generation duration recorded in the scheduling operation to generate a basic parameter set for water supply and power generation;

[0123] Based on the basic parameter set for water supply and power generation, calculate the risk-benefit balance index corresponding to the scheduling operation. The calculation formula is: ;

[0124] Where is the risk-benefit balance index corresponding to the th scheduling operation, is the water supply guarantee rate corresponding to the th scheduling operation, is the power generation output value corresponding to the th scheduling operation, is the installed power generation capacity corresponding to the th scheduling operation, is the dam-break probability index corresponding to the th scheduling operation.

[0125] Specifically, based on the dam-break probability index records corresponding to the quantitative risk index library, first extract the regional water supply achievement and water supply plan value during the time period of each scheduling operation number from the scheduling operation list, and compare the two to obtain the water supply guarantee rate during a certain period. At the same time, read the power generation output value, installed capacity, and power generation duration from the scheduling operation record. The water supply achievement can be obtained from the daily water output statistical data of the regional water plant. Compare the actual total water consumption during the same period with the planned demand to check whether they match. If it is higher than the preset demand, mark the achievement rate as exceeding 1 in the record. If it is lower than the demand, calculate the specific value of the achievement rate according to the difference. At this time, check whether the achievement rate significantly exceeds the normal range of 0 to 1. If it does, trace back the water supply statistical input process. At the same time, obtain the power generation output value from the operation record of the hydropower unit and check whether the unit numbers are consistent. Then read the installed capacity of the power generation for analysis based on the power ratio. The installed capacity is usually in the range of 100MW to 500MW. If it is found that the installed capacity record is 0 or significantly less than this range, it needs to be checked in time. The power generation duration is obtained by subtracting the start time and end time of each continuous operation of the unit. The power generation duration is between 0 hours and 24 hours. If it is detected that the single operation duration is negative, it means that the record is incorrect. After checking the above indicators, catalog the dam-break probability index under each scheduling operation number together with its corresponding water supply and power generation data to ensure that the time intervals of each value correspond correctly. If the water supply data and power generation data records corresponding to a certain number are both complete, mark the status of this number as available. If there are data gaps in some records, mark them as to be supplemented and collect data for the same period in the future to check and fill in the gaps. Finally, integrate the above-matched data in the order of the numbers and add appendices to form a water supply and power generation basic parameter set.

[0126] Formula: , The benefit of the formula is to comprehensively weigh the risks and benefits of a single scheduling operation by combining factors such as water supply guarantee rate, power generation output value, installed capacity, and dam-break probability index. Through operations such as logarithm, absolute value, and square root, parameters of different magnitudes are incorporated into the same framework to achieve a balanced expression between water supply effectiveness and safety risks, providing quantitative reference for scheduling decisions.

[0127] The acquisition steps of are as follows: This parameter represents the water supply guarantee rate, and its value is obtained by the ratio of the regional water supply achievement to the water supply plan value during the same period. After each scheduling operation starts, extract the start and end times of this operation. During this time period, summarize the actual water supply volume and compare it with the pre-developed water supply plan volume. If the actual water supply volume is recorded as and the planned volume is recorded as , then , for example, during a certain operation, the measured water supply volume is 200,000 , planned quantity: 250,000 , then .

[0128] The acquisition steps of are as follows: This parameter represents the power generation output value. During the execution of the dispatching operation, monitor the real-time power readings of the power system, compare the peak and average values of the power recorded every hour. If the difference between the peak and average values is within a certain range (such as 0 MW to 50 MW), it is determined that the power generation load is in a relatively stable state, and then the average power of this time period is selected as , or the average power output can also be calculated from the segmented statistics of the power generation after the unit operation ends, and then compared with the unit log record. For example, during a certain dispatching operation, the average output power of the unit is 120 MW, then MW.

[0129] The acquisition steps of are as follows: This parameter refers to the installed power generation capacity, which reflects the maximum output level designed by the hydropower station unit. If the installed capacity in the record exceeds 500 MW, verify the equipment file. Generally, the installed capacity of a single unit of a medium-sized hydropower facility is between 50 MW and 200 MW. For example, for unit No. 3, its rated installed capacity is queried to be 150 MW, then .

[0130] The acquisition steps of are as follows: This parameter is the dam break probability index, which has been calculated and recorded in the quantified risk index library in the previous text.

[0131] Calculation process:

[0132] First calculate , if = 0.8, , then ,

[0133] Then calculate , if = 120, = 150, then ,

[0134] Multiply the two and take the absolute value: ,

[0135] Take the square root of it: ,

[0136] The denominator part , if = 0.8723, ,

[0137] Divide the numerator 0.6274 by 1.9339, approximately 0.3245, and cube it: ,

[0138] Therefore .

[0139] The result shows that when the water supply guarantee rate is 0.8, the power generation output value is 120 MW, the installed capacity is 150 MW, and the dam break probability index is 0.8723, the risk-benefit balance index is approximately 0.0342. The larger this value is, the higher the proportion of the dispatching operation in terms of water supply and power generation benefits. However, if the value is too small, it means that although it can provide certain power generation or water supply benefits, the relative risk level cannot be ignored. It can be compared with the balance indexes of other operations in the subsequent evaluation to select a more appropriate dispatching plan.

[0140] The steps to obtain the risk constraint effectiveness ranking list are as follows:

[0141] Based on the risk-benefit balance indexes of each dispatching operation, using the risk-benefit balance index as the ranking basis, compare the numerical values item by item and perform a descending order sorting to form a scenario risk-benefit ranking sequence;

[0142] Based on the scenario risk-benefit ranking sequence, call the corresponding scenario numbers in the sorting order, perform index matching item by item, and use the risk ranking order as the index to establish a risk constraint effectiveness ranking list.

[0143] Specifically, based on the risk-benefit balance index of each scheduling operation, first read the pairing content of the corresponding number and its risk-benefit balance index item by item from the previously established scheduling operation list. Compare these balance indexes numerically in the same dimension. When comparing, all record items need to be listed in the order of the number and the risk-benefit balance index among them needs to be extracted. During the extraction process, attention should be paid to whether there are numerical errors or omissions. For example, when it is detected that the balance index of a certain operation number is negative or exceeds an abnormal value such as 2, the water supply guarantee rate, power generation power, and dam-break probability indicators under the scheduling operation need to be checked. After confirming that the input process and calculation process are correct, the process can continue. If the balance index is between 0 and 1, it indicates that the value is relatively reasonable. If it is higher than 1, it is also necessary to check whether it is caused by the water supply achievement exceeding the predetermined planned value too much. After all values are checked and confirmed to be available, these balance indexes are sorted in descending order according to the specified sorting basis. For example, first scan the index records of all operation numbers and count the balance index of each number, then copy the obtained index list as temporary data and compare them in descending order of value. The comparison method can be pairwise comparison. For example, for two consecutive items, judge the size relationship of their balance indexes respectively. If it is greater than or equal to the previous item, their positions are swapped until the position of each record no longer changes. It is also possible to use the method of directly selecting the maximum value and inserting it into the above index list to find the largest balance index in this batch of scheduling operations and place it at the top of the list, and then select the next largest balance index in turn to complete the sorting of all numbers. If there are the same balance indexes for some numbers, they are ranked and supplemented according to the order of the operation trigger time. After each sorting is completed, a scenario risk-benefit sorting sequence arranged from large to small is formed.

[0144] Based on the scenario risk-benefit ranking sequence, check the association between each scenario number in the ranking sequence and the original dispatching operation item by item from the numbered index dataset cited above. Perform sequence number index matching according to the sequence ranking number. If a dispatching operation that matches the current number is found in the index data, read the corresponding hydraulic environment and water supply and power generation information for further recording. At the same time, determine whether the scenario number falls within the group of high-risk dispatching operations. If the dam break probability index is higher than 0.8 or other monitoring parameters exceed the predetermined standard, such scenario numbers need to be marked additionally. Then check whether the same situation also exists in the remaining dispatching operations in the ranking sequence. Gather all the marked high-risk operations together for later reference. If a number in the sequence fails to be successfully paired in the index data, trace back the dispatching record corresponding to this number to check whether it has been regarded as data abnormal in the previous steps. If it is confirmed to be in an abnormal or invalid state, remove this number from the index matching result this time. The remaining numbers that are successfully paired and have no abnormalities continue to be summarized in the sorting order. For each number, retain its previously inserted order identifier. Finally, add the risk ranking order to the number list so that it can clearly view the corresponding scenario and its number from the first to the last order. In this way, a risk constraint effectiveness ranking list can be constructed.

[0145] The steps for obtaining the real-time risk warning signal are as follows:

[0146] Call the dispatching operation number with the earliest ranking number in the risk constraint effectiveness ranking list, extract information item by item, and generate the scenario plan with the highest priority.

[0147] According to the scenario plan with the highest priority, obtain the dam break probability index value, determine the real-time risk level threshold interval into which the dam break probability index value falls, and obtain the real-time risk warning signal.

[0148] Specifically, call the scheduling operation number with the earliest sorting number in the risk constraint efficiency ranking list. First, read the corresponding gate opening period, water level situation, water supply achievement volume, power generation power output and other field contents of this number in the index structure constructed previously. Cross-compare these fields with the dam break probability index and its corresponding period to check for data overlap. If it is found that the water supply achievement volume does not match the planned value or the power generation power output has a too large gap from the average installed capacity, then compare the adjacent operation numbers before and after again to confirm whether the recorded value in this number is within the range of 0 to 1 or 0 MW to 500 MW. If everything is normal, retain this number and check whether it is the first in the ranking list. If it has been determined that it is the earliest number, import the corresponding dam body monitoring information, such as retrieving data on dam body displacement, peak flow velocity, and increase in seepage pressure collected previously, and compare with the pre-established monitoring standards. For example, the increase in seepage pressure should be within the range of 0 MPa to 0.5 MPa. If any record breaks this limit, mark it as abnormal in the comparison log and summarize it to the subsequent inspection items. After confirming that the data corresponding to the number is valid, continue to check other relevant information, including whether the change in gate opening corresponds to the instantaneous slope of the inflow rate within the range of 0 cubic meters per second per minute to 0.5 cubic meters per second per minute defined previously. If all parameters meet the previous safety range, mark this number as executable in the record, and then generate the scenario plan with the highest priority. Archive this plan and provide reference data for the next link, and finally obtain the scenario plan with the highest priority.

[0149] According to the scenario plan with the highest priority, load the dam break probability index value recorded in it, and compare this value with the grading standard set by the project team. This grading standard is obtained by comprehensively considering the dam break risk distribution in the past three years and the results of regular on-site inspections. Divide 0 to 1 into several intervals. For example, 0 to 0.2 is defined as low risk, 0.2 to 0.5 is defined as medium risk, 0.5 to 0.8 is defined as relatively high risk, and exceeding 0.8 is defined as high risk. When the dam break probability index falls into the corresponding interval during this scheduling operation, first append risk grading information to this number in the record. If it is found that the value is higher than 0.8, mark this operation number as high risk and output a real-time risk warning signal at the current moment. If it falls into other risk levels, indicate the level value and then continue to compare the downstream flow, dam foundation seepage pressure, and growth of dam body cracks, etc., and then summarize and generate the latest risk reminder. When the warning signal is output, retain the corresponding timestamp at the same time, and write this warning into the current monitoring information, so as to finally obtain the real-time risk warning signal.

Claims

1. A real-time early warning system for water conservancy risks with multimodal spatio-temporal fusion, characterized in that, The system includes: A multi-modal spatio-temporal data fusion module, which performs spatio-temporal alignment and feature extraction calculations based on the sensor data stream of the reservoir group, remote sensing images, and weather forecasts, obtains the fused spatio-temporal situation value, defines the gate opening and flood discharge flow adjustment options under different windows based on the fused spatio-temporal situation value and the preset scheduling rules, and establishes a potential scheduling operation set; A water conservancy risk quantification and assessment module, which associates the water level and the incoming flow to calculate the hydrodynamic response based on the potential scheduling operation set, calculates the influence degree of each operation option, obtains the scenario hydraulic influence degree, and evaluates the dam-break probability index based on the scenario hydraulic influence degree to construct a quantified risk index library; A resource and risk budget allocation module, which matches the water supply guarantee rate and the power generation benefit index to calculate the balanced performance of each scenario based on the quantified risk index library, obtains the risk-benefit balance index, and filters and ranks the scenarios based on the risk-benefit balance index to generate a risk constraint efficiency ranking list; A scheduling plan early warning and release module, which selects scenarios according to the priority based on the risk constraint efficiency ranking list and generates real-time risk warning signals according to the dam-break probability index of the scenarios; The steps for obtaining the risk-benefit balance index are as follows: Extract the number of each scheduling operation based on the above-mentioned quantified risk index library The corresponding dam-break probability index, synchronously retrieve the regional water supply achievement, the water supply plan value during the corresponding time period, the power generation output value, the installed capacity and the power generation duration recorded in the scheduling operation, and generate a basic parameter set for water supply and power generation; Based on the water supply and power generation basic parameter set, calculate the risk-benefit balance index corresponding to the scheduling operation. The calculation formula is: ； Among them, is the risk-benefit balance index corresponding to the th scheduling operation, is the water supply guarantee rate corresponding to the th scheduling operation, is the power generation output value corresponding to the th scheduling operation, is the installed power generation capacity corresponding to the th scheduling operation, is the dam-break probability index corresponding to the th scheduling operation.

2. The real-time early warning system for water conservancy risks with multimodal spatio-temporal fusion according to claim 1, characterized in that, The steps for obtaining the fused spatio-temporal situation value are as follows: Based on the water level height, incoming flow velocity, and flood discharge rate, perform linear interpolation and outlier removal processing respectively according to the time stamp, align the water level height, incoming flow velocity, and flood discharge rate to the same time series, combine the surface reflectance map and terrain slope map in the same time window of the remote sensing image, and synchronize the precipitation intensity grid sequence within the next three hours in the weather forecast to generate the sensor data stream of the reservoir group, the remote sensing image map, and the weather forecast grid sequence after time alignment; According to the sensor data stream of the reservoir group, the remote sensing image map, and the weather forecast grid sequence after time alignment, map the water level height, incoming flow velocity, flood discharge rate, and precipitation intensity at each moment to the three-dimensional geographical grid according to the regional grid number, and generate a multi-modal spatio-temporal feature vector set under the geographical grid in combination with the surface reflectance and terrain slope at the corresponding position; Based on the multi-modal spatio-temporal feature vector set under the geographical grid, calculate the fused spatio-temporal situation value.

3. The real-time early warning system for water conservancy risks with multi-modal spatio-temporal fusion according to claim 1, characterized in that, The steps for obtaining the potential scheduling operation set are as follows: Based on the fused spatio-temporal situation value, call the gate water level threshold, flood discharge warning line flow, and safety water level interval in the preset scheduling rules, perform parameter threshold comparison item by item, map the fused spatio-temporal situation value to the scheduling conditions in turn, and generate a spatio-temporal situation matching parameter set that meets the gate scheduling conditions; Based on the spatio-temporal situation matching parameter set that meets the gate scheduling conditions, call the gate opening adjustment level and flood discharge flow adjustment level defined for different time windows in the preset scheduling rules, and perform gate opening level selection item by item to generate a potential scheduling operation set.

4. The real-time early warning system for water conservancy risks with multi-modal spatio-temporal fusion according to claim 1, characterized in that, The steps for obtaining the scenario hydraulic influence degree are as follows: Based on the set of potential scheduling operations, call the change amount of the gate opening, the change amount of the flood discharge flow rate, the gate opening duration, the water level monitoring sequence, the inflow monitoring sequence, and the operation trigger time in each scheduling operation, and establish a parameter correspondence table according to the scheduling operation number respectively to generate a set of water level and inflow scheduling response combinations; According to the set of water level and inflow scheduling response combinations, conduct hydrodynamic simulations item by item to obtain the maximum water level rise amplitude, the instantaneous slope of the inflow, the flow disturbance duration, the flood discharge response delay, and the water level lagging fall time corresponding to each scheduling operation, and generate a set of hydrodynamic response parameters for the scheduling operations; Based on the set of hydrodynamic response parameters for the scheduling operations, calculate the scenario hydraulic impact degree, and the calculation formula is: ； Among them, is the scenario hydraulic influence degree of the th scheduling operation, is the maximum rising amplitude of the water level corresponding to the th scheduling operation, is the instantaneous slope of the inflow corresponding to the th scheduling operation, is the duration of the flow disturbance corresponding to the th scheduling operation, is the flood discharge response delay corresponding to the th scheduling operation, is the time for the water level to lag and fall back corresponding to the th scheduling operation, is the starting time of the scheduling corresponding to the th scheduling operation.

5. The real-time early warning system for water conservancy risks with multimodal spatio-temporal fusion according to claim 1, characterized in that, The steps for obtaining the quantified risk index library are as follows: Based on the scenario hydraulic impact degree, extract the scenario hydraulic impact degree values corresponding to each scheduling operation number, and simultaneously retrieve the real-time dam displacement amount, the rise height of the phreatic line around the dam, the peak downstream flow velocity, the rise amplitude of the dam foundation seepage pressure, and the increase in the length of the structural crack corresponding to the number, and uniformly organize the index correspondence relationship to generate a set of basic risk parameters for the scheduling operations; According to the set of basic risk parameters for the scheduling operations, calculate the dam-break probability index corresponding to the scheduling operation, and the calculation formula is: ； Among them, is the dam-break probability index of the th scheduling operation, is the real-time displacement of the dam body corresponding to the th scheduling operation, is the rising height of the phreatic line around the dam body corresponding to the th scheduling operation, is the peak downstream flow velocity corresponding to the th scheduling operation, is the rising amplitude of the seepage pressure at the dam foundation corresponding to the th scheduling operation, is the growth value of the structural crack length corresponding to the th scheduling operation, is the scenario hydraulic influence degree of the th scheduling operation; Based on the dam-break probability index, bind each scheduling operation and the dam-break probability index one by one according to the operation number, and establish an index structure in the order of the numbers to generate a quantified risk index library.

6. The real-time early warning system for water conservancy risks with multi-modal spatio-temporal fusion according to claim 1, characterized in that, The steps for obtaining the risk constraint effectiveness ranking list are as follows: Based on the risk-benefit balance index of each scheduling operation, use the risk-benefit balance index as the ranking basis, compare the numerical sizes item by item, and perform a descending order to form a scenario risk-benefit ranking sequence; Based on the scenario risk-benefit ranking sequence, call the corresponding scenario numbers according to the ranking order, perform serial number index matching item by item, and use the risk ranking order as the index to establish a risk constraint effectiveness ranking list.

7. The real-time early warning system for water conservancy risks with multi-modal spatio-temporal fusion according to claim 1, characterized in that, The steps for obtaining the real-time risk warning signal are as follows: Call the scheduling operation number with the earliest ranking number in the risk constraint effectiveness ranking list, extract the information item by item, and generate the scenario plan with the highest priority; According to the scenario plan with the highest priority, obtain the numerical value of the dam-break probability index, determine the real-time risk level threshold interval into which the numerical value of the dam-break probability index falls, and obtain the real-time risk warning signal.

Images