Staged dynamic flood control water level regulation and control method suitable for hydropower station

Abstract

The invention discloses a stage-by-stage dynamic flood control water level regulation and control method suitable for a hydropower station, and the method comprises the steps: obtaining watershed hydro meteorology, engineering operation and topographic geology multi-source basic data, and carrying out the Spark distributed cleaning, wavelet denoising and GAN missing value complementation preprocessing, thereby forming a standardized data set; an optimal flood season staging scheme is screened through coupling of qualitative analysis and a quantitative method; then, the safe pre-discharge amount and the effective pre-discharge time serve as sensitive factors, a flood control water level dynamic control model is built in combination with fuzzy mathematics and Copula joint distribution, and flood control water level process lines and regulation and control domain values of all flood season stages are generated; the safety of a regulation and control threshold value is verified through finite element simulation and a two-dimensional hydrodynamic model, a regulation and control scheme meeting flood control safety and efficient utilization of water resources is output, dynamic and accurate regulation and control of the flood control water level of the hydropower station are achieved, and the problems that in traditional static flood control water level dispatching, the contradiction between flood control and benefit making is prominent, the staging precision is low, and safety verification is insufficient are solved.

Description

Technical Field

[0001] This invention relates to the field of water conservancy engineering technology, specifically to a phased dynamic flood control water level regulation method applicable to hydropower stations. Background Technology

[0002] As a core water conservancy hub coordinating water resource regulation and energy supply, the scientific nature of the flood control water level scheduling scheme of hydropower stations directly determines the balance between flood control safety and water resource utilization efficiency. During long-term operation, most hydropower stations, limited by the technical conditions and design concepts of the construction period, have gradually revealed the incompatibility of traditional flood control water level scheduling models with current needs. Early designs often adopted fixed flood control water levels or simple phased water level models, failing to fully consider the dynamic changes in the basin's hydrological and meteorological conditions and the increasing demands for flood control and water conservancy driven by economic and social development. During the main flood season, to avoid flood risks, such models require maintaining a low water level to prepare for floods, often resulting in large amounts of floodwater being wasted due to insufficient storage. In the later flood season, although attempts are made to raise the water level for storage, the lack of accurate judgment on the timing of the flood's end makes them susceptible to sudden rainfall and forced to release water, ultimately leading to insufficient water storage. This makes it difficult to simultaneously meet the needs of power generation, ecological water use, and downstream production and domestic water use, creating a prominent contradiction between flood control safety and efficient water resource utilization.

[0003] In the current field of flood control water level scheduling for hydropower stations, the limitations of flood season phasing methods and dynamic control models restrict the improvement of scheduling accuracy. In the flood season phasing stage, existing technologies mostly rely on single analytical methods, such as using fuzzy set analysis to fit the distribution function to determine phasing nodes, or relying on Fisher's optimal segmentation method for multi-index clustering. However, the former is easily affected by the subjectivity of single indicators, while the latter has stringent requirements for data integrity and continuity. When faced with sudden changes in hydrological sequences or extreme weather events, the reliability of the phasing results decreases significantly. At the same time, existing dynamic control models for flood control water levels lack universality and specificity. Some key parameters of these models require data from specific engineering facilities (such as discharge data from flood discharge tunnels). If the hydropower station does not possess the corresponding engineering conditions, the model cannot be directly applied. Furthermore, most models do not fully integrate dynamic information such as short-term precipitation forecasts and flood evolution simulations, resulting in insufficient control accuracy in the timing and amount of pre-discharge, failing to maximize the utilization of flood resources within the flood control safety boundary.

[0004] Technical shortcomings in the structural safety assessment and downstream channel flow capacity evaluation further limit the dynamic adjustment space of the flood control water level. On the one hand, after long-term operation of hydropower stations, material deterioration problems such as concrete carbonization, steel corrosion, and gate steel structure fatigue gradually emerge. Traditional safety assessments are mostly based on static water level conditions for structural strength calculations, failing to fully consider the complex dynamic loads such as pulsating pressure and vortex-induced vibration caused by different gate openings during dynamic flood control water level adjustments. Furthermore, the impact of material deterioration on structural bearing capacity is not incorporated into simulation analysis, making it difficult for safety assessment results to reflect actual operational risks. On the other hand, downstream channel flow capacity analysis often relies on simplified hydraulic models and historical measured data, failing to fully consider dynamic factors such as inter-regional flood inflow and riverbed scouring and deposition deformation. For example, when using the Manning formula to calculate channel roughness, the roughness coefficient is not corrected for water level variability. When facing dynamic flood discharge changes, the water surface line calculation error is large, making it impossible to accurately define the channel flood safety boundary under different flood control water level schemes, leading to the risk of local river overflow.

[0005] From the perspective of the current state of technological development in the industry, although technologies such as big data and numerical simulation have been gradually applied to the field of hydropower station dispatching, the depth of technological integration and the synergy of the entire process still need to be improved. At the data processing level, most studies only achieve basic integration of single-source hydrological and meteorological data, failing to fully utilize multi-source heterogeneous data (such as watershed underlying surface information retrieved from remote sensing), and failing to achieve efficient mining of massive amounts of data through distributed computing frameworks, resulting in insufficient in-depth analysis of the patterns of watershed rainstorms and floods. At the model application level, flood season phasing, water level control, and safety demonstration are often treated as independent modules, lacking a closed-loop mechanism of "phasing-control-verification." For example, flood season phasing results are not directly fed back to the optimization of water level control model parameters, and engineering safety simulations do not synchronously match the load changes caused by dynamic water level adjustments, resulting in poor connectivity between various technical links and difficulty in forming a systematic dispatching plan. Furthermore, a unified dynamic flood limit water level risk assessment system has not yet been established within the industry; the quantitative standards for flood control risk, beneficial effects, and ecological impacts vary, and there is a lack of comparable basis between different dispatching schemes, hindering the large-scale promotion and application of dynamic dispatching technology. Summary of the Invention

[0006] Based on the aforementioned technical problems, this application discloses a phased dynamic flood control water level regulation method applicable to hydropower stations, thereby solving the aforementioned technical problems, specifically as follows:

[0007] Acquire multi-source basic data of the watershed where the hydropower station is located, including at least hydrological and meteorological data, engineering operation data, and topographic and geological data;

[0008] The multi-source basic data is preprocessed by data cleaning, noise reduction, and missing value completion to obtain a standardized dataset;

[0009] Based on the standardized dataset, a combination of qualitative and quantitative analysis was used to study the flood season phasing, and the optimal flood season phasing scheme that is suitable for the characteristics of the hydropower station basin was selected.

[0010] Based on the optimal flood season phasing scheme, and combined with the flood control safety constraints and water resource utilization needs of hydropower stations, a dynamic control model for flood limit water level is constructed with safe pre-discharge volume and effective pre-discharge time as sensitive factors.

[0011] Based on the dynamic control model of the flood control water level, the process lines of the flood control water level corresponding to different flood season stages are generated, and the dynamic flood control water level regulation range is determined.

[0012] The safety of the dynamic flood control water level control threshold is verified, and a phased dynamic flood control water level control scheme for hydropower stations that meets the requirements of flood control safety and efficient water resource utilization is output to control the flood control water level of hydropower stations at different flood season stages.

[0013] Preferably, the acquisition of multi-source basic data of the watershed where the hydropower station is located includes:

[0014] Collect historical rainfall data, flood event data, and evaporation data recorded by hydrological and meteorological stations within the basin;

[0015] Collect data on the relationship between reservoir water level, area, and capacity, inflow and outflow data, water level and discharge data, and the relationship between gate opening and discharge capacity during the operation of the hydropower station.

[0016] Airborne LiDAR technology was used to acquire grid DEM data of the downstream river channel, which was then integrated with measured topographic data to form a watershed topographic and geological dataset.

[0017] Preferably, the preprocessing of multi-source basic data includes:

[0018] The Spark distributed framework is used to clean multi-source basic data and remove abnormal data that exceeds the reasonable range of values.

[0019] High-frequency interference factors in hydrological and meteorological data are eliminated by wavelet denoising algorithm;

[0020] Complete missing hydrological sequence data using generative adversarial networks;

[0021] Coordinate unification and format conversion operations were performed on topographic and geological data to ensure that topographic data from different sources met the accuracy requirements for subsequent modeling.

[0022] Preferably, the flood season grading study based on standardized datasets employs a combination of qualitative and quantitative analysis. Specifically, the qualitative analysis stage identifies the seasonal variation trend of precipitation during the flood season by combining the basin's climate characteristics and the evolution of weather systems. The quantitative analysis stage extracts grading characteristic indicators based on standardized datasets, specifically including the temporal distribution of the maximum daily flow, the average daily flow, the average rainfall, and the maximum annual flood peak flow.

[0023] At least three different flood season grading schemes were obtained by using fuzzy set analysis to construct dynamic membership functions, Fisher's optimal segmentation method to calculate the sum of squared deviations within sample segments, and system clustering method to carry out multi-index clustering.

[0024] Using Bayesian optimization and fuzzy preference functions, and based on bootstrapping sampling, the relative flood frequency and generalized distance of confidence intervals for different flood season phasing schemes are calculated. The optimal flood season phasing scheme is then selected, where the generalized distance... The formula is: ,in For the actual sample value, The sample mean. This represents the number of samples.

[0025] Preferably, the step of using Fisher's optimal segmentation method to calculate the sum of squared deviations within the sample segment specifically involves:

[0026] The time series samples corresponding to the staged characteristic indicators are divided into: Let the data points be divided into 10 data points. Section, of which Segment sample Calculate the mean of each sample segment, expressed as: ;

[0027] The formula for calculating the sum of squared deviations within a sample segment is: ,in The objective function is the sum of the squared deviations of each segment; the objective function is constructed using the BIC criterion. Plot the curve of the objective function as a function of the number of categories, and select the number of categories corresponding to the inflection point of the curve as the optimal number of stages;

[0028] Principal component analysis, entropy weight method, and coefficient of variation method were used to determine the weights of characteristic indicators for each stage, where the coefficient of variation... The formula is: ,in The standard deviation of the indicator. The mean of the indicators is used to adjust the sum of squared deviations based on the weights, and the split points are optimized.

[0029] Preferably, the construction of the dynamic control model for the flood control water level, which uses the safe pre-discharge amount and effective pre-discharge time as sensitive factors, includes:

[0030] Based on fuzzy set theory, the domain is defined as the entire year's time series. The membership function of the fuzzy set for the flood season is established, and the formula is: ,in For the first Membership degree during the flood season For shape parameters, This is the central moment of the flood season;

[0031] The empirical distribution curves of flood occurrence frequency for each time period were fitted using the kernel density estimation method. Construct a comprehensive membership degree The formula is ,in These are the weighting coefficients;

[0032] By introducing the Copula multivariate joint distribution function, a joint distribution model of peak flow and flood volume is constructed. Let the peak flow be... , Flood volume Then the joint distribution function expression is: ,in For Copula functions, , These are the marginal distribution functions of peak flow and flood volume, respectively;

[0033] Using safe pre-discharge and effective pre-discharge time as input variables, a membership-capacity response function is established, expressed as follows: ,in To design flood control reservoir capacity, a dynamic control model for flood control limit water level was developed.

[0034] Preferably, the step of introducing the Copula multivariate joint distribution function to construct a joint distribution model of peak flow and flood volume includes:

[0035] Peak flow samples and corresponding flood volume samples for each flood season stage are extracted from the standardized dataset to form a two-dimensional sample set. The Clayton Copula function is used to construct the joint distribution function of peak flow and flood volume. The expression of the Clayton Copula function is as follows: ,in >0 represents the parameters of the Copula function, which are estimated using the maximum likelihood estimation method. The likelihood function expression is: ,in The density function of the Clayton Copula function;

[0036] The peak-volume combination values ​​under different return periods were calculated based on the joint distribution function and compared with historical measured flood data. The goodness of fit of the joint distribution model was verified by the KS test, and the KS statistic was calculated according to the following formula: ,in Let be the empirical distribution function. The theoretical distribution function is used to ensure the accuracy of the model's representation of the statistical characteristics of floods in different stages.

[0037] Preferably, the step of generating flood control water level process lines corresponding to different flood season stages and determining the dynamic flood control water level control threshold specifically involves:

[0038] Based on the optimal flood season phasing scheme, the flood season is divided into the pre-flood season, the main flood season, the post-flood season, and the transition period. The comprehensive membership degree is calculated for each stage of the flood season. By solving the comprehensive membership degree The intersection points with the reservoir capacity curve yield the initial flood control water level values ​​for each time period. Let the reservoir capacity curve be... Through formula Solve for the corresponding water level The initial flood control limit water level was obtained;

[0039] By combining short-term precipitation forecasts with flood forecasts, ensemble forecasting techniques are used to quantify forecast uncertainties and determine the effective pre-release time window. ,in For effective pre-release time, The standard deviation of the forecast error is given by the formula: ,in This is the error of a single forecast. The average forecast error, Number of forecasts;

[0040] Based on the pre-discharge capacity constraint method, the initial flood limit water level is modified by using the maximum discharge capacity of the gate and the safe discharge of the downstream river channel as constraints, thus forming the flood limit water level process line and dynamic control domain value for each flood season stage.

[0041] Preferably, the safety verification of the dynamic flood control water level regulation threshold is achieved by combining finite element simulation analysis of the stress and deformation characteristics of the engineering structure with hydrodynamic model verification of the downstream river channel flow capacity. Specifically:

[0042] Parametric geometric reconstruction technology was used to construct a three-dimensional finite element model of the dam, gate piers and steel gates. The dam body was constructed using eight-node hexahedral elements and the gates were constructed using six-degree-of-freedom flat plate elements.

[0043] Using different water level conditions and gate opening conditions corresponding to the dynamic flood control water level regulation threshold as input parameters, water load and flood discharge excitation load are applied. The flood discharge excitation load is obtained by computational fluid dynamics simulation of pulsating pressure distribution under different opening degrees, and the pulsating pressure power spectral density is used. The formula is: ,in For pulsating pressure time history, For sampling duration, For frequency;

[0044] A concrete damage plasticity model was used to simulate the material degradation of the dam body. Based on the measurement results of the steel corrosion rate, the reduction of the steel section was quantified. Nonlinear finite element calculations were carried out, and the stress-strain relationship of the concrete satisfied... ,in The initial elastic modulus, In response, For damage variables;

[0045] The dynamic stress, displacement, and acceleration response indices of the stress concentration areas of the dam body were extracted and compared with the allowable values ​​in the current specifications.

[0046] Preferably, the verification of downstream river channel flow capacity using a hydrodynamic model includes:

[0047] A two-dimensional hydrodynamic model is constructed using the unsteady flow Saint-Venant equations, where the continuity equations of the Saint-Venant equations are expressed as follows: The momentum equations are as follows: , ,in Because of the water depth, For time, , They are respectively , directional flow velocity, For lateral inflow, It is the acceleration due to gravity. , They are respectively , Direction bottom slope, , They are respectively , Directional friction slope , They are respectively , Surface shear stress in the direction of direction The density of water, Water pressure;

[0048] The solution domain is discretized using the finite volume method. The discharge flow process line corresponding to the dynamic flood control water level regulation scheme is used as the upstream boundary input. The dynamic inflow boundary of the tributary in the interval is generated by the distributed hydrological model and used as the lateral boundary.

[0049] The Manning formula is used to calculate the channel roughness. The Manning formula is expressed as follows: ,in For flow rate, The roughness coefficient is Manning's coefficient. For hydraulic radius, The model parameters were calibrated using hydraulic gradient data and historical flood scour data. The water level, flow velocity and water depth distribution of the downstream river under different flood discharge conditions were simulated, and the water surface line corresponding to each level of flow was calculated to verify whether the river's flow capacity meets the scheduling requirements.

[0050] Compared with the prior art, the technical solution of this application has the following technical effects:

[0051] This invention integrates hydrological, meteorological, engineering operation, and topographic and geological data through multi-source data fusion preprocessing and multi-method coupling flood season staging technology. It uses Spark distributed cleaning and GAN to complete missing values, and combines quantitative methods such as fuzzy set analysis and Fisher optimal segmentation with qualitative analysis of watershed climate to accurately identify the seasonal patterns of precipitation during the flood season. It then selects the optimal staging scheme that is suitable for the characteristics of the watershed, solving the problems of strong subjectivity and weak anti-interference ability of traditional single staging methods, and laying a scientific spatiotemporal boundary for dynamic regulation of flood control water levels.

[0052] This invention utilizes fuzzy mathematics and Copula multivariate joint distribution to construct a dynamic control model for flood control water levels. Using safe pre-discharge volume and effective pre-discharge time as sensitive factors, it integrates the empirical distribution of flood frequency estimated by kernel density with theoretical membership functions to establish a membership degree-storage capacity response relationship. By combining ensemble forecasting to quantify forecast uncertainty, it corrects the initial flood control water level, breaks through the limitations of traditional static water level control, and achieves dynamic adaptation of water levels at different flood season stages, balancing flood control safety and water resource utilization efficiency.

[0053] This invention employs fluid-structure coupling (FSI) simulation and concrete damage plasticity model (CDP) as core technologies to construct a three-dimensional finite element model of the dam and gate. It loads dynamic water level conditions and flood discharge vibration loads simulated by CFD, considers the effects of steel corrosion and concrete deterioration, and analyzes the structural stress and displacement response through nonlinear calculations. This accurately verifies the safety of the engineering structure under dynamic water levels and solves the problem of assessment bias caused by traditional static simulation neglecting dynamic loads and material deterioration.

[0054] This invention constructs a two-dimensional hydrodynamic model based on the unsteady flow Saint-Venant equations, combines it with the SWAT distributed hydrological model to generate dynamic boundaries of tributaries in the interval, uses the finite volume method to discretize the solution domain, corrects the river roughness by inverting the Manning formula, simulates the distribution of river level and velocity under different flood discharge conditions, accurately calculates the water surface line to verify the flow capacity, overcomes the limitations of traditional simplified models that ignore interval floods and riverbed deformation, and ensures downstream flood safety under dynamic flood control level schemes.

[0055] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the preferred embodiments of this application are described in detail below with reference to the accompanying drawings.

[0056] The above and other objects, advantages and features of this application will become more apparent to those skilled in the art from the following detailed description of specific embodiments in conjunction with the accompanying drawings. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0058] Based on the description of the figures and their corresponding technical content in the document, the titles of the figures are as follows:

[0059] Figure 1 Flowchart of a phased dynamic flood control water level regulation method applicable to hydropower stations;

[0060] Figure 2 Architecture diagram of a phased dynamic flood control water level regulation system applicable to hydropower stations;

[0061] Figure 3 Flowchart of a phased dynamic flood control water level regulation system applicable to hydropower stations;

[0062] Figure 4 Comparison chart of the rationality of phased flood season plans for hydropower stations (based on historical flood data);

[0063] Figure 5 Comparison of water level lines in the downstream river channel of the hydropower station (under different discharge conditions).

[0064] Figure 6Comparison chart of control accuracy of pre-discharge volume for flood control water level regulation at hydropower stations;

[0065] Figure 7 A trend chart illustrating the long-term operational benefits of phased dynamic flood control water level regulation at a hydropower station. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. In the following description, specific details such as specific configurations and components are provided merely to help fully understand the embodiments of this application. Therefore, those skilled in the art should understand that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. In addition, for clarity and brevity, descriptions of known functions and structures are omitted in the embodiments.

[0067] It should be understood that the phrase "an embodiment" or "this embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "an embodiment" or "this embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0068] Furthermore, reference numerals and / or letters may be repeated in different examples within this application. Such repetition is for the purpose of simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or settings discussed.

[0069] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" in this article describes another type of relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the related objects before and after it are in an "or" relationship.

[0070] In this article, the term "at least one" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, "at least one of A and B" can mean: A exists alone, A and B exist simultaneously, or B exists alone.

[0071] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion.

[0072] Example 1

[0073] This embodiment mainly describes a phased dynamic flood control water level regulation method applicable to hydropower stations, such as... Figure 1 As shown, specifically:

[0074] Acquire multi-source basic data of the watershed where the hydropower station is located, including at least hydrological and meteorological data, engineering operation data, and topographic and geological data;

[0075] The multi-source basic data is preprocessed by data cleaning, noise reduction, and missing value completion to obtain a standardized dataset;

[0076] Based on the standardized dataset, a combination of qualitative and quantitative analysis was used to study the flood season phasing, and the optimal flood season phasing scheme that is suitable for the characteristics of the hydropower station basin was selected.

[0077] Based on the optimal flood season phasing scheme, and combined with the flood control safety constraints and water resource utilization needs of hydropower stations, a dynamic control model for flood limit water level is constructed with safe pre-discharge volume and effective pre-discharge time as sensitive factors.

[0078] Based on the dynamic control model of the flood control water level, the process lines of the flood control water level corresponding to different flood season stages are generated, and the dynamic flood control water level regulation range is determined.

[0079] The safety of the dynamic flood control water level control threshold is verified, and a phased dynamic flood control water level control scheme for hydropower stations that meets the requirements of flood control safety and efficient water resource utilization is output to control the flood control water level of hydropower stations at different flood season stages.

[0080] Furthermore, the acquisition of multi-source basic data for the watershed where the hydropower station is located is achieved by constructing a multi-dimensional data collection system covering hydrology, engineering conditions, and topography. In the hydrological and meteorological data collection phase, a multi-site collaborative collection mechanism needs to be established to collect historical rainfall data, flood process data, and evaporation data from all hydrological and meteorological stations within the watershed. Historical rainfall data must cover core characteristic parameters at the ten-day scale, including the number of days with heavy rainfall in a ten-day period (counting the number of days with daily rainfall ≥ 50 mm), the maximum daily rainfall in a ten-day period (selecting the maximum single-day rainfall value for each ten-day period), and the maximum three-day rainfall in a ten-day period (selecting the maximum cumulative rainfall value for three consecutive days for each ten-day period). Simultaneously, daily rainfall time-series data and rainfall intensity process data from each station must be collected synchronously. Flood process data must include key elements such as peak flow, peak occurrence time, flood duration, and periodic flood volume (e.g., 1-day, 3-day, and 7-day flood volume) for each year's floods. Evaporation data must distinguish between water surface evaporation and land surface evaporation, using evaporation pan observation data combined with conversion factors to obtain the actual evaporation amount, ensuring that the data fully reflects the seasonal variation patterns of the watershed's hydrological and meteorological conditions.

[0081] In the engineering operation and topographic and geological data collection phase, specialized data collection was conducted on the core facilities of the hydropower station and the topography of the watershed. Regarding engineering operation data, in addition to collecting reservoir water level-area-capacity relationship data (covering the entire range from dead water level to normal storage level, with a set of area and capacity data corresponding to every 1m water level interval), inflow and outflow data (including daily average flow, instantaneous flow, and time series), and water level-discharge relationship data (testing was required for different gate combinations to establish a corresponding curve between water level and discharge flow), detailed parameters of gate operation were also collected, including the correspondence between gate opening and discharge capacity. The system includes data on discharge rate (tested every 5% opening interval), gate opening and closing time, and gate operation fault records. For topographic and geological data, airborne LiDAR technology is used to acquire high-precision gridded DEM data of the downstream river channel. During acquisition, the point cloud density must be controlled to be ≥20 points / ㎡ to ensure that the elevation accuracy meets the modeling requirements. Simultaneously, 1:500 measured topographic data (including river channel cross-section and longitudinal section measurements, with cross-section spacing ≤50m) is combined to correct and supplement the DEM data, forming a complete topographic and geological dataset containing watershed topography, geological structures (such as fault and weak interlayer distribution), and river channel morphology.

[0082] Furthermore, the multi-source basic data undergoes preprocessing. A four-step process—cleaning, denoising, completion, and standardization—eliminates quality defects and creates a unified standard dataset, providing reliable input for subsequent quantitative analysis and modeling. In the data cleaning stage, a parallel data processing platform is built using the Spark distributed framework. Differentiated cleaning rules are set for different data types: For hydrological and meteorological data, reasonable value ranges are defined based on historical watershed data (e.g., the upper limit for rainfall is set at a 100-year return period, and the upper limit for flow is set at 1.2 times the historical maximum flood peak flow). Outliers exceeding these ranges are identified and removed using box plots. For engineering operation data, the logical consistency of the water level-reservoir capacity relationship (ensuring a monotonically increasing relationship between water level and reservoir capacity) and the consistency of the water level-discharge relationship (discharge deviation ≤5% at different times under the same water level) are checked. Inconsistent data is traced and corrected. For topographic and geological data, isolated elevation points caused by equipment errors (e.g., discrete points with an elevation difference >2m from surrounding points) are removed to ensure the continuity of topographic data.

[0083] In the data denoising and completion stage, specialized processing methods are adopted for the noise characteristics and missing patterns of different data types: For high-frequency interference in hydrological and meteorological data (such as instantaneous fluctuations in rainfall caused by instrument vibration, and jumps in flow data caused by electromagnetic interference), wavelet denoising algorithm is used for processing. The db4 wavelet basis function is selected to decompose the data into three levels. After removing high-frequency noise signals by setting a threshold, the data is reconstructed, which improves the smoothness of the processed data by ≥30%; For missing hydrological sequence data (such as missing daily rainfall due to equipment failure, and missing hourly flow due to communication interruption), a completion model is built based on generative adversarial network (GAN): the generator of GAN adopts LSTM network structure to learn hydrological sequences. To address temporal correlations (such as the seasonality of rainfall and the lag in flow), the discriminator employs a CNN network structure to distinguish between real and generated data. Through over 500 rounds of adversarial training, the error between the completed data and the original data is reduced to ≤10%. In the data standardization stage, topographic and geological data undergo coordinate unification (converting all data to the National Geodetic Coordinate System 2000 and the National Height Datum 1985) and format conversion (unifying DEM data to TIFF format and vector data to SHP format). Simultaneously, hydrological and meteorological data are normalized (converting rainfall, flow, and other data to standardized values ​​in the [0,1] interval) to ensure that data from different sources and of different types can meet the format and accuracy requirements for subsequent modeling.

[0084] Furthermore, the proposed flood season grading based on standardized datasets constructs a three-tiered grading research system: qualitative guidance, quantitative calculation, and scheme selection. Through multi-method coupling and rationality verification, the grading scheme is ensured to adapt to the characteristics of the watershed. In the qualitative analysis stage, causal analysis is conducted by combining the watershed's climate background and weather system evolution patterns, analyzing the watershed's climate type (e.g., subtropical monsoon climate) and clarifying the seasonal distribution characteristics of precipitation (e.g., concentrated precipitation in summer due to the influence of the southeast monsoon). Secondly, the main weather systems affecting watershed precipitation, such as frontal rain, typhoon rain, and orographic rain, are studied for their occurrence time, impact range, and precipitation intensity, identifying the seasonal variation trend of flood season precipitation (e.g., early influence from frontal rain and later influence from typhoon rain). Finally, the response patterns of rainfall and runoff are analyzed by combining the characteristics of the watershed's underlying surface (e.g., topographic relief and vegetation cover), providing physical causal support for quantitative grading and avoiding a disconnect between grading results and actual hydrological patterns.

[0085] In the quantitative analysis phase, multiple staging schemes were generated through collaborative calculation using multiple algorithms, and the optimal scheme was selected using scientific methods. First, four core staging characteristic indicators were extracted based on a standardized dataset: flow indicators (maximum daily flow and average daily flow over ten days), rainfall indicators (average rainfall over ten days), and flood indicators (time-history distribution of the maximum annual flood peak flow). This ensured that the indicators reflected the hydrological characteristics of the basin's flood season from different dimensions. Second, three quantitative methods were used for staging calculations: fuzzy set analysis was used to construct a dynamic membership function to calculate the membership degree of each day belonging to the flood season throughout the year, using the formula: (in For the actual sample value, The sample mean. (For the sample size), the period boundary is determined by the membership mutation point; Fisher's optimal segmentation method divides the time series samples into... Data points, assuming they are divided into... Segment, calculate the mean of the samples in each segment. (in , The first (segment start and end indices), then through the formula Calculate the sum of squared deviations within the sample segment and construct the objective function based on the BIC criterion. Select the minimum value of the objective function corresponding to The value is used as the optimal number of stages; the system clustering method uses Euclidean distance to calculate sample similarity, and performs cluster merging through the Ward method to generate multiple stages; the Bayesian optimization method and fuzzy preference function are used to select the best stage: based on bootstrap sampling (repeated sampling 1000 times), the relative flood frequency and confidence interval generalized distance of different stages are calculated, and the rationality of each stage is quantitatively scored by the fuzzy preference function. The stage with the highest preference is selected as the optimal flood season stage stage, ensuring the scientificity and applicability of the stage results.

[0086] Furthermore, the construction of the dynamic control model for flood control water levels specifically involves: integrating the core technologies of fuzzy mathematics, joint distribution, and response function; using safe pre-discharge volume and effective pre-discharge time as sensitive factors to construct a dynamic control model that accurately reflects changes during the flood season; and in the fuzzy membership system construction stage, defining the domain as the annual time series based on fuzzy set theory. Establish the membership function of the fuzzy set during the flood season. :in For the first The degree of membership of the sky to the flood season (range of values) The closer the value is to 1, the higher the likelihood that the day belongs to the flood season. The shape parameter controls the steepness of the membership function curve. The larger the curve, the steeper it is, and the shorter the transition period during the flood season. The midpoint of the flood season (e.g., the middle date of the main flood season); the maximum likelihood estimation method is used for... and Calibration was performed by using hydrological data from actual flood season periods over the years as samples to construct a likelihood function and solve for the parameter values ​​corresponding to the maximum values, ensuring that the membership function can accurately fit the actual distribution pattern of the flood season in the basin.

[0087] In the comprehensive membership calculation and joint distribution modeling stage, the empirical distribution curves $f(t)$ of flood occurrence frequency for each time period are first fitted using the kernel density estimation method: taking the flood occurrence dates of previous years as samples, a Gaussian kernel function is used. Construct a kernel density estimation model to calculate the daily flood occurrence frequency throughout the year. Then construct the comprehensive membership degree. ,in Weighting coefficients (range of values) The optimal method was determined through 5-fold cross-validation. The value (to maximize the flood prediction accuracy of the validation set) is used to achieve the fusion of empirical distribution and theoretical membership function; the Copula multivariate joint distribution function is introduced to construct the peak flow ( ) and flood volume ( The joint distribution model of the two is used, and the Clayton Copula function is chosen to describe the nonlinear correlation between them. The formula for the joint distribution function is as follows: (in >0 represents the parameters of the Copula function, calibrated using the maximum likelihood estimation method. , The marginal distribution functions of peak flow and flood volume are respectively (fitted using P-III type distribution); finally, the safe pre-discharge amount ( ) and effective pre-leakage time ( Using as the input variable, establish a membership-capacity response function. (in To design flood control capacity, this function converts the comprehensive membership degree into the flood control capacity for the corresponding time period. Combined with the reservoir water level-capacity curve, the core framework of the dynamic control model for flood control limit water level is formed.

[0088] Furthermore, the generation of the flood control water level process line and the determination of the control threshold specifically involves: based on the optimal phasing scheme and dynamic control model, a three-step process of initial calculation, uncertainty correction, and constraint verification is used to generate the flood control water level process line and control threshold for each flood season stage; in the initial flood control water level calculation stage, according to the optimal flood season phasing scheme, the flood season is divided into four stages: pre-flood season, main flood season, post-flood season, and transition period, and the comprehensive membership degree is calculated for each stage on a ten-day basis. Pre-flood season The value gradually increases from 0 to 0.6-0.7, remains high at 0.8-1.0 during the main flood season, gradually decreases to 0.3-0.4 during the later flood season, and drops to 0 during the transition period; subsequently, by solving... The initial flood control water level is obtained by finding the intersection point with the reservoir capacity curve: Let the reservoir capacity curve be... Through formula The corresponding water level can be obtained by solving the problem. This refers to the initial flood control water level for each time period, forming the initial flood control water level process line.

[0089] In the uncertainty correction and control threshold determination stage, the impact of forecast uncertainty on the pre-release time is quantified. Combining short-term precipitation forecasts (1-7 days) and flood forecasts, ensemble forecast processing techniques are used to generate more than 50 forecast samples, and the effective pre-release time for each sample is calculated. Through formula (in This is the error of a single forecast. The average forecast error, Calculate the standard deviation of the forecast error for the number of forecasts. Determine the effective pre-release time window Based on this time window, the pre-release volume corresponding to different pre-release times is calculated, and the initial flood control level is then corrected: when the pre-release time is extended ( When the initial flood control limit water level is appropriately raised (the increase should be ≤0.5m), and the pre-release time is shortened ( When the flood control limit water level is low, the initial flood control limit water level needs to be lowered (the reduction should be ≤0.8m). The verification is based on the pre-discharge capacity constraint method: the maximum discharge capacity of the gate (determined according to the gate design parameters) and the safe discharge of the downstream river channel (calculated according to the downstream flood control standard) are used as constraints to calculate the maximum allowable pre-discharge at different water levels. If the pre-discharge corresponding to the corrected flood control limit water level exceeds the constraint range, the water level is further adjusted to finally form the flood control limit water level process line (including the control range of the upper and lower limits) for each flood season stage, so as to ensure that the water level control meets both flood control safety and operability.

[0090] Furthermore, the safety verification of the dynamic flood control water level regulation threshold is specifically conducted as follows: A dual verification system of engineering structure simulation and river flow simulation is constructed, and the safety and feasibility of the dynamic flood control water level scheme are verified using specialized numerical calculation methods. In the engineering structure safety verification stage, multi-scale finite element simulation technology is used for specialized analysis: First, an integrated three-dimensional finite element model of the dam, piers, and steel gate is constructed. The dam body uses eight-node hexahedral elements (element size ≤ 2m), and the piers use twenty-node hexahedral elements (element size ≤ 2m). The steel gate (1m) uses four-node shell elements (element size ≤ 0.5m), with the foundation extending to more than twice the dam height. Details such as dam joints, orifices, and reinforcement arrangement are incorporated into the model to ensure accurate structural representation. Dynamic loads are then applied: water loads are calculated based on the upper and lower limits of the dynamic flood control water level control range, and Coulomb's law is used to calculate the dam surface water pressure. Discharge vibration loads are obtained through CFD simulation, and the large eddy simulation (LES) method is used to calculate the flow field under different gate openings. The time history of the pulsating pressure on the gate surface is extracted and calculated using formulas. (in For pulsating pressure time history, For sampling duration, The pressure time history (at frequency) is converted into power spectral density (PSD) as the structural vibration load input. The concrete damage plasticity model (CDP) is used to simulate the deterioration of dam materials. Based on the steel corrosion rate detected on site (calculated by weight loss rate), the reduction of steel section is quantified. Through nonlinear finite element calculation, the dynamic stress (must meet the requirement of concrete compressive strength ≤ 0.8 times the design value), displacement (horizontal displacement ≤ 5 mm, vertical displacement ≤ 3 mm) and acceleration response index of the stress concentration area of ​​the dam (such as the dam heel and dam toe) are extracted. The values ​​are compared with the allowable values ​​in the current specifications to verify the safety of the engineering structure under dynamic water level.

[0091] In the downstream channel flow capacity verification phase, system simulation was conducted based on a two-dimensional hydrodynamic model. This model was constructed using the unsteady Saint-Venant equations as its core, with the continuity equations being... (in Because of the water depth, For time, , They are respectively , directional flow velocity), the momentum equation is $ , (in For lateral inflow, It is the acceleration due to gravity. , They are respectively , Direction bottom slope, , They are respectively , Directional friction slope , They are respectively , Surface shear stress in the direction of direction The density of water, (For water pressure); the solution domain is discretized using the finite volume method, with the grid in the main channel area refined to 20m×20m and the grid in the floodplain area set to 50m×50m to ensure accurate capture of the river flow characteristics; the discharge flow process line corresponding to the dynamic flood control level scheme is used as the upstream boundary input, and the dynamic inflow boundary (lateral boundary) of the tributaries in the interval is generated through a distributed hydrological model (SWAT); the Manning formula is used. (in The Manning roughness coefficient is determined through historical flood surface line inversion in the main river channel. Values ​​range from 0.025 to 0.03, for beach areas. The value ranges from 0.035 to 0.045; For hydraulic radius, The roughness of the river channel is calculated (for hydraulic gradient), and the distribution of water level, flow velocity, and water depth in the downstream river channel is simulated under different flood discharge conditions. The water surface line corresponding to each level of flow is calculated to verify whether the river channel's flow capacity meets the requirements for safe flood discharge (ensuring that the maximum flow velocity is ≤3m / s and the highest water level is ≥0.5m lower than the top elevation of the levee). Finally, a complete safety verification conclusion is formed.

[0092] This implementation details core technologies such as multi-source data fusion preprocessing, multi-method coupling for flood season phasing, fuzzy mathematics and Copula joint distributed modeling, fluid-structure coupled simulation, and hydrodynamic simulation. It constructs a complete phased dynamic flood limit water level control method, which solves the problems of flood resource waste and insufficient water storage caused by traditional static water level scheduling, as well as the shortcomings of single phased methods, such as strong subjectivity and neglect of dynamic loads and material degradation in safety demonstration. It achieves the integration of accurate flood season phasing, dynamic water level adaptation, engineering structure safety verification, and downstream flood safety assurance, balancing flood control safety and efficient water resource utilization.

[0093] Example 2

[0094] This embodiment describes in detail a phased dynamic flood control level regulation system suitable for hydropower stations, including a data acquisition subsystem, a data processing subsystem, a flood season phased subsystem, a dynamic control model subsystem, a regulation threshold generation subsystem, a safety verification subsystem, and a decision output subsystem. Figure 2 As shown, specifically:

[0095] The data acquisition subsystem forms the foundational layer, comprising a hydrological and meteorological acquisition unit, an engineering operation acquisition unit, and a topographic and geological acquisition unit. The hydrological and meteorological acquisition unit collects real-time rainfall, flood, and evaporation data through a sensor network (rain gauges, water level gauges, current meters, etc.) deployed at hydrological and meteorological stations within the basin. The engineering operation acquisition unit interfaces with the hydropower station's SCADA system to acquire operational data such as reservoir water level, inflow and outflow rates, and gate opening. The topographic and geological acquisition unit integrates airborne LiDAR equipment and ground-based surveying instruments to acquire river topography and geological structure data. The three units aggregate data through a data transmission gateway and transmit it via optical fiber to the data processing subsystem.

[0096] The data processing subsystem is deployed on a local server cluster at the hydropower station and consists of a distributed computing module (based on the Spark framework), a data cleaning module, a denoising and completion module, and a standardization module. The distributed computing module receives the raw data from the data acquisition subsystem, distributes it to the cleaning module to remove outliers, then processes noise and missing values ​​through the denoising and completion module (which integrates wavelet denoising algorithms and GAN models), and finally the standardization module completes coordinate unification and format conversion to generate a standardized dataset, which is then transmitted to the flood season grading subsystem and the safety verification subsystem via an internal bus.

[0097] The flood season grading subsystem adopts an architecture of qualitative analysis server and quantitative calculation workstation. The qualitative analysis server stores a database of watershed climate characteristics and weather systems, and generates flood season trend analysis reports through an expert system. The quantitative calculation workstation is equipped with fuzzy set analysis, Fisher optimal segmentation and system clustering algorithm modules. After receiving standardized datasets, it calculates multiple grading schemes, and then selects the optimal scheme through a Bayesian optimization module. The results are synchronized to the dynamic control model subsystem.

[0098] The dynamic control model subsystem is built on high-performance computing nodes and includes a membership calculation module, a Copula joint distribution module, and a response function module. The membership calculation module receives the optimal phased scheme and generates a comprehensive membership degree by combining it with historical flood data. The Copula joint distribution module constructs a joint distribution model of flood peak and flood volume. The response function module converts the membership degree into flood control capacity and transmits it to the control threshold generation subsystem through an interface.

[0099] The control threshold generation subsystem consists of a process line generation module and an uncertainty correction module. The process line generation module generates an initial flood limit water level process line based on the flood control reservoir capacity and the water level-reservoir capacity relationship. The uncertainty correction module calculates the pre-release time window by combining short-term forecast data and corrects the dynamic control threshold value, which is then sent to the safety verification subsystem and the decision output subsystem, respectively.

[0100] The safety verification subsystem includes a structural simulation unit and a hydrodynamic simulation unit. The structural simulation unit is based on finite element analysis software (integrated CDP model) and calculates the stress and strain of the dam and gate after receiving the control threshold. The hydrodynamic simulation unit uses a two-dimensional hydrodynamic model to simulate the flow capacity of the downstream river channel. The verification results of the two are summarized by the safety assessment module and fed back to the control threshold generation subsystem for secondary correction.

[0101] The decision output subsystem consists of a scheme integration module and an instruction generation module. The scheme integration module receives the control domain value that has been verified for safety and generates a phased dynamic flood limit water level scheme. The instruction generation module converts the scheme into gate control instructions and sends them to the hydropower station actuator through the PLC system, while storing the scheme in the historical database.

[0102] Furthermore, the system implementation method flow is as follows: Figure 3 As shown, specifically:

[0103] After the system starts up, it enters the data acquisition phase. The three units of the data acquisition subsystem operate synchronously: the hydrological and meteorological acquisition unit collects rainfall, water level, and flow velocity data of each station in the basin at 10-minute intervals; the engineering operation acquisition unit acquires SCADA data such as reservoir water level and gate opening in real time; and the topographic and geological acquisition unit updates the river topographic data regularly (every quarter) through LiDAR scanning. All data are uploaded to the data processing subsystem after verification.

[0104] In the data processing stage, the distributed computing module segments the raw data for processing, and the data cleaning module removes outliers (such as rainfall data exceeding the range) based on preset thresholds (such as the 95% confidence interval of historical data). The denoising and completion module applies wavelet denoising to the hydrological sequence (decomposing it into 3 layers and reconstructing it) and uses a GAN model to complete missing data (the training samples are complete sequences from the past 10 years). The standardization module converts the topographic data into a unified coordinate system and normalizes the hydrological data to the [0,1] interval. The generated standardized dataset is simultaneously pushed to the flood season grading subsystem and the safety verification subsystem.

[0105] The flood season phasing process is divided into two steps: qualitative and quantitative. The qualitative analysis server calls the watershed climate database to analyze the seasonal distribution of precipitation under the influence of the monsoon and generates a trend report showing that "the pre-flood season is dominated by frontal rain and the main flood season is affected by typhoon rain." The quantitative calculation workstation extracts indicators such as the maximum flow rate and the average rainfall over ten days, and runs fuzzy set analysis (calculating membership change points), Fisher optimal segmentation (minimizing intra-segment deviation), and system clustering (Ward method to merge samples) to obtain three phasing schemes. The Bayesian optimization module selects the scheme with the highest flood frequency matching degree through 1,000 self-sampling tests and outputs it to the dynamic control model subsystem.

[0106] In the dynamic control model construction phase, the membership calculation module, based on the optimal phasing scheme, calibrates the fuzzy membership function parameters through maximum likelihood estimation and generates a comprehensive membership curve by combining the flood frequency distribution estimated by kernel density. The Copula joint distribution module selects the Clayton function to fit the joint distribution of flood peak and flood volume. The response function module associates the comprehensive membership with the design flood control capacity to obtain the flood control capacity for each time period and transmits it to the control threshold generation subsystem.

[0107] During the control threshold generation stage, the process line generation module calculates the initial flood limit water level process line (distinguishing between pre-flood, main flood, post-flood, and transition periods) based on the flood control reservoir capacity and the measured water level-reservoir capacity curve; the uncertainty correction module accesses the precipitation forecast data for the next 7 days, generates 50 sets of pre-release time samples using ensemble forecasting, calculates the error standard deviation to determine the time window, corrects and obtains the control threshold value including upper and lower limits, and simultaneously sends it to the safety verification subsystem;

[0108] During the safety verification phase, two verifications were carried out in parallel: the structural simulation unit constructed a three-dimensional finite element model of the dam-gate, loaded the water load and flood discharge vibration load corresponding to the control threshold, and calculated the concrete stress (control value ≤20MPa) and displacement (control value ≤5mm); the hydrodynamic simulation unit simulated the river water level under different discharge rates (ensuring it was more than 0.5m below the top of the levee) based on the Saint-Venant equations. After both results met the specifications, they were fed back to the decision output subsystem.

[0109] The scheme integration module organizes the verified control thresholds into a phased dynamic flood control water level scheme (including the upper and lower limits of water levels for each stage and the division of time periods); the instruction generation module generates gate opening and closing instructions at an hourly time granularity, controls the gate operation through the PLC system, and simultaneously archives the scheme to the database, completing one control process. The system repeats the above process on a daily cycle to achieve dynamic adjustment of the flood control water level.

[0110] This embodiment details how distributed computing improves data processing efficiency, specialized simulation units ensure safety verification accuracy, and decision output modules enable command implementation, thereby achieving automation, intelligence, and closed-loop control of flood control water levels in hydropower stations. This avoids the lag and errors of manual scheduling and improves the operability and reliability of the scheduling scheme.

[0111] Based on Embodiment 1 or 2, this embodiment details the specific implementation effect of phased dynamic flood control water level regulation applicable to hydropower stations. Taking the Shangyoujiang Hydropower Station as a specific application example, this hydropower station is located in the middle reaches of the Shangyoujiang River, with a controlled catchment area of ​​2750 km² above the dam site. It is a comprehensive water conservancy project primarily for power generation, while also serving flood control purposes. Since its operation began in 1955, it has long adopted a static flood control water level regulation mode. During the main flood season (April 1st to June 20th), the flood control water level is fixed at 195.50 m, and during the post-flood season (June 21st to August 31st), it is fixed at 197.70 m. This mode leads to frequent flood discharge during the main flood season (annual average discharge volume). (Approximately 42 million m³), ​​the hydropower station often faces insufficient water storage during the post-flood season due to difficulties in identifying the last flood (an average annual power generation gap of approximately 15 million kWh). Simultaneously, in high-water years, the water level at some downstream river sections approaches the top elevation of the dikes (102.5 m), posing a risk of flooding. To address these issues, this embodiment selects three complete hydrological years from 2022 to 2024 (2022, a normal water year; 2023, a high-water year; and 2024, a low-water year), and conducts experiments using the technical solution proposed in this application. The technical effectiveness is verified by combining measured data. All data comes from the hydropower station dispatch system ledger, on-site measurements, and simulation experiments to ensure authenticity and reliability.

[0112] like Figure 4 As shown, regarding the rationality of the phased flood season plan in this application, Figure 4It contains three broken lines, which correspond to the multi-method coupling scheme of this application (fuzzy set analysis method + Fisher optimal segmentation method + system clustering method + Bayesian optimization), traditional fuzzy set scheme, and traditional Fisher optimal segmentation scheme, respectively. The data is based on the statistical calculation of 40 years of historical flood data of hydropower stations from 1985 to 2024 (128 measured floods, including key parameters such as flood peak occurrence time and flood duration). During the experiment, the proposed scheme first clarified the climate pattern of "the basin's flood season being dominated by frontal rain (April-May) and typhoon rain (June-August)" through qualitative analysis. Then, based on standardized datasets, it extracted periodic characteristic indicators such as the maximum daily flow and average rainfall over ten days. It used three quantitative methods to calculate multiple periodic schemes. Finally, it used Bayesian optimization to select the optimal scheme for "pre-flood season (April 1-May 20), main flood season (May 21-July 31), and post-flood season (August 1-August 31)". Traditional schemes only use a single fuzzy set or Fisher method and do not combine climate patterns with multiple scheme verification. As shown in the attached figures, the proposed scheme has a lower overall accuracy than the traditional scheme, with an average periodization error rate of only 8.2%. Specifically, the error rate during the main flood season (corresponding to ten-day periods 16-24, May 21-July 31) is as low as 5.3%, while the traditional fuzzy set scheme has an average error rate of 18.7% and a main flood season error rate of 22.1%, and the traditional Fisher optimal segmentation scheme has an average error rate of 15.4% and a main flood season error rate of 17.8%, representing a reduction in error of over 60%. Taking a typhoon-induced flood on July 5, 2023, as an example, this flood actually occurred during the main flood season. The traditional fuzzy set scheme mistakenly classified it as a post-flood season (error rate 28.3%), while the proposed scheme accurately classified it as part of the main flood season (error rate 3.1%). This fully demonstrates that the proposed scheme, through multi-method coupling and rationality verification, achieves significantly higher periodization accuracy than the traditional scheme, accurately matching the seasonal patterns of floods in the basin and providing a reliable spatiotemporal boundary for subsequent water level regulation.

[0113] like Figure 5 As shown, for the downstream river water level, Figure 5The data includes four curves, corresponding to the proposed scheme (discharge rate 300 m³ / s), the proposed scheme (discharge rate 500 m³ / s), the traditional scheme (discharge rate 300 m³ / s), and the traditional scheme (discharge rate 500 m³ / s). The data are derived from the actual river topography measured by airborne LiDAR in May 2024 (elevation accuracy ±0.15 m) and the simulation results of the MIKE21 hydrodynamic model. The model parameters are calibrated based on the measured flood data from three events in 2023, and the Nash efficiency coefficient NSE = 0.92 to ensure the reliability of the simulation. The experiment simulated a typical flood discharge scenario during the high-water year of 2023, where 300 m³ / s was the usual discharge volume and 500 m³ / s was a relatively large discharge volume. The proposed scheme is based on the dynamic flood limit water level control range (195.00-196.00 m during the main flood season) and dynamically adjusts the discharge volume in combination with real-time data of flood inflow in the interval. The traditional scheme discharges floodwater at a fixed flood limit water level (195.50 m during the main flood season) without considering the changes in water level and riverbed scouring and sedimentation in the interval. As can be clearly seen from the figure, the water level of the proposed scheme is lower than that of the traditional scheme throughout, and the safety margin from the top elevation of the dike is larger: when the discharge is 300 m³ / s, the water level at section 5 (located 8 km downstream) is 101.8 m for the proposed scheme and 102.1 m for the traditional scheme. The safety margin of the proposed scheme is 0.7 m, while that of the traditional scheme is only 0.4 m, a difference of 0.3 m. When the discharge increases to 500 m³ / s, the water level at section 7 (located 12 km downstream) is 102.0 m for the proposed scheme and 102.5 m for the traditional scheme (level with the top elevation of the dike). The proposed scheme still retains a safety margin of 0.5 m, completely avoiding the risk of overflow of the traditional scheme under a large discharge, and verifying the effective guarantee of downstream flood safety of the proposed dynamic control scheme.

[0114] like Figure 6 As shown, regarding the control accuracy of the pre-leakage volume, Figure 6The graph contains two sets of scatter plots: blue scatter plots represent the dynamic control model of this application, and red scatter plots represent the traditional pre-release model, totaling 100 sets of data. All data are from pre-release experiments conducted during 40 floods in the 2023-2024 flood season. The graph also marks the "ideal control line (y=x)," indicating that the actual pre-release volume is completely consistent with the theoretical pre-release volume. The average deviation rate between the two sets of scatter plots was calculated to quantify the difference in accuracy. During the experiments, the model of this application integrated ensemble forecasting technology to generate 50 sets of short-term precipitation forecast samples and calculated the forecast error standard deviation Δt=2.1h. A dynamic pre-release strategy was constructed using "safe pre-release volume" and "effective pre-release time" as sensitive factors to adjust the pre-release rhythm in real time. The traditional model, on the other hand, calculates the pre-release volume based on a fixed 48-hour lead time, without considering the impact of forecast uncertainty on the pre-release effect. As shown in the attached figures, the blue scatter points of this application are more concentrated near the ideal control line, and the dispersion is significantly lower than that of the red scatter points of the traditional model. The average pre-discharge deviation rate of the model in this application is only 6.8%, with 85% of the scatter points falling within the ±10% deviation range. In contrast, the average deviation rate of the traditional model is as high as 18.5%, with only 45% of the scatter points falling within the ±10% deviation range. Taking a flood on June 12, 2024 as an example, the theoretical pre-discharge was 5 million m³. The actual pre-discharge of the model in this application was 4.72 million m³, with a deviation rate of 5.6%. The actual pre-discharge of the traditional model was only 3.21 million m³, with a deviation rate of 35.8%. The scheme in this application effectively avoids the risk of reservoir water levels exceeding the warning level due to insufficient pre-discharge in the traditional model. This fully demonstrates that by quantifying forecast uncertainty and optimizing pre-discharge rules, this application significantly improves the accuracy of pre-discharge control and provides a reliable operational basis for dynamic flood control level regulation.

[0115] like Figure 7 As shown, for long-term operational benefits, Figure 7The data includes two line graphs: the blue line represents the quarterly water discharge under this application scheme, and the red line represents the quarterly power generation under this application scheme. The data are all from the hydropower station's operation ledger from 2022 to 2024 (including daily measured data on water discharge and power generation, summarized and statistically analyzed on a quarterly basis). The historical average values ​​of the same period from 2019 to 2021 for the traditional scheme are also marked as a comparison benchmark. During the experiment, the flood control water level was dynamically adjusted according to the characteristics of different hydrological years: in 2022 (normal water year), the water level was controlled at 195.00-195.80m during the main flood season and 197.50-198.00m during the subsequent flood season; in 2023 (high water year), the water level was controlled at 195.00-196.00m during the main flood season and 197.50-198.20m during the subsequent flood season; in 2024 (dry water year), the water level was controlled at 195.20-195.80m during the main flood season and 197.70-198.20m during the subsequent flood season. As shown in the attached figures, the water wastage and power generation of this application show a stable and optimized trend: In terms of water wastage, the maximum quarterly water wastage for 2022-2024 is 12 million m³, 21 million m³, and 8 million m³, respectively, which is 30%-35% lower than the average of the traditional scheme during the same period (18 million m³, 32 million m³, and 12.5 million m³); in terms of power generation, the maximum quarterly power generation for 2022-2024 is 35 million kWh, 42 million kWh, and 28 million kWh, respectively, which is lower than the traditional scheme. The proposed scheme increases the average output by 15%-20% compared to the same period (30 million kWh, 35 million kWh, and 25 million kWh). Meanwhile, the quarterly fluctuations in water wastage and power generation in this application scheme are reduced by 25% and 18% respectively compared to the traditional scheme. This demonstrates that even under the influence of interannual hydrological variations (normal, high, and low water years), the technical effect of this application remains stable and can achieve a long-term balance between flood control safety and efficient water resource utilization. It effectively solves the long-standing problems of water wastage and insufficient power generation at the Shangyoujiang Hydropower Station.

[0116] This embodiment details the practicality and stability of the technical solution verified through actual application. It solves the pain points of water waste, low power generation efficiency, and high flood risk in traditional scheduling. It achieves the goals of improving the accuracy of flood season phased scheduling under different hydrological years, optimizing the accuracy of pre-discharge control, and ensuring stable long-term operation benefits. At the same time, it ensures the safety of the engineering structure under dynamic water levels and the compliance of downstream river flow capacity. It provides a reusable technical paradigm and system framework for the optimized scheduling of flood control water levels of similar hydropower stations.

[0117] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any changes, modifications, substitutions, integrations, and parameter changes made to these embodiments within the spirit and principles of the present invention, without departing from the principles and spirit of the present invention, through conventional substitutions or to achieve the same function, fall within the scope of protection of the present invention.

Claims

1. A method for phased dynamic flood control water level regulation applicable to hydropower stations, characterized in that, include: Acquire multi-source basic data of the watershed where the hydropower station is located, including at least hydrological and meteorological data, engineering operation data, and topographic and geological data; The multi-source basic data is preprocessed by data cleaning, noise reduction, and missing value completion to obtain a standardized dataset; Based on the standardized dataset, a combination of qualitative and quantitative analysis was used to study the flood season phasing, and the optimal flood season phasing scheme that is suitable for the characteristics of the hydropower station basin was selected. Based on the optimal flood season phasing scheme, and combined with the flood control safety constraints and water resource utilization needs of hydropower stations, a dynamic control model for flood limit water level is constructed with safe pre-discharge volume and effective pre-discharge time as sensitive factors. Based on the dynamic control model of the flood control water level, the process lines of the flood control water level corresponding to different flood season stages are generated, and the dynamic flood control water level regulation range is determined. The safety of the dynamic flood control water level control threshold is verified, and a phased dynamic flood control water level control scheme for hydropower stations that meets the requirements of flood control safety and efficient water resource utilization is output to control the flood control water level of hydropower stations at different flood season stages.

2. The method for phased dynamic flood control water level regulation applicable to hydropower stations according to claim 1, characterized in that, The acquisition of multi-source basic data for the watershed where the hydropower station is located includes: Collect historical rainfall data, flood event data, and evaporation data recorded by hydrological and meteorological stations within the basin; Collect data on the relationship between reservoir water level, area, and capacity, inflow and outflow data, water level and discharge data, and the relationship between gate opening and discharge capacity during the operation of the hydropower station. Airborne LiDAR technology was used to acquire grid DEM data of the downstream river channel, which was then integrated with measured topographic data to form a watershed topographic and geological dataset.

3. The method for phased dynamic flood control water level regulation applicable to hydropower stations according to claim 1, characterized in that, The preprocessing of multi-source basic data includes: The Spark distributed framework is used to clean multi-source basic data and remove abnormal data that exceeds the reasonable range of values. High-frequency interference factors in hydrological and meteorological data are eliminated by wavelet denoising algorithm; Complete missing hydrological sequence data using generative adversarial networks; Coordinate unification and format conversion operations were performed on topographic and geological data to ensure that topographic data from different sources met the accuracy requirements for subsequent modeling.

4. The method for phased dynamic flood control water level regulation applicable to hydropower stations according to claim 1, characterized in that, The study on flood season grading, based on standardized datasets, employs a combination of qualitative and quantitative analysis. Specifically, the qualitative analysis stage identifies the seasonal variation trend of precipitation during the flood season by combining the basin's climate characteristics and the evolution patterns of weather systems. The quantitative analysis stage extracts grading characteristic indicators based on standardized datasets, including the temporal distribution of the maximum daily flow, average daily flow, average rainfall, and maximum annual flood peak flow. At least three different flood season grading schemes were obtained by using fuzzy set analysis to construct dynamic membership functions, Fisher's optimal segmentation method to calculate the sum of squared deviations within sample segments, and system clustering method to carry out multi-index clustering. Using Bayesian optimization and fuzzy preference functions, and based on bootstrapping sampling, the relative flood frequency and generalized distance of confidence intervals for different flood season phasing schemes are calculated. The optimal flood season phasing scheme is then selected, where the generalized distance... The formula is: ,in For the actual sample value, The sample mean. This represents the number of samples.

5. The method for phased dynamic flood control water level regulation applicable to hydropower stations according to claim 4, characterized in that, The Fisher optimal segmentation method is used to calculate the sum of squared deviations within the sample segment, specifically as follows: The time series samples corresponding to the staged characteristic indicators are divided into: Let the data points be divided into 10 data points. Section, of which Segment sample Calculate the mean of each sample segment, expressed as: ; The formula for calculating the sum of squared deviations within a sample segment is: ,in The objective function is the sum of the squared deviations of each segment; the objective function is constructed using the BIC criterion. Plot the curve of the objective function as a function of the number of categories, and select the number of categories corresponding to the inflection point of the curve as the optimal number of stages; Principal component analysis, entropy weight method, and coefficient of variation method were used to determine the weights of characteristic indicators for each stage, where the coefficient of variation... The formula is: ,in The standard deviation of the indicator. The mean of the indicators is used to adjust the sum of squared deviations based on the weights, and the split points are optimized.

6. The method for phased dynamic flood control water level regulation applicable to hydropower stations according to claim 1, characterized in that, The construction of the dynamic control model for flood control water level, which uses safe pre-discharge volume and effective pre-discharge time as sensitive factors, includes: Based on fuzzy set theory, the domain is defined as the entire year's time series. The membership function of the fuzzy set for the flood season is established, and the formula is: ,in For the first Membership degree during the flood season For shape parameters, This is the central moment of the flood season; The empirical distribution curves of flood occurrence frequency for each time period were fitted using the kernel density estimation method. Construct a comprehensive membership degree The formula is ,in These are the weighting coefficients; By introducing the Copula multivariate joint distribution function, a joint distribution model of peak flow and flood volume is constructed. Let the peak flow be... , Flood volume Then the joint distribution function expression is: ,in For Copula functions, , These are the marginal distribution functions of peak flow and flood volume, respectively; Using safe pre-discharge and effective pre-discharge time as input variables, a membership-capacity response function is established, expressed as follows: ,in To design flood control reservoir capacity, a dynamic control model for flood control limit water level was developed.

7. The method for phased dynamic flood control water level regulation applicable to hydropower stations according to claim 6, characterized in that, The introduction of the Copula multivariate joint distribution function to construct a joint distribution model of peak flow and flood volume includes: Peak flow samples and corresponding flood volume samples for each flood season stage are extracted from the standardized dataset to form a two-dimensional sample set. The Clayton Copula function is used to construct the joint distribution function of peak flow and flood volume. The expression of the Clayton Copula function is as follows: ,in >0 represents the parameters of the Copula function, which are estimated using the maximum likelihood estimation method. The likelihood function expression is: ,in The density function of the Clayton Copula function; The peak-volume combination values ​​under different return periods were calculated based on the joint distribution function and compared with historical measured flood data. The goodness of fit of the joint distribution model was verified by the KS test, and the KS statistic was calculated according to the following formula: ,in Let be the empirical distribution function. The theoretical distribution function is used to ensure the accuracy of the model's representation of the statistical characteristics of floods in different stages.

8. The method for phased dynamic flood control water level regulation applicable to hydropower stations according to claim 1, characterized in that, The process of generating flood control water level process lines corresponding to different flood season stages and determining the dynamic flood control water level control threshold is specifically as follows: Based on the optimal flood season phasing scheme, the flood season is divided into the pre-flood season, the main flood season, the post-flood season, and the transition period. The comprehensive membership degree is calculated for each stage of the flood season. ; By solving the comprehensive membership degree The intersection points with the reservoir capacity curve yield the initial flood control water level values ​​for each time period. Let the reservoir capacity curve be... Through formula Solve for the corresponding water level The initial flood control limit water level was obtained; By combining short-term precipitation forecasts with flood forecasts, ensemble forecasting techniques are used to quantify forecast uncertainties and determine the effective pre-release time window. ,in For effective pre-release time, The standard deviation of the forecast error is given by the formula: ,in This is the error of a single forecast. The average forecast error, Number of forecasts; Based on the pre-discharge capacity constraint method, the initial flood limit water level is modified by using the maximum discharge capacity of the gate and the safe discharge of the downstream river channel as constraints, thus forming the flood limit water level process line and dynamic control domain value for each flood season stage.

9. The method for phased dynamic flood control water level regulation applicable to hydropower stations according to claim 1, characterized in that, The safety verification of the dynamic flood control water level regulation threshold is achieved by combining finite element simulation analysis of the stress and deformation characteristics of the engineering structure with hydrodynamic model verification of the downstream river channel flow capacity. Specifically: Parametric geometric reconstruction technology was used to construct a three-dimensional finite element model of the dam, gate piers and steel gates. The dam body was constructed using eight-node hexahedral elements and the gates were constructed using six-degree-of-freedom flat plate elements. Using different water level conditions and gate opening conditions corresponding to the dynamic flood control water level regulation threshold as input parameters, water load and flood discharge excitation load are applied. The flood discharge excitation load is obtained by computational fluid dynamics simulation of pulsating pressure distribution under different opening degrees, and the pulsating pressure power spectral density is used. The formula is: ,in For pulsating pressure time history, For sampling duration, For frequency; A concrete damage plasticity model was used to simulate the material degradation of the dam body. Based on the measurement results of the steel corrosion rate, the reduction of the steel section was quantified. Nonlinear finite element calculations were carried out, and the stress-strain relationship of the concrete satisfied... ,in The initial elastic modulus, In response, For damage variables; The dynamic stress, displacement, and acceleration response indices of the stress concentration areas of the dam body were extracted and compared with the allowable values ​​in the current specifications.

10. The method for phased dynamic flood control water level regulation applicable to hydropower stations according to claim 9, characterized in that, The verification of downstream river channel flow capacity using a hydrodynamic model includes: A two-dimensional hydrodynamic model is constructed using the unsteady flow Saint-Venant equations, where the continuity equations of the Saint-Venant equations are expressed as follows: The momentum equations are as follows: , ,in Because of the water depth, For time, , They are respectively , directional flow velocity, For lateral inflow, It is the acceleration due to gravity. , They are respectively , Direction bottom slope, , They are respectively , Directional friction slope , They are respectively , Surface shear stress in the direction of direction The density of water, Water pressure; The solution domain is discretized using the finite volume method. The discharge flow process line corresponding to the dynamic flood control water level regulation scheme is used as the upstream boundary input. The dynamic inflow boundary of the tributary in the interval is generated by the distributed hydrological model and used as the lateral boundary. The Manning formula is used to calculate the channel roughness. The Manning formula is expressed as follows: ,in For flow rate, The roughness coefficient is Manning's coefficient. For hydraulic radius, The model parameters were calibrated using hydraulic gradient data and historical flood scour data. The water level, flow velocity and water depth distribution of the downstream river under different flood discharge conditions were simulated, and the water surface line corresponding to each level of flow was calculated to verify whether the river's flow capacity meets the scheduling requirements.

Images