Method for calculating dynamic water storage capacity of reservoir based on hydrology and random forest model

Abstract

This invention provides a method for calculating the dynamic rainwater carrying capacity of reservoirs based on hydrological and random forest models, relating to the field of hydrological technology for small watersheds. This method includes the following steps: determining the maximum target water level of the reservoir; determining parameters; constructing a small reservoir rainwater carrying capacity prediction model based on a random forest regression algorithm, selecting rainfall characteristic parameters, watershed characteristic parameters, and reservoir characteristic parameters; determining evaluation indicators for the prediction model; and calculating and correcting the rainwater carrying capacity hourly. This method proposes four methods for calculating rainwater carrying capacity by constructing a hydrological model and a prediction model for small reservoir rainwater carrying capacity. It can realize the calculation and correction of rainwater carrying capacity during one or more rainfall events, reducing errors caused by rainfall-runoff processes and reservoir scheduling processes, and predicting the reservoir's rainwater carrying capacity corresponding to the rainfall amount at any future rainfall time.

Description

Technical Field

[0001] This invention relates to the field of hydrological technology for small watersheds, specifically a method for calculating the dynamic rainwater carrying capacity of reservoirs based on hydrological and random forest models. Background Technology

[0002] There are many methods for calculating the rainwater carrying capacity of a reservoir, mainly divided into empirical formula method, runoff coefficient method and hydrological model analysis method. The empirical formula method, based on extensive historical and measured data of reservoirs, estimates the reservoir's rainfall carrying capacity using simplified formulas. These formulas typically only consider factors such as catchment area, rainfall intensity, and soil type, without taking into account the dynamic changes in inflow and outflow. Furthermore, the parameters in the formulas are simplified, making parameter values ​​easy to determine and calculations quick and easy. However, the biggest drawback is the high degree of arbitrariness and low accuracy in the calculated results. The runoff coefficient method, based on homogeneity and steady-state assumptions, derives the reservoir's rainfall carrying capacity through empirical formulas. These formulas only consider the reservoir's initial rainfall carrying capacity, catchment area, and runoff coefficient. This method is simple to use and has low data requirements, but its accuracy is limited, ignoring the complexity of hydrological processes. The hydrological model analysis method, by constructing watershed runoff generation and confluence models and reservoir scheduling models, fully considers the interactions and physical mechanisms between hydrological elements. However, due to the large number of parameters considered, the calculation process is relatively complex, although the calculation accuracy is improved. This method is mainly used for typical analyses of design floods or specific rainfall events, but its ability to simulate and model random rainfall events is still insufficient.

[0003] With the development of hydrological models and computer technology, machine learning algorithms have been widely introduced into hydrological research and combined with hydrological models. Compared with traditional hydrological methods, machine learning can quickly process large amounts of data and accurately identify the correlations between data. Therefore, many researchers have conducted extensive research and attempts to use machine learning algorithms for hydrological forecasting and early warning calculations. Lin Wenxiao et al. constructed a hydropower station reservoir water level prediction model based on big data and machine learning technology, providing technical support for medium- and long-term reservoir water level prediction; Wen Kaixiang et al. constructed a reservoir water depth inversion model based on machine learning and satellite remote sensing data, using raw water depth data as the data source, which can quickly determine the reservoir capacity under different conditions; Li Fuchen et al. selected relevant characteristic parameters such as hydrology, underlying surface, and riverside villages, and the calculation results of critical rainfall to construct an extreme gradient lifting algorithm for predicting critical rainfall for different warning periods, predicting the critical rainfall for different warning periods. Summary of the Invention

[0004] (a) Technical problems to be solved

[0005] To address the shortcomings of existing technologies, this invention provides a method for calculating the dynamic rain-carrying capacity of reservoirs based on hydrological and random forest models, which improves the accuracy of rain-carrying capacity analysis and calculation for small reservoirs and provides data support for flood forecasting in reservoir basins.

[0006] (II) Technical Solution

[0007] To achieve the above objectives, the present invention provides a method for calculating the dynamic rainwater carrying capacity of reservoirs based on hydrological and random forest models, comprising the following steps:

[0008] Determine the maximum target water level of the reservoir: Based on the reservoir operation and management plan and on-site investigation, determine the maximum target water level of the reservoir;

[0009] Parameter determination: Rainfall data with different frequencies and durations are input into the HEC-HMS hydrological model for runoff generation and confluence calculation to obtain the runoff generation and confluence processes under different soil moisture contents, rainfall frequencies, and rainfall durations, i.e., the reservoir inflow flood process; a reservoir flood control model is established based on the water balance equation and the reservoir's water level-capacity-discharge curve, and the reservoir water level and its corresponding capacity are calculated according to different initial water levels; based on the reservoir's maximum target water level and water level-capacity curve, the reservoir's rain-carrying capacity is calculated under different soil moisture contents, rainfall frequencies, rainfall durations, and initial water levels for a single rainfall event, and a hydrological model of small reservoir rain-carrying capacity is constructed.

[0010] Based on the random forest regression algorithm, a rain-carrying capacity prediction model for a small reservoir is constructed by selecting rainfall characteristic parameters, watershed characteristic parameters, and reservoir characteristic parameters, and the rain-carrying capacity of the small reservoir under any characteristic parameters is calculated.

[0011] Determine the evaluation indicators for the prediction model: Select the mean square error (MSE), coefficient of determination (r²), and mean absolute error (MAE) to objectively evaluate the prediction model for the rainwater carrying capacity of small reservoirs;

[0012] Rainfall carrying capacity calculation and correction on an hourly basis: Based on the rainfall carrying capacity prediction model and the rainfall carrying capacity hydrological model of small reservoirs, the rainfall carrying capacity is calculated for complete, uninterrupted rainfall, the rainfall carrying capacity is calculated for each time period during the rainfall process with the rainfall starting point as the origin, the rainfall carrying capacity is calculated for each time period during the rainfall process with any rainfall time as the origin, and the time-by-time dynamic rainfall carrying capacity calculation is considered with real-time correction.

[0013] Preferably, the specific steps for determining the maximum target water level of the reservoir include:

[0014] The disaster-causing water level of the most dangerous section of the downstream protected object of the reservoir was determined by on-site investigation. The critical flow was calculated by using the water level-discharge relationship method. The obtained flow was substituted into the water level-storage capacity-discharge curve to obtain the water level and storage capacity corresponding to the critical flow, that is, the target water level and target storage capacity.

[0015] Preferably, the specific steps for determining the parameters include:

[0016] Design storm calculations were performed based on hydrological and storm data from the study area. The design storm frequency, rainfall amounts for different durations, and rainfall time-series distribution were determined according to the design flood standard and check flood standard for small reservoirs. Based on different soil moisture contents, appropriate runoff generation and inflow calculation methods were selected, and an HEC-HMS hydrological model suitable for small reservoir watersheds was constructed. The rainfall time-series distribution for different design storm frequencies and durations was input into the HEC-HMS hydrological model to calculate the runoff generation and inflow process of the watershed, i.e., the inflow flood process of the reservoir. Based on the water balance equation and the reservoir's water level-capacity-discharge curve, a reservoir flood control model was established. The inflow flood process was input into the flood control model to calculate the reservoir's water level and corresponding capacity under different initial water levels. Based on the reservoir's maximum target water level and water level-capacity curve, the reservoir's rain-carrying capacity under different soil moisture contents, rainfall frequencies, rainfall durations, and initial water levels was calculated, thus constructing a hydrological model of small reservoir rain-carrying capacity.

[0017] Preferably, the specific steps for constructing the rainwater carrying capacity prediction model for small reservoirs include:

[0018] Based on the ease of data accessibility, rainfall characteristic parameters, watershed characteristic parameters, and reservoir characteristic parameters were selected as parameters influencing rainwater carrying capacity. Rainfall characteristic parameters include rainfall duration and frequency; watershed characteristic parameter is soil moisture content; and reservoir characteristic parameters include initial water level and target water level. These rainwater carrying capacity influencing parameters were used as independent variables in a small reservoir rainwater carrying capacity prediction model, with the small reservoir's rainwater carrying capacity as the dependent variable. A random forest regression algorithm was used to construct the model. By inputting different characteristic parameters into the constructed model, the rainwater carrying capacity of the small reservoir under different conditions can be obtained.

[0019] Preferably, the specific steps for determining the evaluation index of the prediction model include:

[0020] Select mean squared error (MSE) and coefficient of determination (r) 2 The mean absolute error (MAE) is used to accurately and objectively evaluate the predictive ability of the small reservoir rainwater carrying capacity prediction model.

[0021] Preferably, the rain-holding capacity is calculated and corrected hourly, and the specific steps include:

[0022] Based on the prediction model and hydrological model of the rain-carrying capacity of small reservoirs, the rain-carrying capacity is calculated for complete, uninterrupted rainfall, for each time period during the rainfall process with the rainfall start point as the origin, for each time period during the rainfall process with any rainfall time as the origin, and for dynamic rain-carrying capacity calculation considering real-time correction.

[0023] (III) Beneficial Effects

[0024] This invention provides a method for calculating the dynamic rainwater carrying capacity of reservoirs based on hydrological and random forest models. It has the following beneficial effects:

[0025] 1. Based on the reservoir operation plan and field surveys, determine the maximum target water level of the reservoir; input rainfall data of different frequencies and durations into the HEC-HMS hydrological model for runoff generation and confluence calculation, obtaining the runoff generation and confluence process under different soil moisture contents, rainfall frequencies, and rainfall durations, i.e., the reservoir inflow flood process; establish a reservoir flood control model based on the water balance equation and the reservoir's water level-capacity-discharge curve, calculating the reservoir water level and its corresponding capacity according to different initial water levels; based on the maximum target water level and the reservoir water level-capacity curve, calculate the reservoir's rain-carrying capacity under different soil moisture contents, rainfall frequencies, rainfall durations, and initial water levels, constructing a hydrological model for the rain-carrying capacity of a small reservoir; based on the random forest regression algorithm, select rainfall characteristic parameters, watershed characteristic parameters, and reservoir characteristic parameters to construct a prediction model for the rain-carrying capacity of a small reservoir, calculating the rain-carrying capacity of the small reservoir under any characteristic parameters; select the mean square error (MSE) and the coefficient of determination (r). 2 The mean absolute error (MAE) is used to objectively evaluate the rain-carrying capacity prediction model of small reservoirs. Based on the rain-carrying capacity prediction model and the hydrological model of rain-carrying capacity of small reservoirs, the rain-carrying capacity is calculated for complete, uninterrupted rainfall, the rain-carrying capacity is calculated for each time period during the rainfall process with the rainfall start point as the origin, the rain-carrying capacity is calculated for each time period during the rainfall process with any rainfall time as the origin, and the dynamic rain-carrying capacity is calculated for each time period considering real-time correction.

[0026] 2. This method uses hydrological model analysis to simulate the runoff generation and confluence of floods entering a small reservoir and calculate the reservoir's rain-carrying capacity, thus constructing a hydrological model of the rain-carrying capacity of a small reservoir. The random forest regression algorithm is used to analyze the relationships between datasets, and the constructed rain-carrying capacity prediction model of a small reservoir can calculate the rain-carrying capacity of a small reservoir under different rainfall characteristic parameters, watershed characteristic parameters, and reservoir characteristic parameters.

[0027] 3. The mean squared error (MSE) of the prediction model constructed using this method on the test set is 1.256, and the coefficient of determination (r) is... 2The mean absolute error (MAE) was 0.983, and the mean absolute error (MAE) was 2.036, indicating that the rain-collecting capacity prediction model had a good prediction effect. This provides a new approach for calculating the rain-collecting capacity of small reservoirs, a new technical method for improving the timeliness and accuracy of flash flood disaster forecasts for small reservoirs, and an application path for flash flood early warning work.

[0028] 4. This method constructs a hydrological model and a prediction model for the rain-carrying capacity of small reservoirs, and proposes four methods for calculating rain-carrying capacity. It can calculate and correct the rain-carrying capacity during one or more rainfall events, reduce errors caused by rainfall-runoff processes and reservoir scheduling processes, and predict the rain-carrying capacity of the reservoir at any rainfall time in the future. This effectively extends the flood forecast period, increases the response time of reservoir flood control scheduling, optimizes reservoir regulation, and improves the overall operational efficiency of the reservoir. It is of great significance for improving the flood control capacity of reservoirs and ensuring the safety of reservoir flood control scheduling. Attached Figure Description

[0029] Figure 1 This is a flowchart illustrating the construction process of the reservoir rainfall carrying capacity hydrological model and the random forest model in this invention;

[0030] Figure 2 A flowchart illustrating the method for calculating rain-holding capacity for complete, uninterrupted rainfall.

[0031] Figure 3 A schematic diagram illustrating the method for calculating rain-holding capacity for complete, uninterrupted rainfall.

[0032] Figure 4 A schematic diagram illustrating the calculation of rain-collecting capacity at different time intervals during rainfall (with the origin as the starting point of rainfall);

[0033] Figure 5 A flowchart for calculating the rain-holding capacity at different times during a rainfall event (with any rainfall moment as the origin);

[0034] Figure 6 A schematic diagram illustrating the calculation of time-period dynamic rain-bearing capacity, taking into account real-time corrections;

[0035] Figure 7(a) shows the water level and storage capacity curve of Yekou Reservoir;

[0036] Figure 7(b) shows the water level discharge curve of Yekou Reservoir;

[0037] Figure 8 Comparison chart of predicted and measured values ​​of the rainfall carrying capacity prediction model for Yekou Reservoir;

[0038] Figure 9 Table showing the calculation results of sub-basin parameters for the HEC-HMS hydrological model;

[0039] Figure 10 Table of river parameter calculation results for the HEC-HMS hydrological model;

[0040] Figure 11 Table of Rain-Holding Capacity and its Influencing Parameters;

[0041] Figure 12 Table of evaluation index results for the rainwater carrying capacity prediction model of Yekou Reservoir. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] Example 1:

[0044] like Figures 1-6 As shown, this embodiment of the invention provides a method for calculating the dynamic rainwater carrying capacity of a reservoir based on hydrological and random forest models, including the following steps:

[0045] Determine the maximum target water level of the reservoir: Based on the reservoir operation plan and on-site investigation, determine the maximum target water level of the reservoir.

[0046] Parameter determination: Rainfall data with different frequencies and durations are input into the HEC-HMS hydrological model for runoff generation and confluence calculation to obtain the runoff generation and confluence process under different soil moisture content, rainfall frequency, and rainfall duration, i.e., the reservoir inflow flood process; a reservoir flood control model is established based on the water balance equation and the reservoir's water level-capacity-discharge curve, and the reservoir water level and its corresponding capacity are calculated according to different initial water levels; based on the reservoir's maximum target water level and water level-capacity curve, the reservoir's rain-carrying capacity is calculated under different soil moisture content, rainfall frequency, rainfall duration, and initial water level for a single rainfall event, and a hydrological model of the rain-carrying capacity of a small reservoir is constructed.

[0047] Based on the random forest regression algorithm, a rain-carrying capacity prediction model for a small reservoir is constructed by selecting rainfall characteristic parameters, watershed characteristic parameters, and reservoir characteristic parameters, and the rain-carrying capacity of the small reservoir under any characteristic parameters is calculated.

[0048] Determine the evaluation indicators for the prediction model: Select the mean squared error (MSE) and the coefficient of determination (r). 2 The mean absolute error (MAE) is used to objectively evaluate the rainwater carrying capacity prediction model of small reservoirs.

[0049] Rainfall carrying capacity calculation and correction on an hourly basis: Based on the rainfall carrying capacity prediction model and the rainfall carrying capacity hydrological model of small reservoirs, the rainfall carrying capacity is calculated for complete, uninterrupted rainfall, the rainfall carrying capacity is calculated for each time period during the rainfall process with the rainfall starting point as the origin, the rainfall carrying capacity is calculated for each time period during the rainfall process with any rainfall time as the origin, and the time-by-time dynamic rainfall carrying capacity calculation is considered with real-time correction.

[0050] Furthermore, the specific steps for determining the maximum target water level of the reservoir include:

[0051] The disaster-causing water level of the most dangerous section of the downstream protected object of the reservoir was determined by on-site investigation. The critical flow was calculated by using the water level-discharge relationship method. The obtained flow was substituted into the water level-storage capacity-discharge curve to obtain the water level and storage capacity corresponding to the critical flow, that is, the target water level and target storage capacity.

[0052] Furthermore, the specific steps for determining the parameters include:

[0053] Design storm calculations were performed based on hydrological and storm data from the study area. The design storm frequency, rainfall amounts for different durations, and rainfall time-series distribution were determined according to the design flood standard and check flood standard for small reservoirs. Based on different soil moisture contents, appropriate runoff generation and runoff calculation methods were selected, and an HEC-HMS hydrological model suitable for small reservoir watersheds was constructed. The rainfall time-series distribution for different design storm frequencies and durations was input into the HEC-HMS hydrological model to calculate the runoff generation and runoff process of the watershed, i.e., the inflow flood process of the reservoir. Based on the water balance equation and the reservoir's water level-capacity-discharge curve, a reservoir flood control model was established. The inflow flood process was input into the flood control model to calculate the reservoir's water level and corresponding capacity under different initial water levels. Based on the reservoir's maximum target water level and water level-capacity curve, the difference in capacity ΔW between the reservoir's water level and the maximum target water level was calculated. Substituting this into the rainwater carrying capacity calculation formula, the rainwater carrying capacity was obtained, and a hydrological model of the rainwater carrying capacity of small reservoirs was constructed.

[0054] Based on the actual situation, the calculation processes for watershed runoff generation, confluence, baseflow, and channel flow were performed using the SCS-CN curve method, SCS unit hydrograph method, exponential decay method, and Muskingen method, respectively.

[0055] The fundamental principle of the SCS-CN curve method is the water balance principle and two basic assumptions. The calculation formula is as follows:

[0056]

[0057] In the formula: P t P is the cumulative net rainfall at time t, in mm; I is the rainfall at time t, in mm; adenoted as initial rainfall loss (mm); S represents maximum soil retention (mm).

[0058] Initial rainfall loss I a The expression relating the maximum soil retention capacity S is:

[0059] Ia = 0.2S (2)

[0060] The expression relating the dimensionless parameter CN to the maximum soil retention capacity S is as follows:

[0061]

[0062] In the formula: CN is related to soil type, land use type and soil moisture in the early stage, and generally ranges from 40 to 98. The value of CN has a significant impact on flood simulation.

[0063] The SCS unit hydrograph is a dimensionless, single-peaked unit hydrograph. The required parameters are simple and easy to calculate. The peak flow rate U of the unit hydrograph is... p Unit line peak delay T p The relationship formula is:

[0064]

[0065] In the formula: A is the catchment area, km² 2 C is the conversion factor, with a value of 2.08. Unit hydrostatic peak delay T p The relationship with the unit net rainfall duration is as follows:

[0066]

[0067] In the formula: Δt is the net rainfall duration, in hours; t lag The peak lag time of the sub-basin flood is the time difference between the location of the net rainfall center and the peak value of the unit hydrograph, expressed in hours.

[0068] The exponential decay method refers to the exponential decay of the initial baseflow. It can effectively describe the process of water replenishment from the water storage capacity of a watershed to the river during dry periods. It also has a simple structure, and the calculation formula is as follows:

[0069] Q t =Q0k t (6)

[0070] In the formula: Q t The base flow rate at time t, in m³ / s. 3 / s; k is the drainage coefficient; Q0 is the initial base flow rate, unit: m³ / s. 3 / s.

[0071] The Muskingan method calculates river flow by solving the Muskingan channel storage equation and the water balance equation simultaneously. The Muskingan calculation equation is as follows:

[0072] Q2=C0I2+C1I1+C2Q1 (7)

[0073] in:

[0074]

[0075] In the formula, I1, I2, Q1, and Q2 represent the inflow and outflow at the upstream and downstream cross-sections at the beginning and end of the river period, respectively, and m 3 / s; Δt is the calculation period, h; K is the storage constant, representing the river section propagation time under steady flow conditions; x is the flow proportion factor.

[0076] Based on the water balance equation and the reservoir's water level-capacity-discharge curve, a reservoir flood control model is established. The inflow flood process obtained from the HEC-HMS hydrological model is input into the flood control model to calculate the water level and reservoir capacity under different initial water levels. Based on the reservoir's maximum target water level and water level-capacity curve, the reservoir's rainwater carrying capacity is calculated under different soil moisture content, rainfall frequency, rainfall duration, and initial water levels.

[0077] The formula for calculating rainwater carrying capacity is as follows:

[0078]

[0079] In the formula, P is the reservoir's rainwater carrying capacity (mm); ΔW is the difference between the reservoir's water level and its maximum target water level; α is the runoff coefficient; and F is the reservoir's catchment area (km²). 2 k is the unit conversion factor, which is usually taken as 10.

[0080] Furthermore, based on the random forest regression algorithm, a rain-carrying capacity prediction model for a small reservoir is constructed by selecting rainfall characteristic parameters, watershed characteristic parameters, and reservoir characteristic parameters, and the rain-carrying capacity of the small reservoir under any characteristic parameters is calculated. Specific steps include:

[0081] Rainfall carrying capacity of a reservoir is affected by a variety of parameters, including rainfall duration, rainfall frequency, previous soil moisture content, watershed underlying surface conditions, and initial and target water levels of the reservoir. This disclosure divides the parameters affecting rainfall carrying capacity into rainfall characteristic parameters, watershed characteristic parameters, and reservoir characteristic parameters, and uses them as independent variables in a small reservoir rainfall carrying capacity prediction model. The small reservoir rainfall carrying capacity is used as the dependent variable to construct a small reservoir rainfall carrying capacity prediction model, thereby calculating the rainfall carrying capacity of the small reservoir under any characteristic parameters.

[0082] Furthermore, the specific steps for determining the evaluation index of the prediction model include:

[0083] Determine the evaluation indicators for the prediction model: Select the mean squared error (MSE) and the coefficient of determination (r). 2The mean absolute error (MAE) is used to objectively evaluate the rainwater carrying capacity prediction model of small reservoirs.

[0084] 1) Mean Squared Error (MSE) is the average of the squared differences between the predicted values ​​from the model and the actual observed values. A smaller MSE indicates a better model fit. The calculation formula is shown below:

[0085]

[0086] 2) Coefficient of determination (r) 2 The variance explained by the model is denoted as 0, and its value ranges from 0 to 1. A value closer to 1 indicates a better model fit. The calculation formula is shown below:

[0087]

[0088] 3) Mean Absolute Error (MAE) is the average of the absolute differences between the model's predicted values ​​and the actual observed values. A smaller MAE indicates a better model fit. The calculation formula is shown below:

[0089]

[0090] Furthermore, the rain-holding capacity is calculated and corrected hourly, and the specific steps include:

[0091] (1) Calculation of rain-holding capacity for complete, uninterrupted rainfall:

[0092] The duration of a rainfall event is defined as h, and the rainfall frequency is defined as P. 频率 The rainfall amount is p, the soil moisture content is W, and the initial water level is the reservoir's original water level Z0. These five parameters are used as independent variables and input into a constructed small reservoir rainfall capacity prediction model to obtain the reservoir's rainfall capacity P after this rainfall event. 纳雨 .

[0093] (2) Calculation of rain-holding capacity at different times during the rainfall process (with the start of rainfall as the origin):

[0094] A rainfall event with a known duration of h1 and a rainfall amount of p1 has a rainfall frequency of P. 频率 The initial water level is Z0, the soil moisture content is W, and the data are input into a small reservoir rainfall capacity prediction model to calculate the rainfall capacity as P1. The unknown rainfall duration is h2, and the unknown rainfall amount is p. x The rainfall process will have a duration of h1+h1 and a rainfall amount of p1+p. x Rainfall frequency P 频率 The initial water level is Z0, and the soil moisture content is W. These values ​​are input into a small reservoir rainwater carrying capacity prediction model to calculate the rainwater carrying capacity as P. x This allows us to calculate the reservoir's rain-holding capacity after two rainfall events.

[0095] (3) Calculation of rain-bearing capacity at different times during the rainfall process (any rainfall time is the origin):

[0096] The initial soil moisture content is W0, the initial water level of the reservoir is Z0, and a rainfall event with a duration of h1 and a rainfall amount of p1 is input into the hydrological model of the small reservoir's rain-carrying capacity. After time period h1, a rainfall event with a duration of h2 and a rainfall amount of p2 occurs. At this time, the reservoir water level is Z1, the rain-carrying capacity is P1, and the soil moisture content is W1. The flood that did not enter the reservoir after the first rainfall event after time period h1 is combined with the second flood event. Z1 is used as the initial water level of the reservoir and input into the hydrological model of the small reservoir's rain-carrying capacity to calculate the rain-carrying capacity of the reservoir after the two rainfall events.

[0097] (4) Calculation of time-period dynamic rainfall capacity considering real-time correction:

[0098] The initial soil moisture content is W0, the initial water level of the reservoir is Z0, and the duration of a rainfall event is h. n Rainfall amount is p n The rainfall process is divided into n parts according to the average duration of the rainfall, i.e., nh = h n The initial rainfall duration (h) and its amount are input into a small reservoir's rainfall carrying capacity hydrological model to calculate and predict the inflow flood process. The predicted value is then corrected using the measured inflow flood process with an initial rainfall duration of h. The corrected flood process is then input into the small reservoir's rainfall carrying capacity hydrological model to calculate the corrected reservoir's rainfall carrying capacity. Next, the corrected flood process with a rainfall duration of h is used to correct a flood process with a rainfall duration of 2 hours, and then input into the small reservoir's rainfall carrying capacity hydrological model to calculate the corrected reservoir's rainfall carrying capacity for a rainfall duration of 2 hours. This process is repeated until the total rainfall duration is h. n The flood process was corrected and substituted into the small reservoir's rain-holding capacity hydrological model to calculate the corrected rain-holding capacity of the reservoir.

[0099] Example 2:

[0100] The example disclosed herein is the Yekou Reservoir basin in Wenquan Town, Huancui District, Weihai City, Shandong Province. Yekou Reservoir belongs to the Wuzhu River basin and is located in a low mountain and hilly area, with a basin area of ​​14 km². 2 The main channel is 2.79 km long, with an average gradient of 0.021 m / m. The Yekou Reservoir basin is located in a semi-humid region with an average annual rainfall of 722 mm, exhibiting highly uneven distribution both interannually and intraannually. The Yekou Reservoir protects the lives and property of 10,000 people in 20 downstream villages and safeguards 5.0 km of important transportation routes and infrastructure from flooding. This region is prone to sudden flooding, with short confluence times and rapid rises and falls in floodwaters, making it highly vulnerable to flood disasters. Failure to comply would result in significant political and economic losses.

[0101] The most dangerous cross-section downstream of the reservoir was determined through on-site investigation, and the critical flow rate was calculated to be 325.19 m³ / h using the water level-discharge relationship method. 3 / s, meaning the maximum allowable discharge from the reservoir is 325.19 m³ / s, provided that downstream areas are protected from flooding. 3 / s, using the water level-storage capacity-discharge curve, the reservoir water level and storage capacity corresponding to the maximum allowable discharge flow are obtained, i.e., the target water level is 66.91m, as shown in Figure 7(b), and the target storage capacity is 6.903 million m³. 3 / s, as shown in Figure 7(a).

[0102] The flood control standard for Yekou Reservoir is designed for a 50-year return period and checked against a 500-year return period. The 50-year and 500-year return periods are used as typical rainfall frequencies. Two intermediate scenarios, a 100-year and a 200-year return period, are selected, and a 20-year return period is added, thus defining five typical rainfall frequencies: 20-year, 50-year, 100-year, 200-year, and 500-year. Based on the "Shandong Provincial Hydrological Atlas (1975)," 1, 3, 6, 12, and 24-hour rainfall durations are selected for design rainfall calculations and rainfall time-history allocation. DEM data, land type data, and soil use data of the Yekou Reservoir watershed are collected to construct an HEC-HMS hydrological model, obtaining watershed characteristic parameters, such as... Figure 9 , Figure 10 As shown, based on soil moisture content, it is divided into relatively dry (0.2W) max ), generally (0.5W) max ) and moist (0.8W) max Three scenarios were considered: the rainfall time-history distribution was input into the hydrological model to obtain the runoff process under different soil moisture contents, rainfall frequencies, and rainfall durations, i.e., the reservoir inflow flood process. A reservoir flood control model was established using the water balance formula and the reservoir's water level-capacity-discharge curve. Based on initial reservoir water levels of 60m, 61m, 62m, 63m, 64m, and 64.8m, the reservoir inflow flood process was substituted into the flood control model to calculate the reservoir outflow flood process, obtaining the reservoir water level and corresponding capacity under different initial water levels. The reservoir's rainwater carrying capacity was calculated using the reservoir rainwater carrying capacity calculation formula, resulting in a total of 450 calculation results. A hydrological model of small reservoir rainwater carrying capacity was then constructed.

[0103] The rainwater carrying capacity of a reservoir is influenced by various parameters, including rainfall duration, rainfall frequency, soil moisture content, watershed underlying surface conditions, and initial and target water levels. Based on data availability, rainfall characteristic parameters, watershed characteristic parameters, and reservoir characteristic parameters are selected as parameters affecting rainwater carrying capacity. Rainfall characteristic parameters include rainfall duration, rainfall frequency, and rainfall amount; watershed characteristic parameter is soil moisture content; and reservoir characteristic parameter includes initial water level. These rainwater carrying capacity influencing parameters are used as independent variables in a small reservoir rainwater carrying capacity prediction model, with the small reservoir's rainwater carrying capacity as the dependent variable. A random forest regression algorithm is used to construct the small reservoir rainwater carrying capacity prediction model. Rainwater carrying capacity and its influencing characteristic parameters are as follows: Figure 11 As shown, the 450 calculation results were divided into a training set and a test set in a 5:1 ratio, i.e., 375 groups in the training set and 75 groups in the test set. The training set data was used to train the model, and the test set data was used to evaluate the model's performance, but not for model building.

[0104] Select mean squared error (MSE) and coefficient of determination (r) 2 The mean absolute error (MAE) is used to objectively evaluate the rainwater carrying capacity prediction model for small reservoirs, such as... Figure 12 As shown.

[0105] Based on the prediction model and hydrological model of the rain-carrying capacity of small reservoirs, the rain-carrying capacity is calculated for complete, uninterrupted rainfall, for each time period during the rainfall process with the rainfall start point as the origin, for each time period during the rainfall process with any rainfall time as the origin, and for dynamic rain-carrying capacity calculation considering real-time correction.

[0106] In summary, a model for predicting the rain-carrying capacity of small reservoirs was constructed based on the random forest regression algorithm. This model establishes the relationship between rainfall characteristic parameters, watershed characteristic parameters, reservoir characteristic parameters, and rain-carrying capacity. It can quickly calculate the dynamic rain-carrying capacity of reservoirs under different characteristic parameters. This method is data-driven, does not involve explicit physical processes, and eliminates complex runoff generation and flood control calculations. Furthermore, it does not completely deviate from basic hydrological theory and has a certain mathematical foundation. In practical applications, this method can quickly and accurately estimate the rain-carrying capacity of small reservoirs, meeting the practical needs of small reservoir forecasting and early warning. Simultaneously, it uses three methods—origin calculation, arbitrary starting point calculation, and iterative calculation—to deduce the rain-carrying capacity of reservoirs after a rainfall event. The calculated results are compared with the measured rain-carrying capacity of reservoirs, as shown in the figure. Figure 8 As shown, iterative calculations can correct errors caused by the runoff generation and confluence process and the reservoir flood regulation process in real time, significantly improving the calculation accuracy.

[0107] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for calculating the dynamic water storage capacity of a reservoir based on hydrological and random forest models, characterized by, Includes the following steps: Determine the maximum target water level of the reservoir: Based on the reservoir operation and management plan and on-site investigation, determine the maximum target water level of the reservoir; Parameter determination: Rainfall data with different frequencies and durations are input into the HEC-HMS hydrological model for runoff generation and confluence calculation to obtain the runoff generation and confluence processes under different soil moisture contents, rainfall frequencies, and rainfall durations, i.e., the reservoir inflow flood process; a reservoir flood control model is established based on the water balance equation and the reservoir's water level-capacity-discharge curve, and the reservoir water level and its corresponding capacity are calculated according to different initial water levels; based on the reservoir's maximum target water level and water level-capacity curve, the reservoir's rain-carrying capacity is calculated under different soil moisture contents, rainfall frequencies, rainfall durations, and initial water levels for a single rainfall event, and a hydrological model of small reservoir rain-carrying capacity is constructed. Based on the random forest regression algorithm, a rain-carrying capacity prediction model for a small reservoir is constructed by selecting rainfall characteristic parameters, watershed characteristic parameters, and reservoir characteristic parameters, and the rain-carrying capacity of the small reservoir under any characteristic parameters is calculated. Determine the evaluation indicators for the prediction model: Select the mean square error (MSE), coefficient of determination (r²), and mean absolute error (MAE) to objectively evaluate the prediction model for the rainwater carrying capacity of small reservoirs; Rainfall carrying capacity calculation and correction on an hourly basis: Based on the rainfall carrying capacity prediction model and the rainfall carrying capacity hydrological model of small reservoirs, the rainfall carrying capacity is calculated for complete, uninterrupted rainfall, the rainfall carrying capacity is calculated for each time period during the rainfall process with the rainfall starting point as the origin, the rainfall carrying capacity is calculated for each time period during the rainfall process with any rainfall time as the origin, and the time-by-time dynamic rainfall carrying capacity calculation is considered with real-time correction.

2. The reservoir dynamic storage capacity calculation method based on hydrological and random forest model according to claim 1, characterized in that: The specific steps for determining the maximum target water level of the reservoir include: The disaster-causing water level of the most dangerous section of the downstream protected object of the reservoir was determined by on-site investigation. The critical flow was calculated by using the water level-discharge relationship method. The obtained flow was substituted into the water level-storage capacity-discharge curve to obtain the water level and storage capacity corresponding to the critical flow, that is, the target water level and target storage capacity.

3. The method for calculating the dynamic rainwater carrying capacity of a reservoir based on a hydrological and random forest model according to claim 1, characterized in that: The specific steps for determining the parameters include: Design storm calculations were performed based on hydrological and storm atlases of the study area. The design storm frequency, rainfall amounts for different durations, and rainfall time-series distribution were determined according to the design flood standard and check flood standard for small reservoirs. Based on different soil moisture contents, a runoff generation and runoff calculation method was selected, and an HEC-HMS hydrological model suitable for small reservoir watersheds was constructed. The rainfall time-series distribution for different design storm frequencies and durations was input into the HEC-HMS hydrological model to calculate the runoff generation and runoff process of the watershed, i.e., the inflow flood process of the reservoir. Based on the water balance equation and the reservoir's water level-capacity-discharge curve, a reservoir flood control model was established. The inflow flood process was input into the flood control model to calculate the reservoir's water level and corresponding capacity under different initial water levels. Based on the reservoir's maximum target water level and water level-capacity curve, the reservoir's rain-carrying capacity under different soil moisture contents, rainfall frequencies, rainfall durations, and initial water levels was calculated, thus constructing a hydrological model of the rain-carrying capacity of small reservoirs.

4. The method for calculating the dynamic rainwater carrying capacity of a reservoir based on a hydrological and random forest model according to claim 1, characterized in that: The specific steps for constructing the rainwater carrying capacity prediction model for small reservoirs include: Based on the ease of data accessibility, rainfall characteristic parameters, watershed characteristic parameters, and reservoir characteristic parameters were selected as parameters affecting rainwater carrying capacity. Among them, rainfall characteristic parameters include rainfall duration and rainfall frequency, watershed characteristic parameter is soil moisture content, and reservoir characteristic parameters include initial water level and target water level. The rainwater carrying capacity influencing parameters were used as independent variables in the small reservoir rainwater carrying capacity prediction model, and the small reservoir rainwater carrying capacity was used as the dependent variable. The random forest regression algorithm was used to construct the small reservoir rainwater carrying capacity prediction model. Different characteristic parameters were input into the constructed small reservoir rainwater carrying capacity prediction model to obtain the reservoir rainwater carrying capacity under different conditions.

5. The reservoir dynamic capacity calculation method based on hydrological and random forest model according to claim 1, characterized in that: The specific steps for determining the evaluation index of the prediction model include: Select mean squared error (MSE) and coefficient of determination (r) 2 The mean absolute error (MAE) is used to accurately and objectively evaluate the predictive ability of the small reservoir rainwater carrying capacity prediction model.

6. The reservoir dynamic storage capacity calculation method based on hydrological and random forest model according to claim 1, characterized in that: The hourly calculation and correction of the rain-holding capacity includes the following specific steps: Based on the prediction model and hydrological model of the rain-carrying capacity of small reservoirs, the rain-carrying capacity is calculated for complete, uninterrupted rainfall, for each time period during the rainfall process with the rainfall start point as the origin, for each time period during the rainfall process with any rainfall time as the origin, and for dynamic rain-carrying capacity calculation considering real-time correction.

Images