Reservoir water surface evaporation capacity calculation method based on genetic algorithm

Abstract

The invention belongs to the technical field of water resources and atmospheric climate, and discloses a reservoir water surface evaporation capacity calculation method based on a genetic algorithm, which comprises the steps of data collection, variable screening, model construction, model verification and evaporation capacity calculation. According to the method, variable screening and model construction are optimized through a genetic algorithm, the synergistic effect of key factors such as water temperature (Tw) and net radiation (Rn) is mainly captured, and an air temperature-humidity interaction item is introduced to reflect the inhibition effect. According to the independent data set verification of the Three Gorges Reservoir in 2018-2021, the mean absolute error (MAE) of the model is reduced to 0.21 mm / d, the precision is improved by 108% compared with the Penman formula (MAE = 1.297 mm / d), the measurement and calculation deviation of the evaporation capacity of the complex water area is remarkably reduced, and the actual thermodynamic process is better met.

Description

Technical Field

[0001] This invention belongs to the field of water resources and atmospheric climate technology, and specifically relates to a method for calculating the evaporation of reservoir water surface based on a genetic algorithm. Background Technology

[0002] Evaporation from reservoir surfaces is a key parameter in the water resource cycle system, directly affecting reservoir water storage regulation, water supply and demand balance analysis, watershed ecological environment assessment, and water conservancy project scheduling decisions. Especially for mega-scale water conservancy projects like the Three Gorges Reservoir, accurate calculation of evaporation is of great significance for ensuring flood control safety, power generation efficiency, smooth navigation, and ecological water use.

[0003] Calculating reservoir evaporation is a core aspect of scientific water resource management, and its accuracy directly impacts the quality of decisions regarding water conservancy project scheduling, ecological water use security, and optimal water resource allocation. Against the backdrop of global climate change and frequent extreme hydrological events, the impact of reservoir evaporation on regional water resource balance is becoming increasingly prominent. The annual evaporation from the Three Gorges Reservoir is equivalent to the capacity of a medium-sized reservoir; therefore, accurate calculation of evaporation is of strategic significance for ensuring water supply security, smooth navigation, and ecological balance in the middle and lower reaches of the Yangtze River.

[0004] The evaporation process in deep-water reservoirs exhibits unique characteristics that distinguish it from natural water bodies. Large reservoirs possess enormous heat capacity; the thermal reserves of the Three Gorges Reservoir cause temperature changes to lag behind natural water temperature by 2-3 weeks, and the typical summer temperature gradient forms a thermodynamic barrier. The canyon topography results in significant spatial variations in net radiation flux; and the rough dynamics of the water surface lead to a non-linear relationship between wind speed and evaporation. Water surface evaporation is a crucial pathway for water loss in lakes and reservoirs; globally, evaporation losses from reservoirs exceed the total consumption of industrial and domestic water. The Three Gorges Dam project is located at the junction of the Sichuan Basin and the middle and lower reaches of the Yangtze River Plain, controlling a drainage area of ​​approximately 1 million km² in the upper reaches of the Yangtze River. 2 It accounts for 55.6% of the total area of ​​the Yangtze River basin. Since the Three Gorges Reservoir began impounding water, its water area has increased significantly, reaching 1084 km² at the normal water level of 175 meters. 2 Evaporation loss from the water surface affects the water balance of the reservoir to a certain extent. Research on water surface evaporation is of great significance for the water balance analysis of the Three Gorges Reservoir and the scientific scheduling of the Three Gorges Water Control Project.

[0005] Due to the difficulty of direct observation, estimating water surface evaporation is crucial. Currently, research and practice on reservoir evaporation have entered a development stage of estimating long-term reservoir (lake) evaporation using various meteorological driving data and satellite remote sensing data. Texas A&M University in the United States has considered the wind speed function method affected by water surface size, used the equilibrium temperature method to estimate water heat storage, and estimated the evaporation of a large number of reservoirs worldwide, including the Three Gorges Reservoir, based on the Penman formula. A team from the Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, used a similar method to estimate the monthly evaporation of 916 large reservoirs in my country from 1984 to 2018. However, these studies mostly focus on multiple reservoirs and lakes in a region, resulting in lower accuracy and failing to meet the requirements for precise calculation of water surface evaporation.

[0006] Currently, the mainstream methods for calculating evaporation from large reservoirs both domestically and internationally are the Penman formula (Penman, 1948) and its improved version, the FAO Penman-Monteith model (Allen, 1998). The Penman model is a formula based on energy balance and water vapor diffusion principles, capable of estimating pan evaporation by incorporating conventional meteorological data. This model is applicable to different time scales, including hours, days, and months, and the hourly-scale model is more accurate when hourly environmental data is available. The Penman formula calculates evaporation by coupling thermodynamic and aerodynamic equations, integrating parameters such as net radiation, air temperature, and humidity. However, it is based on the assumption of a "homogeneous water body," which cannot adapt to the complex characteristics of deep-water reservoirs. Taking the Three Gorges Reservoir as an example, its average water depth exceeds 70 meters, and summer water temperature stratification is significant (the temperature difference between the surface and bottom layers can reach 15℃). The thermal inertia of the water body and the vertical stratification effect are ignored by the Penman formula, resulting in an underestimation of evaporation by as much as 38% in an actual measurement in August 2016 when the surface water temperature reached 29.3℃. Furthermore, Penman's empirical estimation of net radiation does not consider the differences in reservoir topography. For example, the net radiation flux in the waters in front of the Three Gorges Dam differs from that in the tributary areas by 12-18%, directly causing a deviation of over 20% in the calculation of regional evaporation. Penman's comprehensive formula, derived from evaporation pan observation experiments and integrating energy balance and turbulent transport theory, includes radiation and aerodynamic terms. However, it originates from "sufficiently small" wetted surfaces such as evaporation pans, and in application, it overestimates the evaporation of large water surfaces such as lakes. It is necessary to determine the wind speed function to eliminate the deviation. However, the width of river reservoirs like the Three Gorges Reservoir varies significantly in time and space, making it difficult to determine the appropriate wind speed function. On the other hand, because lakes and reservoirs have lower roughness, higher specific heat capacity, and lower heat conduction rate compared to land, and because of the unique turbulent diffusion capacity of fluids, lakes and reservoirs have huge heat storage and release potential, resulting in a seasonal phase deviation between water surface evaporation and net radiation. This is the phenomenon of evaporation lag. Deep water bodies typically possess greater thermal storage capacity, with changes in stored thermal energy potentially exceeding 50% of net radiation. Their evaporation also exhibits a longer lag time compared to shallow lakes. In river-type deep-water reservoirs, stored thermal energy causes a significant lag in peak evaporation, necessitating accurate quantification of these changes. However, the widely used equilibrium temperature method assumes a depth of 20m for all water bodies deeper than 20m and fails to account for temperature stratification, resulting in substantial errors in deep water bodies like the Three Gorges Reservoir. Therefore, an accurate method for estimating surface evaporation applicable to the Three Gorges Reservoir and its upstream cascade reservoirs is needed. Accurate quantification of various components, including advection input and stored thermal energy, is crucial for improving the accuracy of surface evaporation estimation methods.

[0007] While the FAOPenman-Monteith model, as an internationally accepted standard, is theoretically more complete, it requires 8-10 parameters (such as net radiation, wind speed, and water vapor pressure). This not only demands a dense monitoring network but also incurs high equipment maintenance costs, making it difficult to apply comprehensively in reservoir areas with complex terrain and limited monitoring conditions. Furthermore, the model's aerodynamic resistance parameterization scheme is based on an ideal flat underlying surface design, but the canyon terrain of the Three Gorges Reservoir distorts the wind field (e.g., the relationship between wind speed at 10 meters and wind speed at 2 meters above the water is non-linear), leading to resistance calculation errors of 25-40%. The dynamic changes in water surface roughness (0.0002-0.005 m) further amplify the error to ±35%. In addition, the surface impedance parameter characterizing vegetation stomatal resistance in the model has no physical meaning for open water bodies, and after simplification, it is prone to overestimating evaporation in high-wind-speed areas.

[0008] The technology for calculating reservoir evaporation is developing in four directions: ① Multi-source data fusion: assimilation of data from ground monitoring stations, satellite remote sensing, and numerical models; ② Lightweight model structure: deployment of models using edge computing devices; ③ Transparent calculation process: analysis of evaporation contribution based on SHAP values; ④ Expanded application scenarios: from single reservoirs to collaborative optimization of cascade reservoir groups.

[0009] While traditional statistical models attempt to simplify calculations, they generally fail to consider the synergistic effect of water temperature (Tw) and net radiation (Rn), limiting their applicability under complex hydrological and meteorological conditions. In summary, existing technologies have significant shortcomings in terms of computational accuracy, parameter complexity, and terrain adaptability, necessitating a method for calculating evaporation that is tailored to the characteristics of large, deep-water reservoirs and balances accuracy and efficiency. Summary of the Invention

[0010] To overcome the problems of existing technologies, this invention proposes a method for calculating reservoir surface evaporation based on a genetic algorithm. Using this method to calculate reservoir surface evaporation can greatly reduce its mean absolute error (MAE) and significantly improve the calculation accuracy.

[0011] The objective of this invention is achieved as follows:

[0012] This invention provides a method for calculating reservoir evaporation based on a genetic algorithm, the method comprising the following steps:

[0013] Step 1, Data Collection:

[0014] Collect historical meteorological data of the reservoir, including the average water temperature T. a Wind speed u, average water vapor pressure above water e s Relative humidity γ, mean atmospheric pressure above water e a Water temperature T w Net radiation R n, and measured data of the evaporation rate E from the reservoir surface.

[0015] Step 2, Variable Selection:

[0016] Strongly correlated variables are retained first: Calculate the absolute value of the correlation coefficient between each variable and the evaporation E (|*r*|), remove weakly correlated variables with |*r*|≤0.3, and retain significant influencing factors with |*r*|>0.3 as core parameters.

[0017] Step 3, Model Building:

[0018] Using the selected core parameters as input variables, a model for calculating the evaporation rate of reservoir surface is constructed based on a genetic algorithm. The model structure includes, but is not limited to, linear regression models, nonlinear regression models, or artificial neural network models. The parameters and structure of the model are optimized through a genetic algorithm so that the model can accurately reflect the relationship between each parameter and the evaporation rate of the water surface.

[0019] Step 4, Model Validation:

[0020] The historical meteorological data collected in Step 1 for the reservoir is divided into a training set and a validation set. The model is trained using the training set, and during training, the model parameters are continuously adjusted using a genetic algorithm to minimize the error between the model's predicted values ​​and the measured evaporation. The validation set is used to monitor the training process. Training is stopped when the error on the validation set no longer decreases significantly, resulting in the optimized model for calculating reservoir surface evaporation. Step 5: Evaporation Calculation:

[0021] The real-time meteorological data is input into the optimized model, and after model calculation, the predicted value of the current water surface evaporation of the reservoir is output.

[0022] Furthermore, in step 2, the variable selection process, the genetic algorithm optimizes the selection of variables through the following constraint mechanisms:

[0023] (1) Simplification penalty orientation: Introduce a variable number penalty term in the fitness function of the genetic algorithm. When the number of selected variables increases by 1, the mean absolute error (MAE) of the model is required to decrease by >0.01mm / d. At the same time, redundant parameters with a contribution rate of <5% to the calculation of evaporation are automatically eliminated.

[0024] (2) Evolutionary stability constraint: In the crossover and mutation operations of the genetic algorithm, for variables with a high absolute value of the correlation coefficient (|*r*|) with evaporation E, the retention probability of their gene loci is >80%; the termination condition for variable screening is: the model accuracy fluctuation corresponding to the optimal variable combination output by the genetic algorithm for 50 consecutive generations is <0.1%.

[0025] Furthermore, the evaporation calculation model obtained in step 3 is as follows:

[0026] E = 0.273T w +0.188R n -0.00251T a (γ)

[0027] In the formula, E is the evaporation rate from the reservoir surface, and T is the evaporation rate from the reservoir surface. w For water temperature, R n For net radiation, T a γ represents the average air temperature over water, and γ represents the relative humidity.

[0028] The present invention provides a method for calculating reservoir water surface evaporation based on a genetic algorithm, which has the following significant advantages compared to existing technologies:

[0029] 1. Significantly Improved Calculation Accuracy: By optimizing variable selection and model construction through genetic algorithms, the model focuses on capturing the synergistic effects of key factors such as water temperature (Tw) and net radiation (Rn), and introduces an air-humidity interaction term to reflect the inhibition effect. Validated using an independent dataset from the Three Gorges Reservoir from 2018 to 2021, the model's mean absolute error (MAE) was reduced to 0.21 mm / d, a 108% improvement in accuracy compared to the Penman formula (MAE = 1.297 mm / d). This significantly reduces the measurement deviation of evaporation in complex water bodies and better reflects actual thermodynamic processes.

[0030] 2. Parameter Simplification and Cost Reduction: Utilizing the variable selection mechanism of a genetic algorithm, only core parameters significantly correlated with evaporation (|*r*|>0.3) (such as Tw, Rn, air temperature Ta, and relative humidity γ) are retained, while redundant variables (such as time and vapor pressure ea) are eliminated. Compared to the 8-10 parameters required by the FAOPenman-Monteith model, this invention reduces the number of parameters by more than 50%, reducing reliance on dense monitoring networks and significantly saving data acquisition and equipment maintenance costs, making it particularly suitable for reservoir areas with limited monitoring conditions.

[0031] 3. Adaptability to Complex Water Area Characteristics: Addressing the characteristics of large, deep-water reservoirs such as water temperature stratification and topographic heterogeneity, the model enhances its responsiveness to key factors like vertical water temperature differences and uneven distribution of net radiation by employing evolutionary stability constraints in genetic algorithms (e.g., retention probability of highly correlated variable gene loci > 80%). Compared to the traditional model's "homogeneous water body assumption," this invention is better suited to the evaporation mechanisms of complex water areas like the Three Gorges Reservoir, with regional calculation bias controlled within 20%. By selecting appropriate radiation and meteorological input data, the evaporation rate of different sections of the cascade reservoirs in recent years is estimated. Based on the operational status of the cascade reservoirs and the dynamic changes in reservoir surface area, the evaporation loss is estimated.

[0032] 4. Enhanced Model Stability and Generalization Ability: The introduction of a variable quantity penalty term and a convergence termination condition (optimal solution fluctuation <0.1% over 50 consecutive generations) effectively avoids overfitting and ensures the model's robustness (fit accuracy 0.765) on an independent validation dataset (2018-2021). The global optimization characteristics of the genetic algorithm make the model highly adaptable to changes in hydrological and meteorological conditions, and it can be extended to other similar terrain scenarios for measuring evaporation from large reservoirs.

[0033] 5. Supporting precise water resource management: High-precision, low-cost evaporation calculation results can provide a scientific basis for reservoir scheduling, such as optimizing water storage strategies to cope with seasonal evaporation losses and accurately assessing ecological water demand, thereby helping to improve water resource utilization efficiency and the comprehensive benefits of water conservancy projects. Attached Figure Description

[0034] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0035] Figure 1 This is a correlation graph of the formula fitting accuracy;

[0036] Figure 2 This is a comparison chart of the calculation results of the evaporation calculation model constructed by the method of this invention and the measured values. Detailed Implementation

[0037] The present invention will be further described in detail below with reference to preferred embodiments. More details are set forth in the following description in order to provide a full understanding of the present invention. However, the present invention can obviously be implemented in many other ways different from those described herein. Those skilled in the art can make similar extensions and derivations based on actual application situations without departing from the spirit of the present invention. Therefore, the scope of protection of the present invention should not be limited by the content of this specific embodiment.

[0038] Example 1:

[0039] This embodiment takes the Three Gorges Reservoir as an example and provides a method for calculating reservoir surface evaporation based on a genetic algorithm, specifically including:

[0040] Step 1, Data Collection:

[0041] Basic hydrological data were collected, and meteorological observation data from meteorological departments along the reservoir area over many years were compiled and organized. Data verification and consistency checks were conducted to verify the accuracy of the data. Based on the observation data, correlation analysis was performed using mathematical statistics and mathematical modeling methods, and corresponding relationships were proposed. On this basis, the distribution patterns and trends of water surface evaporation in various regions were studied and analyzed.

[0042] Specifically, two measured monthly meteorological datasets were collected from the Three Gorges Reservoir for the periods of 2014-2017 and 2018-2021, including the average surface temperature T. a Wind speed u, average water vapor pressure above water e s Relative humidity γ, mean atmospheric pressure above water e a Water temperature T w Net radiation R n , and measured data of the evaporation rate E from the reservoir surface.

[0043] The specific data is shown in Table 1 below.

[0044] Table 1

[0045]

[0046] Modeling period (2014-2017): 48 monthly datasets; Validation period (2018-2021): 48 monthly datasets.

[0047] The spatial distribution of the monitoring network is as follows: 12 monitoring points cover typical water areas of the Three Gorges Reservoir: the point in front of the dam (10°22′32″E, 31°02′56″N, 71km from the dam); the middle section of the reservoir (Zigui section); and the confluence area of ​​tributaries (Xiangxi River and Daning River).

[0048] Monthly-scale data generation:

[0049] Every morning at 8:00 AM, the water level in the evaporation dish is measured using a measuring needle with a minimum graduation of 0.1 mm. To reduce observational uncertainty, two readings are taken for each measurement. If the difference between the two readings does not exceed 0.2 mm, their average value is taken as the water level in the evaporation dish for that day, thus determining the daily evaporation. During the observation process, a rain gauge installed on the raft is used to assess the impact of precipitation on the water level.

[0050] parameter Original collection Monthly value generation method Evaporation rate E Daily measurement of evaporating dish (±0.1 mm) Monthly average value = Σ Daily value / Number of days in the month <![CDATA[Water temperature T w > Continuous recording (±0.2℃) Monthly average = Arithmetic mean of hourly averages <![CDATA[Net radiation R n > Radiometer sampling every 10 minutes (±5%) Monthly average = Daily average arithmetic mean

[0051] Step 2, Variable Selection:

[0052] Genetic algorithms were proposed by American computer scientist John Holland in the early 1960s and have been widely developed and applied in the decades since. It is a heuristic search and optimization algorithm based on biological evolution theory, simulating the evolutionary process in nature to find optimal solutions for complex search and optimization problems. This algorithm uses mathematical methods and computer simulations to transform the problem-solving process into processes similar to selection, crossover, and mutation in biological evolution, and then iterates these processes until a value meeting preset conditions is reached. It can perform global searches in a large solution space, with a high probability of finding the global optimum, rather than being limited to local optima; it is highly adaptable, capable of adapting to various types of optimization problems without requiring in-depth understanding of the specific structure of the problem, making its application wide-ranging; it gradually approaches the global optimum, providing multiple excellent solutions for the user to choose from, and progressively approaches the optimal solution during the iteration process; and it achieves better optimization results, often faster than some conventional optimization algorithms when solving complex combinatorial optimization problems. This innovative approach, which integrates physical mechanisms and evolutionary optimization, not only inherits the core advantages of genetic algorithms—global optimization, strong adaptability, and avoidance of local optima—but also addresses the limitations of traditional algorithms in complex hydrological scenarios through domain knowledge constraints.

[0053] In this invention, a genetic algorithm automatically identifies core variables that are significantly related to water surface evaporation from meteorological parameters. It achieves parameter selection driven by physical mechanisms through correlation coefficient thresholds and contribution rate analysis, dynamically optimizes linear and nonlinear models and neural network structures, and simultaneously optimizes the mathematical form and coefficient weights of the model. The complexity of the model is controlled by a penalty term, and convergence stability constraints are used to ensure generalization ability.

[0054] Its use involves the following four steps:

[0055] ① Import the dependent and independent variable data into the program;

[0056] ② Select the symbolic functions to be used in the program;

[0057] ③ A custom accuracy standard is used to evaluate the predicted value of the expression and the measured value of the validation group. Expressions that do not meet the standard are discarded. Then, the built-in probability function is used to perform a crossover operation on the remaining expressions to generate new expressions. This process is similar to the genetic operation in biological evolution, which obtains a better solution through continuous selection and crossover.

[0058] ④ Select the appropriate expression as the optimal solution from a series of formulas with different precision and complexity generated by the program.

[0059] Strongly correlated variables should be retained first: Calculate the absolute value of the correlation coefficient between each variable and the evaporation E (|r*|), and remove weakly correlated variables (such as e) where |r*| ≤ 0.3. a (Time), retain significant impact factors (such as T) where |*r*|>0.3. w R n (), as the core parameter.

[0060] Physical constraint: Water temperature T w Weights are forced to be positive (basics of evaporation thermodynamics); relative humidity γ weights are forced to be negative (verification of the inhibition effect); net radiation R n Threshold R that is positively correlated with evaporation 2 >0.6.

[0061] Simplification-oriented penalty: Introduce a variable quantity penalty term in the fitness function: For each additional variable, the model accuracy is required to improve by >0.01mm / d (MAE compensation threshold); redundant parameters with a contribution rate of <5% are automatically eliminated.

[0062] Evolutionary stability constraints: In crossover and mutation operations, the probability of retention of high |*r*| variable gene loci is >80%; Termination condition: The fluctuation of the optimal solution is <0.1% for 50 consecutive generations (to ensure the convergence and stability of coefficients).

[0063] The results of the fitting accuracy calculation are shown in Table 2.

[0064] Table 2

[0065]

[0066] Step 3, Model Building:

[0067] Using the selected core parameters as input variables, a genetic algorithm is used to perform nonlinear fitting on the dataset, resulting in the optimized evaporation calculation model:

[0068] E = 0.273T w +0.188R n -0.00251T a (γ)

[0069] In the formula, E is the evaporation rate from the reservoir surface, and T is the evaporation rate from the reservoir surface. w For water temperature, R n For net radiation, T a γ represents the average air temperature over water, and γ represents the relative humidity.

[0070] Figure 1 The correlation graph of the formula fitting accuracy is shown.

[0071] Step 4, Model Validation:

[0072] The model constructed in step 3 was verified using independent time period data (2018-2021 measured dataset). While ensuring the accuracy of the calculation results, fewer calculation parameters were used, making it simpler and more convenient. The MAE of the verified model was 0.21.

[0073] Step 5, Evaporation Calculation:

[0074] Input real-time monitoring T w R n T a γ, output the predicted value of water surface evaporation E.

[0075] Figure 2 The results show a comparison between the calculation results of the evaporation calculation model constructed by the method of the present invention and the measured values.

[0076] In addition, using the measured data from the Three Gorges Reservoir from 2018 to 2021, the calculation accuracy of the evaporation calculation model described in this embodiment was compared with that of the Penman formula. The results are shown in the table below:

[0077] Verification Result Comparison Table

[0078] Model MAE (mm / d) This invention model 0.21 Penman-Monteith 1.297

[0079] The results shown in the table above demonstrate that the method described in this application can significantly reduce computational complexity while maintaining accuracy, and the mean absolute error (MAE) is improved by 108% compared to the Penman formula, making it suitable for calculating evaporation in deep water areas such as the Three Gorges Reservoir.

[0080] Finally, it should be noted that the above is only used to illustrate the technical solution of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred arrangement, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solution of the present invention (such as the application of a formula, the order of steps, etc.) without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for calculating reservoir evaporation based on a genetic algorithm, characterized in that, The method includes the following steps: Step 1, Data Collection: Collect historical meteorological data of the reservoir, including the average water temperature T. a Wind speed u, average water vapor pressure above water e s Relative humidity γ, mean atmospheric pressure above water e a Water temperature T w Net radiation R n , and measured data of the evaporation rate E of the reservoir surface; Step 2, Variable Selection: Strongly correlated variables are retained first: Calculate the absolute value of the correlation coefficient between each variable and the evaporation E (|*r*|), remove weakly correlated variables with |*r*|≤0.3, and retain significant influencing factors with |*r*|>0.3 as core parameters; Step 3, Model Building: Using the selected core parameters as input variables, a model for calculating the evaporation of water surface in a reservoir is constructed based on a genetic algorithm. The model structure includes, but is not limited to, linear regression models, nonlinear regression models, or artificial neural network models. The parameters and structure of the model are optimized through a genetic algorithm so that the model can accurately reflect the relationship between each parameter and the evaporation of water surface. Step 4, Model Validation: The historical meteorological data of the reservoir collected in step 1 is divided into a training set and a validation set. The model is trained using the training set. During the training process, the model parameters are continuously adjusted using a genetic algorithm to minimize the error between the model's predicted value and the measured evaporation. The training process is monitored using the validation set. When the error on the validation set no longer decreases significantly, the training is stopped, resulting in the optimized model for calculating the evaporation of the reservoir surface. Step 5, Evaporation Calculation: The real-time meteorological data is input into the optimized model, and after model calculation, the predicted value of the current water surface evaporation of the reservoir is output.

2. The method according to claim 1, characterized in that, In step 2, the variable selection process, the genetic algorithm optimizes the selection of variables through the following constraint mechanisms: (1) Simplification penalty orientation: Introduce a variable number penalty term in the fitness function of the genetic algorithm. When the number of selected variables increases by 1, the mean absolute error (MAE) of the model is required to decrease by >0.01mm / d. At the same time, redundant parameters with a contribution rate of <5% to the calculation of evaporation are automatically eliminated. (2) Evolutionary stability constraint: In the crossover and mutation operations of the genetic algorithm, for variables with a high absolute value of the correlation coefficient (|*r*|) with evaporation E, the retention probability of their gene loci is >80%; the termination condition for variable screening is: the model accuracy fluctuation corresponding to the optimal variable combination output by the genetic algorithm for 50 consecutive generations is <0.1%.

3. The method according to claim 1, characterized in that, The evaporation calculation model obtained in step 3 is as follows: E=0.273T w +0.188R n -0.00251T a (c) In the formula, E is the evaporation rate from the reservoir surface, and T is the evaporation rate from the reservoir surface. w For water temperature, R n For net radiation, T a γ represents the average air temperature over water, and γ represents the relative humidity.

4. The method according to claim 1, characterized in that, In step 1, the time scale of the historical meteorological data is monthly or daily.

5. The method according to claim 1 or 4, characterized in that, In step 1, the measured data comes from a monitoring network covering typical water areas of the reservoir, and the monitoring network includes at least three monitoring points: the area in front of the dam, the middle section of the reservoir, and the confluence area of ​​tributaries.

6. The method according to claim 1, characterized in that, In step 4, the ratio of the training set to the validation set is 7:3 to 8:

2.

7. The method according to claim 1, characterized in that, In step 4, the validation metrics for the model include mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R²). When R² ≥ 0.6, the model is considered to have passed training.

Images