Reservoir ecological operation method and system based on classification algorithm and decision boundary diagram

Abstract

The present application relates to the technical field of river ecological protection, and specifically discloses a reservoir ecological scheduling method and system based on a classification algorithm and a decision boundary diagram. The method includes: collecting spawning data and environmental data, constructing a machine learning model with environmental data as an independent variable and spawning data as a dependent variable; screening the first two independent variables with the highest variable importance as target variables, and constructing a decision boundary diagram of the target variable; constructing multiple ecological scheduling schemes based on the decision boundary diagram; determining the optimal scheduling date for implementing each ecological scheduling scheme; selecting an ecological scheduling scheme as the optimal scheme based on the optimal scheduling date, and before implementing the optimal scheme, if the environmental scheduling value is affected by environmental rainfall, adjusting the optimal scheduling date based on the actual environment of the target area. The present invention greatly improves the efficiency and scientific nature of the formulation of ecological scheduling schemes.

Description

Technical Field

[0001] The present application relates to the technical field of river ecological protection, and in particular to a reservoir ecological scheduling method and system based on a classification algorithm and a decision boundary diagram. Background Art

[0002] Under natural conditions, the spawning, foraging, and survival activities of fish have formed a long-term adaptive relationship with the aquatic habitat conditions of the river. Dam diversion and water storage have caused significant changes in the aquatic habitat of the river, which has an adverse impact on aquatic organisms such as fish. To solve this problem, existing technologies use ecological scheduling technology to change the operation and scheduling mode of reservoirs. During the fish spawning season, the normal spawning activities of fish are ensured by adjusting the discharge method and the outflow flow rate. However, excessive or insufficient flow rates have an adverse effect on the spawning activities of fish. Therefore, accurately identifying the environmental requirements of fish spawning activities is a prerequisite for the scientific implementation of ecological scheduling.

[0003] In the prior art, Chinese patent document publication number CN119180721A discloses an ecological scheduling method for increasing the spawning rate of fish that lay sticky eggs in reservoirs. This method, based on meteorological conditions and dam gate openings, predicts and dynamically adjusts to sudden events such as rainfall and dam gate openings, while avoiding the economic costs and timeliness of instantaneous regulation and response associated with real-time monitoring, thereby promoting fish spawning. Another example is Chinese patent document publication number CN109447848A, which discloses an ecological scheduling system for cascade hydropower stations suitable for fish that lay drifting eggs. This system uses cascade hydropower stations as scheduling targets and, based on the spawning and reproduction needs of these fish, creates optimal flow and water level conditions for their spawning and reproduction through coordinated scheduling of upstream and downstream cascade hydropower stations, thereby more specifically promoting the spawning and reproduction of these fish.

[0004] However, fish spawning is affected not only by hydrological and hydrodynamic conditions such as flow, but also by water quality and environmental factors such as water temperature. This makes it difficult to achieve ideal ecological effects by controlling the outflow of a single reservoir without considering the synergistic effects of multiple factors. Summary of the Invention

[0005] In order to solve the problems raised in the above background technology, the present application provides a reservoir ecological scheduling method and system based on a classification algorithm and a decision boundary diagram.

[0006] In order to achieve the above-mentioned object of the invention, the present invention proposes a reservoir ecological scheduling method based on a classification algorithm and a decision boundary diagram, comprising:

[0007] Collecting spawning data and environmental data from a target area, and building a machine learning model using the environmental data as an independent variable and the spawning data as a dependent variable;

[0008] Performing importance analysis on the independent variables based on the output results of the machine learning model to obtain the variable importance of each independent variable;

[0009] Selecting the first two independent variables with the highest importance as target variables, constructing a grid data set about the target variables, and constructing a decision boundary graph based on the machine learning model and the grid data set;

[0010] Constructing multiple ecological scheduling schemes based on the decision boundary graph, and obtaining an environmental scheduling value for each of the ecological scheduling schemes;

[0011] Obtaining influencing factors of the environmental scheduling value, obtaining historical data of the influencing factors in the target area, and analyzing the historical data to determine the optimal scheduling date for implementing each of the ecological scheduling plans;

[0012] An ecological scheduling scheme is selected as an optimal scheme based on the optimal scheduling date. Before implementing the optimal scheme, if the environmental scheduling value is affected by environmental rainfall, the optimal scheduling date is adjusted based on the actual environment of the target area.

[0013] Furthermore, determining the optimal scheduling date includes the following steps:

[0014] Dividing the target variable of the ecological scheduling plan into a first variable and a second variable, determining a first judgment condition based on the first variable, and determining a second judgment condition based on the second variable;

[0015] A first time window is set, where the first time window is a time window for fish spawning. A first statistical value of the first variable within the first time window is determined based on the historical data. A second time window is determined in the first time window based on the first statistical value, and the first statistical value within the second time window meets the first judgment condition. A second statistical value of the second variable within the second time window is determined based on the historical data. A third time window is determined in the second time window based on the second statistical value, and the second statistical value within the third time window meets the second judgment condition. The third time window is defined as the optimal scheduling date.

[0016] Furthermore, adjusting the optimal solution includes the following steps:

[0017] Locate multiple upstream points, where the rainfall at the upstream points directly affects the inflow of the dam area, predict the daily rainfall at each upstream point within the optimal scheduling date to obtain the predicted rainfall, establish a conversion model, convert the predicted rainfall based on the conversion model to obtain the predicted inflow of each upstream point, determine the inflow weight of each upstream point, perform weighted summation of the predicted inflow based on the inflow weight, and obtain the predicted inflow of the dam area; if the predicted inflow of each day within the optimal scheduling date is less than a critical threshold, implement the optimal plan on the optimal scheduling date; otherwise, adjust the optimal scheduling date of the optimal plan.

[0018] Furthermore, the conversion model converts the predicted rainfall including the following steps:

[0019] The conversion model includes a first conversion function and a second conversion function, which determine the soil saturation value and basic flow of each upstream point. If the predicted rainfall is less than the soil saturation value, the predicted rainfall is converted into a first flow based on the first conversion function, and the sum of the first flow and the basic flow is used as the predicted inflow flow. Otherwise, the predicted rainfall is divided into a first part and a second part based on the soil saturation value, the first part is converted into a first inflow flow based on the first conversion function, and the second part is converted into a second inflow flow based on the second conversion function, and the sum of the first inflow flow, the second inflow flow and the basic flow is used as the predicted inflow flow.

[0020] Furthermore, determining the inflow weight includes the following steps:

[0021] setting a water flow arrival time based on the distance between the upstream point and the target area, obtaining historical rainfall data of the upstream point in a historical time period, defining a time period in which the historical rainfall data appears as a first time period, and determining a second time period in which the rainfall flow arrives at the dam area based on the water flow arrival time;

[0022] Splitting the first time period into a plurality of first sub-time periods, and correspondingly splitting the second time period into a plurality of second sub-time periods, obtaining the inflow flow of the upstream point in the first sub-time period, and the inflow flow of the dam area in each of the second sub-time periods, setting an initial weight for each upstream point, and weighted summing the inflow flow of each of the first sub-time periods based on the initial weight to obtain a plurality of influencing quantities, constructing a first sequence based on the influencing quantities, and constructing a second sequence based on the inflow flow;

[0023] Performing a correlation analysis on the first sequence and the second sequence to obtain a correlation coefficient, adjusting the initial weight and then recalculating the correlation coefficient, repeating the process multiple times, and using the initial weight corresponding to the maximum correlation coefficient as the corresponding import weight.

[0024] Furthermore, the target variable affected by rainfall is set as the first variable.

[0025] Furthermore, selecting one of the ecological scheduling schemes as the best scheme includes the following steps:

[0026] The power generation benefit of each ecological scheduling scheme is calculated, and the ecological scheduling scheme with the highest power generation benefit is selected as the optimal scheme.

[0027] Furthermore, the variable importance is the SHAP value of each of the environmental data.

[0028] Furthermore, the environmental data includes water temperature, water flow, transparency, water level and sediment content, and the machine learning model is a binary tree machine learning model.

[0029] The present invention also provides a reservoir ecological scheduling system based on a classification algorithm and a decision boundary diagram, which is used to implement the above-mentioned reservoir ecological scheduling method based on a classification algorithm and a decision boundary diagram. The system includes:

[0030] a preprocessing module that collects spawning data and environmental data from a target area, constructs a machine learning model using the environmental data as an independent variable and the spawning data as a dependent variable, and performs importance analysis on the independent variables based on outputs of the machine learning model to obtain the variable importance of each independent variable;

[0031] A construction module is provided for selecting the first two independent variables with the highest importance as target variables, constructing a grid data set for the target variables, constructing a decision boundary graph based on the machine learning model and the grid data set, constructing multiple ecological scheduling schemes based on the decision boundary graph, and obtaining an environmental scheduling value for each of the ecological scheduling schemes;

[0032] an optimization module for obtaining influencing factors of the environmental scheduling value, obtaining historical data of the influencing factors in the target area, and analyzing the historical data to determine an optimal scheduling date for implementing each of the ecological scheduling plans;

[0033] An adjustment module selects one of the ecological scheduling schemes as the optimal scheme based on the optimal scheduling date, and before implementing the optimal scheme, if the environmental scheduling value is affected by environmental rainfall, adjusts the optimal scheduling date based on the actual environment of the target area.

[0034] On the one hand, the present invention builds a machine learning model by collecting spawning and environmental data in the target area and analyzing the importance of variables. It can accurately screen out the key environmental variables that have the greatest impact on fish spawning, and then construct a decision boundary diagram based on the key variables, providing implementers with a flexible decision-making basis and technical support. By comparing the positions of different scheduling plans in the decision boundary diagram and the corresponding probability of occurrence of target events, implementers can quickly and intuitively select the reservoir scheduling plan with the best effect, and intuitively present the favorable areas for fish spawning under different environmental conditions, greatly improving the efficiency and scientific nature of scheduling plan formulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 A flowchart of the steps of a reservoir ecological scheduling method based on a classification algorithm and a decision boundary diagram is provided for this application;

[0036] Figure 2 Schematic diagram of the analysis results of the importance of variables in Table 1 for this application;

[0037] Figure 3 This is a schematic diagram of the analysis results of the importance of variables in Table 2 of this application;

[0038] Figure 4 This is the decision boundary diagram corresponding to Table 1 of this application;

[0039] Figure 5 This is the decision boundary diagram corresponding to Table 2 of this application;

[0040] Figure 6 This is a structural diagram of a reservoir ecological scheduling system based on a classification algorithm and a decision boundary diagram for this application. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0042] It is understood that the terms "first," "second," etc., used herein may be used to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish a first element from another element. For example, a first xx script may be referred to as a second xx script, and similarly, a second xx script may be referred to as a first xx script without departing from the scope of this application.

[0043] like Figure 1 As shown in FIG, a reservoir ecological operation method based on a classification algorithm and a decision boundary diagram includes:

[0044] S1: Collect spawning data and environmental data from the target area, and build a machine learning model with environmental data as the independent variable and spawning data as the dependent variable.

[0045] Environmental data include water temperature, water flow, transparency, water level and sediment content, and the machine learning model is a binary tree machine learning model.

[0046] S2: Perform importance analysis on the independent variables based on the machine learning model to obtain the variable importance of each independent variable.

[0047] The variable importance is the SHAP value of each environment data.

[0048] Specifically, spawning data includes spawning and not spawning, environmental data includes daily average water temperature, water temperature 2-day variation, water temperature 3-day variation, daily average flow, flow 2-day variation, flow 3-day variation, daily average transparency, transparency 2-day variation, transparency 3-day variation, daily average water level, water level 2-day increment, water level 3-day increment, daily average sediment content, sediment content 2-day increment, sediment content 3-day increment. In addition, other specific statistical values ​​can also be included, such as mean, maximum and minimum values, and other environmental indicators and characteristic values ​​related to fish spawning that may affect fish spawning in theory or practice can also be used in the present technology. In actual use, different indicators can be flexibly adopted according to local conditions, according to different fish breeding habits. For example, the spawning data and environmental data of copper fish from Zhutuo to Jiangjin section of the upper Yangtze River collected are shown in Table 1 below. The spawning data and environmental data of the four major carps from Yidu to Shashi section of the middle Yangtze River collected are shown in Table 2 below.

[0049] Table 1

[0050]

[0051] Table 2

[0052]

[0053] In addition, the constructed machine learning model is a binary tree machine learning model, which specifically includes support vector machine, random forest, adaptive boosting machine, gradient boosting machine, extreme random tree, extreme gradient boosting machine, Light gradient boosting machine, CatBoost classifier, etc. The evaluation indicators of the machine learning model include Accuracy, AUC, Recall, Precision, F1, Kappa, MCC, etc. The importance evaluation of the input data of the machine learning model, that is, the importance evaluation of the independent variables is based on the SHAP value.

[0054] S3: Filter the top two independent variables with the highest variable importance as the target variable, build a grid dataset about the target variable, and build a decision boundary graph based on the machine learning model and the grid dataset.

[0055] For the data shown in Table 1 above, this embodiment constructs a Catboost model and calculates the variable importance of each independent variable. The results of variable importance are as follows: Figure 2 As shown, based on Figure 2 It can be determined that the top two independent variables with the highest variable importance are daily average water temperature and daily average flow. For the data shown in Table 2 above, this embodiment constructs an extreme random tree model and calculates the variable importance of each independent variable. The results of variable importance are as follows: Figure 3 As shown, based on Figure 3 It can be determined that the first two independent variables with the highest variable importance are the 3-day increase in water and the average daily water level.

[0056] S4: Construct multiple ecological scheduling schemes based on the decision boundary graph and obtain the environmental scheduling value of each ecological scheduling scheme.

[0057] The decision boundary diagram constructed based on Table 1 is as follows Figure 4 As shown in Table 2 above, the decision boundary diagram is constructed as follows Figure 5 As shown in the decision boundary diagram, the blue area is the area that is favorable for spawning, and the darker the color, the higher the favorability for spawning. By selecting multiple points in the decision boundary diagram, each point corresponds to an ecological scheduling scheme, and the horizontal and vertical coordinates corresponding to each point are the corresponding environmental scheduling values. Figure 4 As shown in Figure 1, for Scheme 1 and Scheme 2, Scheme 1 is more conducive to the spawning of copper fish, and its corresponding environmental scheduling values ​​are water temperature 23°C and water flow rate 7400 m³ / s. Figure 5 As shown, among Scheme 1, Scheme 2 and Scheme 3, Scheme 1 is obviously more suitable for spawning of the four major carps.

[0058] S5: Obtain the influencing factors of the environmental scheduling value, obtain the historical data of the influencing factors in the target area, and analyze the historical data to determine the optimal scheduling date for implementing each ecological scheduling plan.

[0059] Water flow and temperature are also affected by environmental factors. For example, when heavy rain occurs in the upstream river area, the water inflow to the dam area may surge, making it impossible for the dam to release water according to the planned flow rate. Therefore, it is necessary to select a date with relatively stable rainfall to avoid sudden rainfall that interferes with the ecological scheduling plan. Similarly, dates with smaller temperature fluctuations should be selected to avoid frequent switching of water release layers when the dam releases water. As mentioned above, it is necessary to obtain historical data on the influencing factors and analyze the historical data to determine the date when the influencing factors are relatively stable to implement the corresponding ecological scheduling plan. For example, if the ecological scheduling plan requires a water temperature of 24°C, it is necessary to determine a date when the water temperature is stable at 24°C. The specific determination method will be introduced later.

[0060] S6: Select an ecological scheduling scheme as the best scheme based on the best scheduling date. Before implementing the best scheme, if the environmental scheduling value is affected by environmental rainfall, adjust the best scheduling date based on the actual environment of the target area.

[0061] After determining the optimal scheduling date, the optimal ecological scheduling option must be selected from the available options. Based on the previous discussion, multiple options may exist in the decision boundary diagram, all of which are beneficial to fish spawning, so the optimal option must be selected. The dam's power generation efficiency can be considered when making this selection. Finally, before implementing the selected optimal option, for example, one week before implementation, the optimal option's implementation plan can be appropriately adjusted based on weather forecasts. For example, if heavy rainfall upstream is detected before water release, placing significant flow pressure on the downstream dam area, the optimal option's implementation date can be adjusted appropriately.

[0062] On the one hand, the present invention collects spawning and environmental data from the target area to build a machine learning model, and analyzes the importance of variables to accurately screen out the key environmental variables that have the greatest impact on fish spawning. A decision boundary diagram is then constructed based on the key variables, providing implementers with a flexible decision-making basis and technical support. By comparing the positions of different scheduling plans in the decision boundary diagram and the corresponding probability of occurrence of target events, implementers can quickly and intuitively select the most effective reservoir scheduling plan, intuitively presenting the favorable areas for fish spawning under different environmental conditions, and greatly improving the efficiency and scientific nature of scheduling plan formulation.

[0063] Another aspect of the present invention is that it fully considers the impact of environmental factors on water flow and temperature. By analyzing historical data on these factors, the optimal scheduling date is determined, preventing unexpected environmental conditions from interfering with the ecological scheduling plan. Furthermore, factors such as the dam's power generation efficiency are comprehensively considered when selecting the optimal solution. Before implementation, the solution can be adjusted based on the actual environment to ensure that ecological scheduling can effectively promote fish spawning while also addressing the reservoir's other functional requirements, thus guaranteeing the feasibility and effectiveness of the ecological scheduling solution in practical applications. Finally, before implementing the selected optimal solution, the implementation plan of the optimal solution is appropriately adjusted based on weather forecasts to ensure that the optimal solution can be effectively implemented.

[0064] In this embodiment, determining the optimal scheduling date includes the following steps:

[0065] The target variable of the ecological scheduling scheme is divided into a first variable and a second variable, a first judgment condition is determined based on the first variable, and a second judgment condition is determined based on the second variable.

[0066] The target variable affected by rainfall is set as the first variable.

[0067] A first time window is set, where the first time window is a time window for fish spawning. A first statistical value of a first variable within the first time window is determined based on historical data. A second time window is determined in the first time window based on the first statistical value. The first statistical value within the second time window meets a first judgment condition. A second statistical value of a second variable within the second time window is determined based on historical data. A third time window is determined in the second time window based on the second statistical value. The second statistical value within the third time window meets a second judgment condition. The third time window is defined as the optimal scheduling date.

[0068] When determining the primary and secondary variables, the variable most susceptible to sudden environmental impacts is set as the primary variable, and the variable least susceptible to sudden environmental impacts is set as the secondary variable. For example, for temperature and water flow, water flow is more susceptible to sudden rainfall, so it is set as the primary variable. Similarly, for water level and three-day water level rise, water level is more susceptible to sudden rainfall, so it is set as the primary variable.

[0069] The following steps are explained using water temperature and flow as examples. For water temperature, the first criterion is that the average daily water temperature for five consecutive days must fluctuate within 2°C of the temperature specified in the ecological scheduling plan. The first time window can be determined based on the fish spawning period, which runs from April to June each year. The first time window is then calculated by taking the average daily temperature over the past 10 years within the first time window and using this average as the first statistical value. For example, if the ecological scheduling plan specifies a water temperature of 24°C and the average daily water temperature between April 15th and 30th each year is found to be between 22°C and 26°C, then the second time window is set between April 15th and 30th.

[0070] For water flow, the second judgment condition is that the daily average water flow for 5 consecutive days should be within the range of 2000m³ / s specified in the ecological scheduling plan, and the probability of daily extreme values ​​is less than 20%. In this embodiment, the daily average water flow and the distribution probability of the extreme water flow on each date in the second time window are used as the second statistical value. If the water temperature is set at a water flow of 7500m³ / s in the ecological scheduling plan, and it is found that the daily average water flow between April 15 and April 20 each year is within the range of 5500m³ / s-9500m³ / s, and the probability of daily extreme values ​​is less than 20%, then April 15 to April 20 will be used as the third time window, and the third time window will be defined as the optimal scheduling date for the ecological scheduling plan.

[0071] In this embodiment, adjusting the optimal solution includes the following steps:

[0072] Locate multiple upstream points. The rainfall at the upstream points directly affects the inflow of the dam area. Predict the daily rainfall of each upstream point on the optimal scheduling date to obtain the predicted rainfall. Establish a conversion model. Based on the conversion model, convert the predicted rainfall to obtain the predicted inflow of each upstream point. Determine the inflow weight of each upstream point. Based on the inflow weight, perform weighted summation of the predicted inflow to obtain the predicted inflow of the dam area. If the daily predicted inflow on the optimal scheduling date is less than the critical threshold, implement the best plan on the optimal scheduling date. Otherwise, adjust the optimal scheduling date of the best plan.

[0073] If the optimal solution includes river flow or water level, the upstream inflow from the dam will affect these factors. The following method can be used to determine the upstream points: First, multiple upstream points are selected and the inflow weight of each upstream point is determined. If the inflow weight is small, the upstream point is deleted. In other words, the upstream points and the corresponding weights are obtained simultaneously. One week before the optimal scheduling date, the rainfall at each upstream point on the optimal scheduling date is predicted. The corresponding rainfall can be directly obtained from the relevant weather forecast.

[0074] A conversion model is then established to convert the predicted rainfall at each upstream point into a predicted inflow. Based on the previously determined inflow weights, the predicted inflows at each upstream point are weighted and summed to obtain the daily predicted inflow to the dam area. If the daily predicted inflow is low, it indicates that upstream rainfall will not affect the implementation of the optimal plan. This means that large upstream inflows will not cause forced releases from the dam, preventing planned flow and water level adjustments. Therefore, the implementation date of the optimal plan does not need to be changed. Otherwise, the implementation date of the optimal plan does need to be changed.

[0075] In this embodiment, the conversion model converts the predicted rainfall including the following steps.

[0076] The conversion model includes a first conversion function and a second conversion function, which determine the soil saturation value and basic flow of each upstream point. If the predicted rainfall is less than the soil saturation value, the predicted rainfall is converted into the first flow based on the first conversion function, and the sum of the first flow and the basic flow is used as the predicted inflow flow. Otherwise, the predicted rainfall is divided into a first part and a second part based on the soil saturation value, the first part is converted into the first inflow flow based on the first conversion function, and the second part is converted into the second inflow flow based on the second conversion function, and the sum of the first inflow flow, the second inflow flow and the basic flow is used as the predicted inflow flow.

[0077] Here is an example of a formula for the first conversion function and the second conversion function. In other embodiments, the first conversion function and the second conversion function can also use other functions, or use neural network models trained with different training sets. The first conversion function of this embodiment is, ,in, It is the rainfall conversion coefficient when the soil is not saturated, and its value is between 0.1 and 0.5, which is determined by experiments. is the predicted rainfall that does not exceed the soil saturation value. The second conversion function is, ,in, is the surface runoff coefficient after soil saturation, and its value is between 0.6 and 0.9. is the predicted rainfall above the soil saturation value. Base flow is the base flow of the river at the upstream point, which can be measured when there is no rainfall.

[0078] Specifically, determining the soil saturation value involves the following steps:

[0079] Obtain historical flow data of an upstream point, the historical flow data including historical rainfall and historical flow values, set a first saturation value, select two historical flow data as first data and second data respectively, calculate a corresponding first theoretical flow value based on the first data and the first saturation value, calculate a first difference between the first theoretical flow value and the corresponding historical flow value, obtain a second saturation value after adjusting the initial saturation value based on the first difference, calculate a second theoretical flow value based on the second data and the second saturation value, calculate a second difference between the second theoretical flow value and the corresponding historical flow value, and establish a linear function of the difference with respect to the soil saturation value based on the first difference and the second difference.

[0080] Based on the linear function, the soil saturation value corresponding to the difference value of 0 is determined and used as the third saturation value. The third difference corresponding to the third saturation value is calculated. A parabolic function is constructed based on the first difference, the second difference and the third difference. The soil saturation value corresponding to the difference value of 0 in the parabolic function is set as the soil saturation value.

[0081] This embodiment obtains historical flow data for a certain upstream point A, which includes the historical rainfall at the upstream point and the historical flow values ​​for the same period. When determining the soil saturation value, a numerical value is first set as the first saturation value based on experience. A piece of historical flow data is selected, and the historical rainfall therein is divided into a first part and a second part based on the first saturation value. The first and second conversion functions are used to calculate the first inflow flow and the second inflow flow, and these two are added to the base flow to obtain the first theoretical flow value. The first difference between the first theoretical flow value and the corresponding historical flow value is calculated. If the first theoretical flow value is less than the historical flow value, the first difference is a negative number, indicating that the calculated value is too small and the soil saturation value is set too high. If the second difference is a positive number, it indicates that the soil saturation value is set too low.

[0082] Based on the above calculation results, the soil saturation value is increased or decreased accordingly. A second historical flow data point is selected and a second theoretical flow value is calculated. This value is compared with the historical flow value to obtain a second difference. A coordinate system is then established with the soil saturation value as the horizontal axis and the difference as the vertical axis. Two coordinate points are marked within the system: the first coordinate point represents the first difference and the corresponding first saturation value, and the second coordinate point represents the second difference and the corresponding second saturation value. The corresponding linear function can be obtained based on the coordinates of these two coordinate points.

[0083] The linear function must exist at the point where the horizontal axis intersects. The horizontal coordinate of this point is the third saturation value. Select another piece of historical flow data, calculate the third difference corresponding to the data based on the third saturation value, and mark the third difference and the third saturation value in the coordinate system in the form of the third coordinate point. Finally, use a parabola to fit the first coordinate point, the second coordinate point, and the third coordinate point. The point where the fitted function intersects the horizontal axis of the coordinate system is the soil saturation value to be obtained.

[0084] In this embodiment, determining the import weight includes the following steps:

[0085] The water flow arrival time is set based on the distance between the upstream point and the target area, and the historical rainfall data of the upstream point in the historical time period is obtained. The time period when the historical rainfall data appears is defined as the first time period. The second time period when the rainfall flow reaches the dam area is determined based on the water flow arrival time.

[0086] The following example illustrates this: upstream points A and B exist, with the corresponding dam area being downstream point Q. Rainfall at upstream point A will affect the river flow at downstream point Q one hour later. Rainfall at upstream point B will affect the river flow at downstream point Q two hours later. To determine the weights, we need to find a time period in the historical data where both upstream points A and B experienced rainfall. This is called the first time period. The second time period is the hour following the first time period.

[0087] The first time period is divided into multiple first sub-time periods, and the second time period is correspondingly divided into multiple second sub-time periods. The inflow flow of the upstream point in the first sub-time period and the inflow flow of the dam area in each second sub-time period are obtained. An initial weight is set for each upstream point. The inflow flow of each first sub-time period is weighted and summed based on the initial weight to obtain multiple influencing quantities. A first sequence is constructed based on the influencing quantities, and a second sequence is constructed based on the inflow flow.

[0088] Upstream points A and B experience rainfall simultaneously between 11:00 and 13:00, corresponding to a second time period of 12:00 to 14:00. The first time period is split into multiple first sub-time periods at one-hour intervals, and the second sub-time period is split into second sub-time periods. The inflow from upstream points A and B is then obtained for each first sub-time period. The initial weights for both upstream points A and B are set to 0.5, and the weighted sum is calculated to obtain the impact. Since there are three first sub-time periods, the first sequence has three values. Similarly, the second sequence, constructed based on the inflow from downstream point Q, also includes three values.

[0089] Perform correlation analysis on the first and second sequences to obtain the correlation coefficient. Adjust the initial weight and calculate the correlation coefficient again. After repeating this process multiple times, use the initial weight corresponding to the maximum correlation coefficient as the corresponding import weight.

[0090] Calculate the Pearson correlation coefficient between the first and second sequences. A larger Pearson correlation coefficient indicates a stronger correlation between the two sequences, which in turn indicates a more appropriate initial weight setting. After the first calculation is completed, adjust the initial weights of upstream points A and B to 0.4 and 0.6. Then, regenerate the first and second sequences and calculate the corresponding Pearson correlation coefficients. Repeat this iteration multiple times, and use the initial weight corresponding to the maximum Pearson correlation coefficient as the inflow weight for upstream points A and B.

[0091] In this embodiment, selecting an ecological scheduling scheme as the optimal scheme includes the following steps:

[0092] Calculate the power generation benefit of each ecological scheduling scheme, and select the ecological scheduling scheme with the highest power generation benefit as the best scheme.

[0093] This implementation calculates the dispatch benefit based on the following method. First, the optimal dispatch date corresponding to the ecological dispatch scheme is divided into multiple time periods. The power generation benefit of the ecological dispatch scheme is calculated based on the first formula. The first formula is: ,in, For power generation efficiency, is the total number of time periods included, For the The power generation capacity in each time period is determined by the turbine efficiency, water head height and discharge flow. is the electricity price in the tth time period, The length of a single period.

[0094] like Figure 6 As shown, the present invention also provides a reservoir ecological scheduling system based on a classification algorithm and a decision boundary diagram, which is used to implement the above-mentioned reservoir ecological scheduling method based on a classification algorithm and a decision boundary diagram. The system includes:

[0095] The preprocessing module collects spawning data and environmental data from the target area, builds a machine learning model with environmental data as the independent variable and spawning data as the dependent variable, and performs importance analysis on the independent variables based on the output of the machine learning model to obtain the variable importance of each independent variable;

[0096] The construction module selects the top two independent variables with the highest importance as target variables, constructs a grid dataset about the target variables, constructs a decision boundary graph based on the machine learning model and the grid dataset, constructs multiple ecological scheduling schemes based on the decision boundary graph, and obtains the environmental scheduling value of each ecological scheduling scheme;

[0097] The optimization module obtains the factors affecting the environmental scheduling value, obtains the historical data of the influencing factors in the target area, and analyzes the historical data to determine the optimal scheduling date for implementing each ecological scheduling plan;

[0098] The adjustment module selects an ecological scheduling scheme as the best scheme based on the best scheduling date. Before implementing the best scheme, if the environmental scheduling value is affected by the environmental rainfall, the best scheduling date is adjusted based on the actual environment of the target area.

[0099] It should be understood that the various technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the various technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0100] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included outside the scope of protection of the present invention.

Claims

1. A reservoir ecological operation method based on classification algorithm and decision boundary diagram, characterized in that: Collecting spawning data and environmental data from a target area, and building a machine learning model using the environmental data as an independent variable and the spawning data as a dependent variable; Performing importance analysis on the independent variables based on the output results of the machine learning model to obtain the variable importance of each independent variable; Selecting the first two independent variables with the highest importance as target variables, constructing a grid data set about the target variables, and constructing a decision boundary graph based on the machine learning model and the grid data set; Constructing multiple ecological scheduling schemes based on the decision boundary graph, and obtaining an environmental scheduling value for each of the ecological scheduling schemes; Obtaining influencing factors of the environmental scheduling value, obtaining historical data of the influencing factors in the target area, and analyzing the historical data to determine the optimal scheduling date for implementing each of the ecological scheduling plans; selecting one of the ecological scheduling schemes as an optimal scheme based on the optimal scheduling date, and before implementing the optimal scheme, if the environmental scheduling value is affected by environmental rainfall, adjusting the optimal scheduling date based on the actual environment of the target area; Determining the optimal dispatch date includes the following steps: Dividing the target variable of the ecological scheduling plan into a first variable and a second variable, determining a first judgment condition based on the first variable, and determining a second judgment condition based on the second variable; Setting a first time window, where the first time window is a time window for fish spawning, determining a first statistical value of the first variable within the first time window based on the historical data, determining a second time window within the first time window based on the first statistical value, where the first statistical value within the second time window satisfies the first judgment condition, determining a second statistical value of the second variable within the second time window based on the historical data, determining a third time window within the second time window based on the second statistical value, where the second statistical value within the third time window satisfies the second judgment condition, and defining the third time window as the optimal scheduling date; Adjusting the optimal solution includes the following steps: Locate multiple upstream points, where the rainfall at the upstream points directly affects the inflow of the dam area, predict the daily rainfall at each upstream point within the optimal scheduling date to obtain the predicted rainfall, establish a conversion model, convert the predicted rainfall based on the conversion model to obtain the predicted inflow of each upstream point, determine the inflow weight of each upstream point, perform weighted summation of the predicted inflow based on the inflow weight, and obtain the predicted inflow of the dam area; if the predicted inflow of each day within the optimal scheduling date is less than a critical threshold, implement the optimal plan on the optimal scheduling date; otherwise, adjust the optimal scheduling date of the optimal plan.

2. The method according to claim 1, characterized in that The conversion model converts the predicted rainfall and comprises the following steps: The conversion model includes a first conversion function and a second conversion function, which determine the soil saturation value and basic flow of each upstream point. If the predicted rainfall is less than the soil saturation value, the predicted rainfall is converted into a first flow based on the first conversion function, and the sum of the first flow and the basic flow is used as the predicted inflow flow. Otherwise, the predicted rainfall is divided into a first part and a second part based on the soil saturation value, the first part is converted into a first inflow flow based on the first conversion function, and the second part is converted into a second inflow flow based on the second conversion function, and the sum of the first inflow flow, the second inflow flow and the basic flow is used as the predicted inflow flow.

3. The method according to claim 1, characterized in that Determining the inflow weight includes the following steps: setting a water flow arrival time based on the distance between the upstream point and the target area, obtaining historical rainfall data of the upstream point in a historical time period, defining a time period in which the historical rainfall data appears as a first time period, and determining a second time period in which the rainfall flow arrives at the dam area based on the water flow arrival time; Splitting the first time period into a plurality of first sub-time periods, and correspondingly splitting the second time period into a plurality of second sub-time periods, obtaining the inflow flow of the upstream point in the first sub-time period, and the inflow flow of the dam area in each of the second sub-time periods, setting an initial weight for each upstream point, and weighted summing the inflow flow of each of the first sub-time periods based on the initial weight to obtain a plurality of influencing quantities, constructing a first sequence based on the influencing quantities, and constructing a second sequence based on the inflow flow; Performing a correlation analysis on the first sequence and the second sequence to obtain a correlation coefficient, adjusting the initial weight and then recalculating the correlation coefficient, repeating the process multiple times, and using the initial weight corresponding to the maximum correlation coefficient as the corresponding import weight.

4. The method according to claim 1, wherein The target variable affected by rainfall is set as the first variable.

5. The method according to claim 1, wherein Selecting one of the ecological scheduling schemes as the best scheme includes the following steps: The power generation benefit of each ecological scheduling scheme is calculated, and the ecological scheduling scheme with the highest power generation benefit is selected as the optimal scheme.

6. The method according to claim 1, characterized in that The variable importance is the SHAP value of each of the environmental data.

7. The method according to claim 1, characterized in that The environmental data includes water temperature, water flow, transparency, water level and sediment content, and the machine learning model is a binary tree machine learning model.

8. A reservoir ecological operation system based on a classification algorithm and a decision boundary diagram, used to implement a reservoir ecological operation method based on a classification algorithm and a decision boundary diagram as claimed in any one of claims 1 to 7, characterized in that: The system includes: a preprocessing module that collects spawning data and environmental data from a target area, constructs a machine learning model using the environmental data as an independent variable and the spawning data as a dependent variable, and performs importance analysis on the independent variables based on outputs of the machine learning model to obtain the variable importance of each independent variable; A construction module is provided for selecting the first two independent variables with the highest importance as target variables, constructing a grid data set for the target variables, constructing a decision boundary graph based on the machine learning model and the grid data set, constructing multiple ecological scheduling schemes based on the decision boundary graph, and obtaining an environmental scheduling value for each of the ecological scheduling schemes; an optimization module for obtaining influencing factors of the environmental scheduling value, obtaining historical data of the influencing factors in the target area, and analyzing the historical data to determine an optimal scheduling date for implementing each of the ecological scheduling plans; An adjustment module selects one of the ecological scheduling schemes as the optimal scheme based on the optimal scheduling date, and before implementing the optimal scheme, if the environmental scheduling value is affected by environmental rainfall, adjusts the optimal scheduling date based on the actual environment of the target area.

Images