Ecological scheduling scheme optimization simulation coupling decision-making method based on artificial intelligence

Abstract

The invention discloses an ecological scheduling scheme optimization simulation coupling decision-making method based on artificial intelligence. The method comprises the steps of determining ecological scheduling starting time and ecological scheduling key parameters; determining a suitable interval of ecological scheduling key parameters required by fish breeding; constructing a reservoir multi-objective optimization ecological scheduling model for promoting natural breeding of fishes by applying an optimization algorithm; solving an optimization scheme of reservoir ecological scheduling in different scenes; arranging the optimal scheduling scheme to form a scheduling track sequence data set; constructing a reservoir ecological scheduling scheme simulation decision model based on a long short-term memory network deep learning algorithm; constructing a loss function with a physical mechanism; and taking the scheduling track sequence data set as training data, and autonomously learning an ecological optimal scheduling rule of the reservoir from the data set. According to the method, the ecological scheduling outflow scheme for various actual working condition scenes can be quickly given, the optimization decision of the ecological scheduling scheme is assisted, and the ecological scheduling scheme is improved from a static state to a real-time dynamic state.

Description

Technical Field

[0001] The present invention belongs to the technical field of reservoir operation, and specifically relates to an optimization simulation coupling decision-making method for ecological operation schemes based on artificial intelligence. Background Art

[0002] While fully exerting the beneficial effects of water conservancy and hydropower projects, it is inevitable to change the natural hydrological regime and sediment process of rivers, driving changes in environmental states such as river hydrology and hydrodynamics, and thus destroying aquatic organisms, especially the original growth conditions of fish, and affecting the natural reproduction of fish. To address the above negative ecological impacts brought about by the construction of water conservancy and hydropower projects and coordinate the contradictory relationship between the social and economic benefits of the projects and the ecological environment, the concept of ecological operation was proposed.

[0003] The operation model is an important prerequisite for supporting the formulation and optimization decision-making of ecological operation schemes. Conventional operation models mainly utilize runoff regulation theory and hydropower calculation methods, and have problems such as over-reliance on the experience of operation personnel and unsatisfactory operation effects. Therefore, conventional operation models have gradually been replaced by optimized operation models. Optimized operation models are good at formulating the optimal reservoir outflow scheme according to elements such as hydrology, water conservancy, and hydropower to meet the water use requirements of different objectives to the greatest extent, effectively improving the operation effect. However, they also have problems such as the curse of dimensionality and poor real-time performance. With the rapid development of artificial intelligence algorithms, the research on exploring the advantages of coupling optimization algorithms and simulation algorithms to solve reservoir operation schemes has gradually become a research hotspot in the discipline. However, how to couple the advantages of multi-objective optimal operation and deep learning algorithms and explore the improvement of ecological operation schemes from static to real-time dynamic is a blank field in current research and also a technical bottleneck that urgently needs to be broken through for optimizing ecological operation effects. Summary of the Invention

[0004] In view of the above problems, the present invention is proposed to provide an optimization simulation coupling decision-making method for ecological operation schemes based on artificial intelligence that overcomes the above problems or at least partially solves the above problems.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] An optimization simulation coupling decision-making method for ecological operation schemes based on artificial intelligence, the method comprising the following steps:

[0007] S1. Determine the starting time of ecological operation and the key parameters of ecological operation;

[0008] S2. Establish the response relationship between the spawning scale of the target fish for operation and the key parameters of ecological operation, and determine the appropriate interval of the key parameters of ecological operation required for fish reproduction;

[0009] S3. Apply an optimization algorithm to construct a multi-objective optimal ecological operation model of the reservoir to promote natural fish reproduction;

[0010] S4. Design typical operating condition scenarios, apply the reservoir optimal operation model for promoting natural fish reproduction, and solve the optimal ecological operation plans for the reservoir under different scenarios;

[0011] S5. Organize the optimal operation plans to form a dataset of operation trajectory sequences;

[0012] S6. Build a simulation decision model for the reservoir ecological operation plan based on the deep learning algorithm of long short-term memory network;

[0013] S7. Construct a loss function with physical mechanism when the ecological operation discharge process exceeds the appropriate interval of key parameters or violates the water balance;

[0014] S8. Use the dataset of operation trajectory sequences as training data to autonomously learn the optimal ecological operation rules of the reservoir from the dataset.

[0015] Optionally, in step S1, determine the starting time of ecological operation according to the historical spawning time and spawning water temperature requirements of fish.

[0016] Optionally, in step S2, the key parameters of ecological operation required for fish reproduction include the continuous rising water time, initial rising water flow, peak flood flow, and daily flow growth rate.

[0017] Optionally, in step S3, the multi-objective optimized ecological operation model takes into account power generation and ecological requirements, and takes the maximum power generation and the highest ecological satisfaction degree as the optimization operation objectives. The objective function expression is as follows:

[0018]

[0019] S = max(S d + S o + S p + S v )

[0020] In the formula, E represents the total power generation of the reservoir during the ecological operation period T, with the unit of kW·h; k represents the comprehensive output coefficient of the hydropower station; Q t represents the discharge of the hydropower station at time t, with the unit of m 3 / s; H t represents the average power generation head of the hydropower station at time t, with the unit of m; Δt is the time length of the calculation period; S represents the total suitability degree of the reservoir discharge process for fish reproduction; S d , S o , S p and S v respectively represent the suitability degrees of the rising water duration, initial rising water flow, peak flood flow, and daily flow growth rate.

[0021] Optionally, the optimization algorithm in step S3 includes the second-generation non-dominated sorting genetic algorithm and the ant colony algorithm.

[0022] Optionally, the typical operating condition scenarios in step S4 include the historical inflow process or the runoff process at the dam site, typical wet, normal, and dry years, and extreme droughts and floods.

[0023] Optionally, in step S6, the expression of the simulation decision model for the reservoir ecological operation plan is:

[0024] Q O t = f(t, Tem, Q i t , Q O t-1 , Z t , Z t-1 )

[0025] Where: Q O t is the reservoir discharge at the current moment, f is the scheduling rule function of the reservoir optimized ecological operation plan, t is the date, Tem is the water temperature, Q i t is the reservoir inflow in the current period, Q O t-1 is the reservoir discharge in the previous period, Z t is the reservoir operating level in the current period, Z t-1 is the reservoir operating level in the previous period.

[0026] Optionally, in step S6, the evaluation indexes of the calculation accuracy of the simulation decision model for the reservoir ecological operation plan adopt the ratio of root mean square error to standard deviation RSR and the Nash efficiency coefficient NSE, where RSR and NSE are calculated by the following formula:

[0027]

[0028] Where, s i represents the model simulation result, o i represents the observed result, represents the average observed value; if RSR < 0.5 and NSE > 0.75, it means that the simulation decision model of the ecological operation plan meets the accuracy requirements.

[0029] Optionally, in step S7, the calculation formula of the loss function with physical mechanism is:

[0030] Loss = w o Loss o + w b Loss b + w h Loss h

[0031] In the formula, Loss is a loss function with a physical mechanism, and Loss o is the root mean square error between the target out - flow and the out - flow simulated by the model. Loss b is the penalty function for the reservoir out - flow violating the water balance constraint. Loss h is the penalty function for the reservoir out - flow exceeding the appropriate interval of the key parameters. w o 、w b 、w h are the weight coefficients of Loss o 、Loss b 、Loss h respectively.

[0032] Optionally, in step S8, 80% of the scheduling trajectory sequence dataset is used as training data to autonomously learn the optimal scheduling rules for the reservoir ecology from the dataset, and the remaining 20% is used as test data to test the model performance and obtain the decision result of the reservoir ecological scheduling scheme.

[0033] In summary, due to the adoption of the above - mentioned technical solutions, the beneficial effects of the present invention are as follows:

[0034] 1. Based on the LSTM algorithm in the field of deep learning, the present invention learns the scheduling rules from the ecological scheduling optimization scheme trajectory. Compared with traditional machine - learning models such as the back - propagation neural network (BP) and support vector regression (SVR), the deep - learning model has stronger learning ability, can effectively extract the latent scheduling rules behind the data, and give a more accurate ecological scheduling out - flow process scheme;

[0035] 2. The present invention constructs a penalty function for the reservoir out - flow violating the water balance constraint and a penalty function for the reservoir out - flow exceeding the appropriate interval of the key parameters. Compared with the traditional loss function that only considers the prediction error, this method can ensure that the decision result of the ecological scheduling scheme is more in line with the physical reality;

[0036] 3. By coupling the advantages of the optimization scheduling and the deep - learning algorithm, the present invention can use the optimization algorithm to solve the ecological scheduling optimization scheme under various historical or design scenarios, use the deep - learning algorithm to propose the latent scheduling rules behind the scheme, quickly give the ecological scheduling out - flow scheme for various actual working condition scenarios, assist in the optimization decision of the ecological scheduling scheme, and realize the improvement of the ecological scheduling scheme from static to real - time dynamic. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is a schematic flow chart of an ecological scheduling scheme optimization simulation coupling decision method provided by an embodiment of the present application;

[0038] Figure 2 The suitability curve of the key parameters for ecological regulation provided by the embodiments of the present application;

[0039] Figure 3 The solution results of the ecological regulation optimization scheme for some years;

[0040] Figure 4 The comparison chart of the optimization and simulation decision results of the ecological regulation scheme. Specific implementation manners

[0041] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0042] The Three Gorges Reservoir is located on the main stream of the Yangtze River and is the last stage of the controlled cascade reservoirs on the main stream of the upper reaches of the Yangtze River. Research shows that the operation of the Three Gorges Reservoir inevitably has many impacts on the Yangtze River ecosystem. As a representative species of the middle and lower reaches of the Yangtze River ecosystem, the four major Chinese carps are also species that are greatly affected by the operation of the Three Gorges Reservoir. Since 2011, the Three Gorges Reservoir has carried out ecological regulation experiments for 12 consecutive years for the natural reproduction of fish species that produce drifting eggs such as the four major Chinese carps downstream. This embodiment takes the Three Gorges Reservoir as the implementation object and details the compilation and real-time optimization decision-making process of the reservoir ecological regulation scheme for promoting the natural reproduction of fish species that produce drifting eggs by using the steps proposed in this embodiment.

[0043] Please refer to Figures 1-4 , this embodiment provides an artificial intelligence-based optimization simulation coupling decision-making method for ecological regulation schemes, and the method includes the following steps:

[0044] S1. Determine the starting time of ecological regulation and the key parameters of ecological regulation.

[0045] In step S1, the starting time of ecological regulation is determined according to the historical spawning time and spawning water temperature requirements of fish.

[0046] According to the breeding habits of fish species that produce drifting eggs in the Yangtze River Basin, their spawning time is generally from May to July every year. When the water temperature reaches 18°C, preferably greater than 20°C, combined with the incoming water conditions, ecological regulation is carried out at an opportune moment.

[0047] S2. Establish the response relationship between the spawning scale of the target fish for regulation and the key parameters of ecological regulation, and determine the appropriate interval of the key parameters of ecological regulation required for fish reproduction.

[0048] In step S2, according to the results of literature research, the key ecological operation parameters affecting the natural reproduction of fish species that produce drifting eggs such as the four major Chinese carps were screened out. The key ecological operation parameters required for fish reproduction include the duration of continuous rising water, the initial rising water flow rate, the peak flood flow rate, and the daily growth rate of the flow rate.

[0049] And the suitability curves of the above key parameters were plotted, as shown in the appendix Figure 2 .

[0050] Regarding the duration of rising water, Yibolu et al. (1988) believed through investigation that the four major Chinese carps start spawning about 2 days after the river water begins to rise; the spawning duration is more than 4 days, ranging from 4 to 18 days; Li Bo et al. (2021) believed that maintaining the rising water time for about 4 days is beneficial to the reproductive activities of the four major Chinese carps; Wang Junna et al. (2012) proposed through research that the suitable duration of rising water is 5 - 8 days; therefore, in the present invention, the suitable range of the duration of rising water is 2 - 8 days, and the most suitable range is 4 - 8 days.

[0051] Regarding the initial rising water flow rate, Li Bo et al. (2021) believed through research that the range of the initial rising water flow rate suitable for the reproduction of the four major Chinese carps is 6400 - 18900 m 3 / s, Li Chaoda et al. (2021) believed that the suitable initial flow rate is 7030 - 16200 m 3 / s, Xu Wei et al. (2020) believed that reaching 14000 m 3 / s is suitable; therefore, in this patent, the suitable range of the initial flow rate of the rising water section is 6400 - 18900 m 3 / s, and the most suitable range is 14000 - 16200 m 3 / s.

[0052] Regarding the peak flood flow rate, Li Bo et al. (2021) believed that the flow rate distribution range during spawning is 9400 - 38575 m 3 / s, the peak flood flow rate suitable for the reproduction of the four major Chinese carps is 16425 - 32300 m 3 / s, Wang Yue et al. (2017) believed that the peak flood flow rate should be greater than 20000 m 3 / s; Liu Han et al. (2023) believed that the most suitable peak flood flow rate is 19610 m 3 / s; therefore, in this patent, the suitable range of the peak flood flow rate of the rising water section is 9400 - 38575 m 3 / s, and the most suitable range is 20000 - 32300 m 3 / s.

[0053] Regarding the daily growth rate of the flow rate, Wang Junna et al. (2012) proposed through research that the daily growth rate of the flow rate is higher than 900 m 3 / s·d is beneficial to fish reproduction. Wang Yue et al. (2017) believe that the appropriate daily growth rate range is 1295 - 2825 m 3 / s·d, and Li Bo et al. (2021) believe that the appropriate daily growth rate range is 1925 - 4533 m 3 / s·d. Therefore, this patent takes 900 - 4533 m 3 / s·d as the appropriate range of daily flow growth rate, and 1925 - 2825 m 3 / s·d as the optimal range.

[0054] S3. Construct a multi-objective optimal ecological operation model of the reservoir to promote natural fish reproduction by using an optimization algorithm.

[0055] The optimization algorithm in step S3 includes the second-generation non-dominated sorting genetic algorithm and the ant colony algorithm. Taking the key parameters reaching the optimal range as the ecological operation goal, the second-generation non-dominated sorting genetic algorithm is applied to construct a multi-objective optimal ecological operation model of the reservoir to promote natural fish reproduction.

[0056] In step S3, the multi-objective optimal ecological operation model takes into account power generation and ecological needs, and takes the maximum power generation and the highest ecological satisfaction degree as the optimal operation goals. The objective function expression is as follows:

[0057]

[0058] S = max(S d + S o + S p + S v )

[0059] In the formula, E represents the total power generation of the reservoir during the ecological operation period T, with the unit of kW·h; k represents the comprehensive output coefficient of the hydropower station; Q t represents the downstream discharge of the hydropower station at time t, with the unit of m 3 / s; H t represents the average power generation head of the hydropower station at time t, with the unit of m; Δt is the time length of the calculation period; S represents the total suitability of the reservoir discharge process for fish reproduction; S d , S o , S p and S v respectively represent the suitability of the rising water duration, the initial rising water flow, the flood peak flow and the daily flow growth rate.

[0060] The calculation formulas for each index are as follows:

[0061]

[0062] In the formula, sgn() is a step function, Q t+1 - Qt > 0, sgn(Q t+1 -Q t ) = 1, Q t+1 -Q t = 0, sgn(Q t+1 -Q t ) = 0, Q t+1 -Q t <0, sgn(Q t+1 -Q t ) = -1, Q t+1 is the discharge of the hydropower station at time t + 1.

[0063]

[0064] In the formula, Q 1 represents the initial flow rate.

[0065]

[0066] In the formula, max Q t represents the maximum flow rate during the ecological operation, where t = 1, 2,..., T.

[0067]

[0068] In the formula, ΔQ represents the daily increase rate of the flow rate during the ecological operation, and the calculation formula is:

[0069] ΔQ = (max Q t -Q 1 ) / (t - 1).

[0070] The operation constraints mainly consider the conventional operation constraints of the reservoir.

[0071] ① Water balance constraint

[0072] V t+1 = V t + (I t -Q t )·Δt

[0073] In the formula, V t and V t+1 are the water storage volumes of the reservoir at times t and t + 1 respectively, I t is the inflow rate of the reservoir at time t, and Q t is the outflow rate of the reservoir at time t. The evaporation loss of the reservoir is ignored in the water balance constraint.

[0074] ② Reservoir water storage constraint

[0075]

[0076] In the formula, is the water storage of the reservoir at time t.

[0077] ③ Reservoir discharge flow constraint

[0078]

[0079] In the formula, are the minimum and maximum discharge capacities of the reservoir at time t.

[0080] ④ Hydropower station output constraint

[0081]

[0082] In the formula, are the minimum and maximum outputs of the power station units at time t.

[0083] ⑤ Reservoir water level constraint

[0084] Reservoir water level constraint:

[0085]

[0086] Daily decline of the water level of the Three Gorges Reservoir:

[0087] ΔZ d ≤0.6m

[0088] In the formula, are the lowest and highest reservoir water levels. ΔZ d is the daily variation range of the water level of the Three Gorges Reservoir.

[0089] S4. Design typical operating condition scenarios, apply the reservoir optimal operation model for promoting natural fish reproduction, and solve the optimal ecological operation schemes of the reservoir under different scenarios.

[0090] The typical operating condition scenarios in step S4 include historical inflow processes or runoff processes at the dam site, typical wet, normal, and dry years, and various operating condition scenarios faced during reservoir operation such as extreme droughts and floods.

[0091] Based on the long-term historical runoff and water temperature process data from May to July during 1962 - 2022 at the dam site of the Three Gorges Reservoir, apply the multi-objective optimal ecological operation model of the reservoir for promoting natural fish reproduction, and solve to generate the optimal ecological operation schemes of the reservoir. The solution results of the optimal ecological operation schemes for some years are shown in the appendix Figure 3 The multi-objective optimal ecological operation model of the reservoir for promoting natural fish reproduction can reasonably select the operation timing according to the reservoir inflow conditions and give an ecological operation scheme with a continuous rising water process.

[0092] S5. Organize the optimal operation schemes to form a dataset of operation trajectory sequences.

[0093] Sort out the optimal scheduling plan to form a scheduling trajectory sequence data set containing the date t, water temperature Tem, current period inflow Q of the reservoir i t , current moment outflow Q of the reservoir O t , outflow Q of the previous period of the reservoir O t-1 , current period operating water level Z of the reservoir t , operating water level Z of the previous period of the reservoir t-1 .

[0094] S6. Construct a simulation decision model for the reservoir ecological scheduling plan based on the long short-term memory network deep learning algorithm.

[0095] Taking the current moment outflow Q of the reservoir O t as the model output variable and the other factors as the input variables, construct a simulation decision model for the reservoir ecological scheduling plan based on the long short-term memory network deep learning algorithm.

[0096] In step S6, the expression of the simulation decision model for the reservoir ecological scheduling plan is:[[]]

[0097] Q O t = f(t, Tem, Q i t , Q O t-1 , Z t , Z t-1 )

[0098] In the formula: Q O t is the current moment outflow of the reservoir, f is the scheduling rule function of the reservoir optimized ecological scheduling plan, t is the date, Tem is the water temperature, Q i t is the current period inflow of the reservoir, Q O t-1 is the outflow of the previous period of the reservoir, Z t is the current period operating water level of the reservoir, Z t-1 is the operating water level of the previous period of the reservoir.

[0099] In step S6, the evaluation indexes of the calculation accuracy of the simulation decision model for the reservoir ecological scheduling plan adopt the ratio RSR of the root mean square error to the standard deviation and the Nash efficiency coefficient NSE, where RSR and NSE are calculated by the following formula:[[]]

[0100]

[0101]

[0102] In the formula, s i represents the model simulation result, o i represents the observation result, represents the average observed value; if RSR < 0.5 and NSE > 0.75, it indicates that the ecological regulation scheme simulation decision model meets the accuracy requirements.

[0103] The test results are as shown in the appendix Figure 4 The prediction accuracy index of the model, RSR is equal to 0.26, NSE is equal to 0.93, and the calculation time is 0.52 s. It can give real-time ecological regulation optimization decision-making schemes for various working conditions.

[0104] S7. Construct a loss function with physical mechanism based on the fact that the ecological regulation discharge process exceeds the appropriate interval of key parameters or violates the water balance.

[0105] In step S7, the calculation formula of the loss function with physical mechanism is:

[0106] Loss = w o Loss o + w b Loss b + w h Loss h

[0107] In the formula, Loss is the loss function with physical mechanism, Loss o is the root mean square error between the target discharge flow and the model-simulated discharge flow, Loss b is the penalty function for the reservoir discharge flow violating the water balance constraint, Loss h is the penalty function for the reservoir discharge flow exceeding the appropriate interval of key parameters, w o 、w b 、w h are the weight coefficients of Loss o 、Loss b 、Loss h respectively.

[0108] For this embodiment, to ensure the model accuracy and at the same time ensure that the outflow process meets the water balance and the requirements of key parameters for fish spawning, w o 、w b 、w h are respectively set to 0.3, 0.4, and 0.3.

[0109] S8. Use the scheduling trajectory sequence dataset as the training data to autonomously learn the optimal scheduling rules of the reservoir ecology from the dataset.

[0110] In step S8, 80% of the scheduling trajectory sequence dataset is used as training data to autonomously learn the optimal scheduling rules for the reservoir ecosystem from the dataset, and the remaining 20% is used as test data to test the model performance and obtain the decision result of the reservoir ecological scheduling scheme.

[0111] In this embodiment, the LSTM algorithm in the field of deep learning is used to learn the scheduling rules from the ecological scheduling optimization scheme trajectory. Compared with traditional machine learning models such as the backpropagation neural network (BP) and support vector regression (SVR), the deep learning model has stronger learning ability, can effectively extract the latent scheduling rules behind the data, and give a more accurate ecological scheduling outflow process scheme.

[0112] In this embodiment, a penalty function for the reservoir outflow violating the water balance constraint and a penalty function for the reservoir outflow exceeding the appropriate interval of the key parameters are constructed. Compared with the traditional loss function that only considers the prediction error, this method can ensure that the decision result of the ecological scheduling scheme is more in line with the physical reality.

[0113] In this embodiment, by coupling the advantages of the optimization scheduling and the deep learning algorithm, the optimization algorithm can be used to solve the ecological scheduling optimization scheme under various historical or design scenarios, and the deep learning algorithm can be used to propose the latent scheduling rules behind the scheme, quickly give the ecological scheduling outflow scheme for various actual working condition scenarios, assist in the optimization decision of the ecological scheduling scheme, and realize the improvement of the ecological scheduling scheme from static to real-time dynamic.

[0114] The above content is a further detailed description of the present invention in combination with specific preferred embodiments. It cannot be determined that the specific implementation of the present invention is only limited to these descriptions. For those of ordinary skill in the technical field to which the present invention belongs, without departing from the concept of the present invention, several simple deductions or substitutions can still be made, which should all be regarded as belonging to the protection scope of the present invention.

Claims

1. An artificial intelligence-based ecological scheduling scheme optimization simulation coupling decision-making method, characterized in that: The method comprises the following steps: S1. Determine the starting time and key parameters of ecological scheduling; S2. Establish the response relationship between the spawning scale of the target fish species and the key parameters of ecological scheduling, and determine the appropriate range of the key parameters of ecological scheduling required for fish reproduction; S3. Apply optimization algorithms to construct a multi-objective optimization ecological scheduling model for reservoirs that promote natural reproduction of fish; S4. Design typical operating scenarios, apply the reservoir optimization scheduling model that promotes natural fish reproduction, and solve the optimization scheme of reservoir ecological scheduling under different scenarios; S5, sort out the optimal scheduling plan to form a scheduling trajectory sequence data set; S6. Construct a simulation decision model for reservoir ecological dispatching scheme based on long short-term memory network deep learning algorithm; S7. Construct a loss function with a physical mechanism based on the ecological dispatching discharge process that exceeds the appropriate range of key parameters or violates the water balance; S8. Use the scheduling trajectory sequence dataset as training data and autonomously learn the optimal scheduling rules for reservoir ecology from the dataset.

2. The method for optimizing and simulating coupled decision-making of ecological scheduling scheme based on artificial intelligence according to claim 1, characterized in that: In step S1, the ecological scheduling start time is determined according to the historical spawning time of fish and the spawning water temperature requirements.

3. The method for optimizing simulation coupling decision-making of ecological scheduling scheme based on artificial intelligence according to claim 1 is characterized in that: In step S2, the key parameters of ecological scheduling required for fish reproduction include the duration of flooding, the initial flow of flooding, the peak flow and the daily growth rate of flow.

4. The artificial intelligence-based ecological scheduling scheme optimization simulation coupling decision-making method according to claim 1 is characterized in that: In step S3, the multi-objective optimization ecological scheduling model takes into account both power generation and ecological needs, and takes the maximum power generation and the highest ecological satisfaction as the optimization scheduling objectives. The objective function expression is as follows: S=max(S d +S o +S p +S v ) Where, E represents the total power generation of the reservoir during the ecological dispatch period T, in kW h; k represents the comprehensive output coefficient of the hydropower station; Q t It represents the discharge flow of the hydropower station in period t, in m 3 / s;H t represents the average power generation head of the hydropower station during period t, in m; Δt is the length of the calculation period; S represents the overall suitability of the reservoir discharge process for fish reproduction; S d , S o , S p and S v They respectively represent the suitability of flood duration, initial flood flow, peak flow and daily flow growth rate.

5. The artificial intelligence-based ecological scheduling scheme optimization simulation coupling decision-making method according to claim 1, characterized in that: The optimization algorithms in step S3 include a second generation non-dominated sorting genetic algorithm and an ant colony algorithm.

6. The artificial intelligence-based ecological scheduling scheme optimization simulation coupling decision-making method according to claim 1, characterized in that: The typical operating conditions in step S4 include historical inflow processes or runoff processes at the dam site, typical years of wet and dry seasons, and extreme droughts and floods.

7. The artificial intelligence-based ecological scheduling scheme optimization simulation coupling decision-making method according to claim 1 is characterized in that: In step S6, the expression of the reservoir ecological dispatching scheme simulation decision model is: Q O t =f(t,Tem,Q i t ,Q O t-1 ,Z t ,Z t-1 ) Where: Q O t is the outflow of the reservoir at the current moment, f is the dispatching rule function of the reservoir optimization ecological dispatching plan, t is the date, Tem is the water temperature, Q i t is the inflow of the reservoir in the current period, Q O t-1 is the outflow of the reservoir in the previous period, Z t is the operating water level of the reservoir in the current period, Z t-1 It is the operating water level of the reservoir in the previous period.

8. The artificial intelligence-based ecological scheduling scheme optimization simulation coupling decision-making method according to claim 1, characterized in that: In step S6, the evaluation index of the calculation accuracy of the reservoir ecological dispatching scheme simulation decision model adopts the ratio of the root mean square error to the standard deviation RSR and the Nash efficiency coefficient NSE, where RSR and NSE are calculated by the following formula: In the formula, s i Represents the model simulation results, o i Represents the observation results, Represents the average observation value; if RSR<0.5, NSE>0.75, it means that the ecological scheduling scheme simulation decision-making model meets the accuracy requirements.

9. The artificial intelligence-based ecological scheduling scheme optimization simulation coupling decision-making method according to claim 1, characterized in that: In step S7, the calculation formula of the loss function with physical mechanism is: Loss=w o Loss o +w b Loss b +w h Loss h In the formula, Loss is a loss function with a physical mechanism, Loss o is the root mean square error between the target outbound flow and the model simulated outbound flow, Loss b is the penalty function for the reservoir outflow violating the water balance constraint, Loss h is the penalty function for the outflow of the reservoir exceeding the suitable interval of the key parameters, w o 、w b 、w h Loss o 、Loss b 、Loss h The weight coefficient of .

10. The artificial intelligence-based ecological scheduling scheme optimization simulation coupling decision-making method according to claim 1, characterized in that: In step S8, 80% of the scheduling trajectory sequence data set is used as training data to autonomously learn the optimal scheduling rules for reservoir ecology from the data set, and the remaining 20% ​​is used as test data to test the model performance and obtain the decision results of the reservoir ecological scheduling plan.

Images