NSGA-III algorithm-based cascade reservoir group multi-objective optimization scheduling method

Abstract

The invention discloses a cascade reservoir group multi-target optimization scheduling method based on an NSGA-III algorithm. The method comprises the following steps: establishing a multi-target scheduling model taking power generation benefit maximization, ecological water shortage minimization, water supply shortage minimization and shipping interruption day number minimization as target functions; solving the multi-target scheduling model by adopting an NSGA-III algorithm to generate a plurality of optimal solutions under constraint conditions so as to form an optimal solution set; and selecting an optimal solution from the optimal solution set as an optimal scheduling scheme. The method aims at solving the multi-target optimization problem of how to realize power generation benefit maximization, ecological water shortage minimization, water supply shortage minimization and shipping interruption day number minimization on the premise of ensuring flood control safety in cascade reservoir group multi-target scheduling.

Description

Technical Field

[0001] This invention relates to the field of cascade reservoir scheduling technology, specifically to a multi-objective optimization scheduling method for cascade reservoir groups based on the NSGA-Ⅲ algorithm. Background Technology

[0002] Cascade reservoir operation is a crucial technology in water resource management, and existing methods mostly employ single-objective or simple multi-objective optimization models. However, these methods often struggle to effectively handle multi-objective operation requirements under complex hydrological conditions, particularly in balancing flood control, power generation, irrigation, and ecological needs. Key problems include: 1. Single objective: Existing methods often focus on a specific objective, neglecting the coordination of other important objectives, such as the conflict between power generation and ecological needs.

[0003] 2. Low computational efficiency: Traditional optimization algorithms are difficult to handle large-scale reservoir systems, especially when facing multiple optimization objectives. Real-time scheduling computation efficiency is low and it is difficult to adapt to complex hydrological changes.

[0004] 3. Lack of flexibility: In the face of sudden hydrological conditions, the existing dispatching scheme is not flexible enough and cannot be dynamically adjusted, which reduces the effectiveness of the dispatching scheme.

[0005] In recent years, the application of intelligent optimization algorithms in reservoir scheduling has received widespread attention, among which genetic algorithms have shown significant advantages. The Non-Dominated Sorting Genetic Algorithm III (NSGA-III) is a classic multi-objective optimization algorithm that can obtain the Pareto optimal solution set without sacrificing individual diversity, making it particularly suitable for solving multi-objective conflict problems. Therefore, to meet the needs of cascade reservoir scheduling, there is an urgent need to develop a multi-objective optimization scheduling method based on NSGA-III that can simultaneously satisfy various benefit requirements. Summary of the Invention

[0006] This invention provides a multi-objective optimization scheduling method for cascade reservoir groups based on the NSGA-Ⅲ algorithm. It aims to solve the multi-objective optimization problem in the multi-objective scheduling of cascade reservoir groups, which is to maximize power generation benefits, minimize ecological water shortage, minimize water supply shortage, and minimize the number of days of navigation interruption while ensuring flood control safety.

[0007] The technical solution adopted in this invention is as follows: A multi-objective optimization scheduling method for cascade reservoir groups based on the NSGA-Ⅲ algorithm includes the following steps: Step 1: Establish a multi-objective scheduling model with the objective functions of maximizing power generation benefits, minimizing ecological water shortage, minimizing water supply shortage, and minimizing shipping interruption days; Step 2: Use the NSGA-Ⅲ algorithm to solve the multi-objective scheduling model, thereby generating several optimal solutions under the constraints to form an optimal solution set; Step 3: Select an optimal solution from the set of optimal solutions as the optimal scheduling scheme.

[0008] In step 1, the objective function is specifically: ; ; ; ; in, , , , These represent power generation benefits, ecological water shortage, water supply shortage, and number of days of shipping disruption, respectively. It refers to the number of hydroelectric power stations; It refers to the number of power generation periods; It is the overall output coefficient of hydropower station i; It is the average head of hydropower station i during stage t; It is the power generation flow of hydropower station i in stage t; It is the duration of each scheduling phase; It is the difference between the ecological flow and the downstream flow of hydropower station i during time period t; It is the water supply demand flow of hydropower station i at stage t; It is the actual water supply flow of hydropower station i in stage t; This is an indicator variable for navigation interruption at hydropower station i during time period t. If the outflow is lower than the lower limit of navigation or exceeds the upper limit of navigation, then... Otherwise, it is 0.

[0009] In step 1, the multi-objective scheduling model includes relevant constraints: ① The constraints on flood control storage capacity and water level limits are: ; in: This represents the water level of the i-th reservoir during time period t; This indicates the flood control limit water level of reservoir i; This represents the capacity of reservoir i during time period t; This indicates the flood control limit capacity of reservoir i, that is, the maximum capacity allowed during the flood season.

[0010] It also includes other constraints: ②Flow constraints: ; in: Let be the minimum outflow from the i-th reservoir during the t-th time period; Let be the outflow from reservoir i during time period t. Let be the maximum outflow from reservoir i during time period t.

[0011] ③ Load constraints: ; in: Let t be the minimum load requirement of the power grid on the power station for the i-th reservoir during the t-th time period; Let t be the power generation output of the i-th reservoir during the t-th time period; This represents the maximum output of the i-th reservoir during the t-th time period.

[0012] ④ Water balance constraints of each reservoir: ; in: Let be the water storage volume of reservoir i at the end of time period t; Let be the water storage volume of reservoir i at the beginning of time period t; Let be the inflow rate of the i-th reservoir during the t-th time period; Let be the outflow from reservoir i during time period t. The time period is long.

[0013] Step 2 includes the following steps: S2.1: Initial population generation: Set the number of iterations for the genetic algorithm Randomly generated containing The initial population of individuals: ; in, This represents the randomly generated initial population; Each represents an individual in the population; each individual This indicates the water level or outflow of each reservoir at different times: ; in: , For the first The reservoir during the period The lower and upper limits of the water level; Representing an interval Random numbers on the array.

[0014] This process ensures the diversity of initial solutions, which is beneficial for global search.

[0015] S2.2: Offspring generation: In the During generational iteration, the parent population By applying crossover and mutation operators, a new offspring population is generated: ; in, This indicates the generation of a new offspring population; Each represents an individual in the offspring population.

[0016] Crossover operations promote the combination of superior genes, while mutation operations enhance the population's exploratory capabilities, thereby avoiding getting trapped in local optima.

[0017] S2.3: Population Merging: Merging the parent and child generations yields a size of Expanding population: ; in, Indicates the expansion size is The population in the first In the next iteration, the parent population will be... With offspring population The merged overall set.

[0018] The expanded population includes both existing excellent solutions and newly generated potential solutions, providing a larger solution space for subsequent screening.

[0019] S2.4: Non-dominated sorting and reference point selection: right Perform non-dominated sorting and divide the area into multiple frontier layers:

[0020] In the non-dominated sorting process, the algorithm expands the population based on the superiority or inferiority of multi-objective fitness. Divide into layers.

[0021] First layer It consists of all solutions that are not dominated by any other individual, i.e., the best non-dominated front (Pareto Front) in the current iteration. Second floor It consists of solutions dominated only by individuals in the first layer, and is the second best. Third layer : It consists of solutions dominated by the individuals in the first two layers, and so on.

[0022] therefore, The non-dominated frontier layers represent different levels of superiority and inferiority; the number of layers is... It depends on the dominance structure of the current solution set.

[0023] Add solutions to new populations layer by layer ,until When the last layer When the remaining capacity is exceeded, the reference point method is used: S2.4.1. Normalize the objective function value; S2.4.2. Constructing a set of reference points , Each direction vector in the target space represents the reference position where the algorithm expects to allocate solutions in that direction.

[0024] S2.4.3. Establish the "adsorption relationship" between the individual and the reference point; S2.4.4. Select the individual closest to the reference point to join the new population.

[0025] This mechanism ensures that the solution set is evenly distributed in the target space, thereby increasing the diversity of solutions.

[0026] S2.5: Elite Retention: The next generation of the parent population is formed through selection operations:

[0027] in: It is the updated population after iteration; This represents the set of retained individuals obtained after non-dominated sorting and reference point filtering.

[0028] This strategy ensures that excellent solutions from the previous generation are retained, while combining new solutions to maintain the population's evolutionary capacity, thus enabling the algorithm to have convergence and stability.

[0029] S2.6: Pareto solution set output: When the number of iterations reaches its maximum value When outputting the final non-dominated solution set:

[0030] in, For the final non-dominated solution set; Iteration The set of solutions of degree 1.

[0031] This solution set is the Pareto optimal frontier that satisfies the constraints, and constitutes a candidate scheme for decision-making in the multi-objective scheduling problem.

[0032] Step 3, after obtaining the Pareto optimal solution set, requires selecting one solution as the final scheduling scheme based on actual operational needs. The Pareto optimal solution set reflects the balance between multiple objectives such as power generation efficiency, ecological water supply, water shortage, and uninterrupted navigation, providing decision-makers with diverse candidate solutions. Under the premise of meeting flood control safety constraints, decision-makers can make selections based on their specific periodic preferences or operational priorities. For example, when the grid load is high and power generation demand is strong, the solution with the highest power generation efficiency can be prioritized; while during the dry season or ecological protection period, the solution that maximizes ecological flow and water supply needs can be chosen.

[0033] This selection process not only supports the formulation of medium- and long-term pre-schedule plans but also allows for dynamic updates based on real-time hydrological data. When external hydrological conditions change abruptly, the scheduling model can quickly generate new Pareto solutions, allowing decision-makers to adjust the selected plan accordingly, thereby achieving real-time scheduling and optimization. This mechanism ensures that the cascade reservoir group can maintain the synergistic optimization of flood control safety and multi-objective benefits even in complex and uncertain hydrological environments, significantly improving the adaptability and practicality of the scheduling plan.

[0034] It also includes step 4: establishing and training a neural network-based prediction model to predict water level data and rainfall.

[0035] The prediction model is trained by taking historical water level data at the same time point over a certain period, the weather type during that period, and the upper and lower limits of the weather temperature. The water level data and rainfall at the same time point on the second day after that period are taken as the output of the prediction model. The trained prediction model is then used to predict water level data and rainfall.

[0036] A multi-objective optimization scheduling system for a cascade reservoir group based on the NSGA-III algorithm is characterized by comprising: Model building module: Used to build a multi-objective scheduling model with the goals of maximizing power generation benefits, minimizing ecological water shortage, minimizing water supply shortage, and minimizing shipping interruption days under the constraint of flood control safety; The solution module is used to solve the multi-objective scheduling model using the NSGA-Ⅲ algorithm, thereby generating several optimal solutions under constraints to form an optimal solution set; The selection module is used to select an optimal solution from the set of optimal solutions as the optimal scheduling scheme according to actual needs.

[0037] Prediction module: Used to build and train a neural network-based prediction model, learn from historical water level data, weather type, upper and lower limits of weather temperature, and output water level data and rainfall prediction results for future periods, providing input data support for the scheduling model.

[0038] A storage medium storing a computer program, which, when executed by a processor, implements the aforementioned multi-objective optimization scheduling method for a cascade reservoir group based on NSGA-Ⅲ.

[0039] An electronic device includes at least one processor and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to execute the aforementioned multi-objective optimization scheduling method for a cascade reservoir group based on NSGA-III.

[0040] This invention presents a multi-objective optimization scheduling method for a cascade reservoir group based on the NSGA-III algorithm, with the following technical advantages: 1) Optimization of multi-objective benefits: Under the premise of ensuring flood control safety, this invention can maximize power generation benefits, reduce ecological water shortage and water supply shortage, and ensure smooth navigation.

[0041] 2) Improve scheduling flexibility: The scheduling scheme can be dynamically adjusted according to real-time hydrological conditions to ensure the flexibility and responsiveness of the scheduling process.

[0042] 3) Improved computational efficiency: The NSGA-Ⅲ algorithm can effectively handle the multi-objective scheduling problem of large-scale reservoir groups, and its computational efficiency is higher than that of traditional algorithms.

[0043] 4) Collaborative optimization: Through comprehensive evaluation of different scheduling schemes, decision-makers can select the optimal scheduling scheme based on the balance of multiple objectives, thereby achieving collaborative scheduling of the cascade reservoir group. Attached Figure Description

[0044] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a flowchart of the scheduling method of the present invention.

[0045] Figure 2 This is a schematic diagram of the prediction model. Detailed Implementation

[0046] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0047] This invention aims to solve the multi-objective optimization problem in the multi-objective scheduling of cascade reservoir groups, namely, how to maximize power generation benefits, minimize ecological water shortage, minimize water supply shortage, and minimize navigation interruption days while ensuring flood control safety. Flood control safety is a constraint, not an optimization objective, but its importance ensures the safety of the entire system. Specifically, this embodiment 1 provides a multi-objective optimization scheduling method for cascade reservoir groups based on NSGA-III, including: S1. Under the constraint of flood control safety, establish a multi-objective scheduling model with the objective functions of maximizing power generation benefits, minimizing ecological water shortage, minimizing water supply shortage, and minimizing the number of days of shipping interruption. The constraints specifically refer to the upper limits of flood control storage capacity and water level. By setting upper limits for flood control storage capacity and water level as constraints, flood control safety is ensured to guarantee that the reservoir can effectively regulate floodwaters and avoid flood threats during the flood season. The flood control safety objective is not used as an optimization goal, but rather as a constraint that is strictly adhered to during operation.

[0048] The objective function is specifically: ; ; ; ; in, , , , These represent power generation benefits, ecological water shortage, water supply shortage, and number of days of shipping disruption, respectively. It refers to the number of hydroelectric power stations. It refers to the number of power generation periods. It is the overall output coefficient of hydropower station i. It is the average head of hydropower station i during stage t. It is the power generation flow of hydropower station i in stage t. It is the duration of each scheduling phase; It is the difference between the ecological flow and the downstream flow of hydropower station i during time period t; It is the water supply demand flow rate of hydropower station i in stage t. It is the actual water supply flow of hydropower station i in stage t; This is an indicator variable for navigation interruption at hydropower station i during time period t. If the outflow is lower than the lower limit of navigation or exceeds the upper limit of navigation, then... Otherwise, it is 0.

[0049] In the objective function, the first objective function To maximize power generation benefits: Through reservoir scheduling, the goal is to maximize power generation benefits, taking into account the comprehensive utilization of flow and water level.

[0050] Second objective function To minimize ecological water shortages, reservoir flow is managed to ensure water use for ecological purposes while minimizing water scarcity.

[0051] The third objective function To minimize water shortages, the goal is to reduce water shortage problems as much as possible while meeting the irrigation needs of residents and agriculture.

[0052] The fourth objective function To minimize shipping disruption days: regulate reservoir discharge to reduce the number of shipping disruption days caused by flow changes.

[0053] S2. The NSGA-III algorithm is used to solve the multi-objective scheduling model, thereby generating several optimal solutions under the constraints to form an optimal solution set; The NSGA-III algorithm is an advanced multi-objective optimization algorithm that can optimize multiple objectives simultaneously and generate a set of Pareto optimal solutions. By adding flood control safety constraints, it ensures flood control safety requirements during the scheduling process, while optimizing the other four objectives (power generation, ecology, water supply, and navigation).

[0054] The specific process is as follows: A1. When using the NSGA-III algorithm for optimization, the population needs to be initialized first. The maximum number of iterations for the genetic algorithm should be set. And randomly generated within the feasible solution space. Individuals, forming the initial population Each individual Composed of the water level or outflow of each reservoir at different times, satisfying ,in , The first The reservoir during the period The lower and upper limits of the water level. This step ensures diversity of knowledge, providing an initial foundation for the global search.

[0055] A2. During the iteration process, the parent population New offspring populations are generated through crossover and mutation operations. Crossover operations promote the combination of superior information through the exchange of gene fragments, while mutation operations introduce new search directions by randomly perturbing some gene values, thereby maintaining population diversity and avoiding premature convergence.

[0056] A3. Merge the parent generation and the child generation to obtain a size of Expanding population In the expanded population, individuals contain both excellent solutions from the previous generation and newly generated potential solutions, providing a larger solution space for subsequent selection.

[0057] For each individual in the population, evaluation is performed according to multiple objective functions: (1) Maximize power generation benefits: Through reservoir scheduling, the goal is to maximize power generation benefits, taking into account the comprehensive utilization of flow and water level.

[0058] ; (2) Minimize ecological water shortage: By regulating reservoir flow, ensure water use for ecological environment and minimize ecological water shortage.

[0059] ; (3) Minimize water shortage: Minimize water shortage problems as much as possible while meeting the irrigation needs of residents and agriculture.

[0060] ; (4) Minimize shipping interruption days: Schedule reservoir discharge to reduce shipping interruption days caused by flow changes.

[0061] ; Each objective function is designed based on actual operational requirements to ensure a comprehensive consideration of various benefits.

[0062] A4. Regarding the expansion of the population Perform non-dominated sorting and divide into multiple front layers. The new parental group absorbs individuals from these frontal layers layer by layer until it reaches population size. If the last layer exceeds the capacity limit, a reference point mechanism is used for filtering, which involves normalizing the objective function values ​​and constructing a set of reference points. Furthermore, based on the distance relationship between individuals and the reference point, individuals closest to the reference point are preferentially selected to join the new group. This mechanism ensures a uniform distribution of information in the target space and enhances the diversity of the Pareto front.

[0063] A5. After forming a new generation of parental population, an elite preservation strategy is adopted, denoted as... This strategy ensures that excellent solutions from the previous generation are preserved, while combining newly generated solutions to maintain the population's evolutionary capacity, thus enabling the algorithm to achieve convergence and stability.

[0064] A6. When the number of iterations reaches its maximum value. The algorithm terminates when the convergence condition is met, and outputs the final non-dominated solution set.

[0065] This solution set is the Pareto optimal frontier that satisfies the constraints, forming candidate solutions for the multi-objective scheduling problem and providing decision-makers with diverse choices.

[0066] S3. Based on actual needs, select an optimal solution from the set of optimal solutions as the optimal scheduling scheme.

[0067] The multi-objective scheduling model generates scheduling schemes and allows for corresponding selections: the optimized Pareto optimal solution set allows decision-makers to select the optimal scheduling scheme that can achieve power generation benefits, ecological water supply, water shortage prevention, and uninterrupted navigation, while meeting flood control safety constraints, based on actual needs.

[0068] After obtaining the Pareto solution set (optimal solution), for example, if the current demand for power generation is very high or the decision-maker prefers a certain amount of power generation, then the decision-maker can select the solution with the largest power generation for corresponding scheduling.

[0069] This multi-objective scheduling model enables real-time scheduling and optimization. It can be used not only for pre-scheduling but also for continuously adjusting and optimizing the scheduling scheme based on feedback from real-time hydrological data, ensuring that it can maintain efficient scheduling when responding to sudden hydrological changes.

[0070] S4. Establish and train a neural network-based prediction model. The prediction model is used to predict water level data and rainfall. The historical water level data at the same time point in a certain period, the weather type during that period, and the upper and lower limits of the weather temperature are used as inputs to the prediction model. The water level data and rainfall at the same time point on the second day after that period are used as outputs to train the prediction model. The trained prediction model is then used to predict water level data and rainfall.

[0071] That is, time series forecasting and dynamic optimization: introduce time series forecasting analysis based on historical water level data and future weather forecasts to predict future reservoir water levels and rainfall, and dynamically adjust individual fitness and scheduling strategies based on the forecast results.

[0072] like Figure 2 As shown, for example, historical water level data at 00:00 for each day of a week, the weather type of that week, and the upper and lower limits of the weather temperature are selected as inputs to the prediction model. Then, the water level data and rainfall at the same time on the second day after that week are used as outputs of the prediction model. The historical water level data, the weather type of that week, and the upper and lower limits of the weather temperature are input into the prediction model, and then the prediction model outputs the predicted water level data and rainfall. Then, the predicted water level data and rainfall are compared with the water level data and rainfall at the same time on the second day after that week to calculate the loss function, thereby training the prediction model.

[0073] In summary, this invention uses flood control safety as a constraint and combines it with the NSGA-Ⅲ algorithm to achieve multi-objective optimal scheduling of a cascade reservoir group. Specifically, 1. Construction of a multi-objective scheduling model: It simultaneously considers objectives such as power generation, ecology, water supply and shipping, and achieves dynamic balance of scheduling schemes.

[0074] 2. Application of NSGA-III algorithm: The NSGA-III algorithm is introduced to solve multi-objective optimization problems, generate Pareto optimal solution sets, and improve the diversity of scheduling schemes and the flexibility of decision selection.

[0075] 3. Flood control safety constraints: By introducing flood control safety as a constraint, the safety of the system under complex hydrological conditions is ensured, while optimizing other objectives.

[0076] The method in this invention can be applied to the scheduling of cascade reservoir groups, and can also be extended to other fields with similar multi-objective scheduling needs, such as water resource management and energy scheduling.

[0077] Example 2: A multi-objective optimization scheduling system for a cascade reservoir group based on NSGA-III is also disclosed to execute the multi-objective optimization scheduling method for a cascade reservoir group based on NSGA-III described in Example 1, including: The model building module is used to establish a multi-objective scheduling model under the constraint of flood control safety, with the objectives of maximizing power generation benefits, minimizing ecological water shortage, minimizing water supply shortage, and minimizing the number of days of shipping interruption. The solution module is used to solve the multi-objective scheduling model using the NSGA-Ⅲ algorithm, thereby generating several optimal solutions under constraints to form an optimal solution set; The selection module is used to select an optimal solution from the set of optimal solutions as the optimal scheduling scheme according to actual needs.

[0078] Prediction module: Used to build and train a neural network-based prediction model, learn from historical water level data, weather type, upper and lower limits of weather temperature, and output water level data and rainfall prediction results for future periods, providing input data support for the scheduling model.

[0079] Example 3: An electronic device is also disclosed, including at least one processor and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to execute a multi-objective optimization scheduling method for a cascade reservoir group based on NSGA-III as described in Example 1.

[0080] Example 4: A storage medium storing a computer program is also disclosed, characterized in that the computer program, when executed by a processor, implements the multi-objective optimization scheduling method for a cascade reservoir group based on NSGA-Ⅲ described in Example 1.

Claims

1. A multi-objective optimal scheduling method for cascade reservoirs based on NSGA-III algorithm, characterized in that The method comprises the following steps: Step 1: establishing a multi-objective scheduling model with the objective functions of maximizing power generation benefit, minimizing ecological water shortage, minimizing water supply shortage and minimizing shipping interruption days; Step 2: solving the multi-objective scheduling model by using an NSGA-III algorithm, thereby generating a plurality of optimal solutions under the constraint conditions to form an optimal solution set; Step 3: selecting an optimal solution from the optimal solution set as an optimal scheduling scheme.

2. The multi-objective optimal scheduling method for cascade reservoirs based on NSGA-III algorithm according to claim 1, characterized in that: In step 1, the objective functions are specifically as follows: ； ； ； ； wherein, , , , respectively represent the power generation benefit, the ecological water shortage amount, the water supply shortage amount, and the navigation interruption days; is the number of hydropower stations; is the number of power generation periods; is the comprehensive output coefficient of the hydropower station i; is the average water head of the hydropower station i in the stage t; is the power generation flow of the hydropower station i in the stage t; is the time length of each dispatching stage; is the difference between the ecological flow and the discharge flow of the hydropower station i in the t period; is the water supply demand flow of the hydropower station i in the stage t; is the actual water supply flow of the hydropower station i in the stage t; is an indication variable of the navigation interruption of the hydropower station i in the time period t, if the discharge flow is lower than the navigation lower limit or exceeds the navigation upper limit, then , otherwise 0.

3. The multi-objective optimal scheduling method for cascade reservoirs based on NSGA-III algorithm according to claim 2, characterized in that: In step 1, the multi-objective scheduling model comprises relevant constraint conditions: ① the constraint condition of flood control storage capacity and upper limit of water level is as follows: ； wherein: represents the water level of the ith reservoir at time period t; represents the flood control limit water level of reservoir i; represents the storage of reservoir i at time period t; represents the flood control limit storage of reservoir i, i.e. the maximum storage allowed during the flood season; Meanwhile, other constraint conditions are also included: ② flow constraint: ； wherein: is the minimum outflow of the ith reservoir in the tth period; is the outflow of the ith reservoir in the tth period; is the maximum outflow of the ith reservoir in the tth period; ③ load constraint: ； wherein: is the minimum load requirement of the power grid to the power plant of the ith reservoir at the tth period; is the power generation output of the ith reservoir at the tth period; is the maximum output of the ith reservoir at the tth period; ④ water balance constraint of each reservoir: ； wherein: is the storage capacity of the ith reservoir at the end of the tth time period; is the storage capacity of the ith reservoir at the beginning of the tth time period; is the inflow to the ith reservoir during the tth time period; is the outflow from the ith reservoir during the tth time period; is the length of the time period.

4. The multi-objective optimal scheduling method for cascade reservoirs based on NSGA-III algorithm according to claim 3, characterized in that: The step 2 comprises the following steps: S2.1: initial population generation: Setting the number of iterations of the genetic algorithm , randomly generating an initial population of individuals: ； wherein, represents an initial population randomly generated; respectively represents each individual in the population; each individual represents the water level or outflow of each reservoir at each time period: ； wherein: , is the lower and upper water level of the th reservoir at time interval ; denotes a random number on the interval ; S2.2: offspring generation: In the first In the parent population Applying the crossover and mutation operators, a new offspring population is generated: ； wherein, represents the creation of a new child population; respectively represent each individual in the child population; S2.3: population merging: The parents and offspring are combined to obtain an extended population of size N = Nparent + Noffspring ； in, Indicates the expansion size is The population in the In the next iteration, the parent population will be... With offspring population The merged overall set; S2.4: non-dominated sorting and reference point selection: For non-dominated sorting is performed, dividing into multiple front layers: ； non-dominated front layers representing different levels of quality, number of layers depends on the dominance relation structure of the current solution set; Add solutions to new population layer by layer , until ; when the last layer exceeds the remaining capacity, adopt the reference point method; S2.5: elite reservation: The next generation of parent population is formed through the selection operation: wherein: is the updated population after iteration; denotes the set of reserved individuals after non-dominated sorting and reference point filtering; S2.6: Pareto solution set output: When the number of iterations reaches the maximum value the final non-dominated solution set is output: wherein, is the final set of non-dominated solutions; denotes the iteration set of solution sets. The solution set is the Pareto optimal front that meets the constraint conditions, and constitutes the candidate schemes for decision selection of the multi-objective scheduling problem.

5. The multi-objective optimal scheduling method for cascade reservoirs based on NSGA-III algorithm according to claim 4, characterized in that: In S2.4, the reference point method comprises the following steps: S2.4.1: normalizing the objective function values; S2.4.

2. Constructing the set of reference points , respectively represent each direction vector in the target space, representing the reference position at which the algorithm expects to allocate the solution in that direction; S2.4.3: establishing an "adsorption relationship" between the individual and the reference point; S2.4.4: selecting the individual closest to the reference point to join the new population.

6. The multi-objective optimal scheduling method for cascade reservoirs based on NSGA-III algorithm according to claim 5, characterized in that: After obtaining the Pareto optimal solution set, an optimal solution needs to be selected as the final scheduling scheme according to the actual operation requirements in step 3; under the premise of meeting the flood control safety constraint, the decision maker can select according to the target preference or operation emphasis in a specific period; for example, when the power grid load is high and the power generation demand is high, the solution with the maximum power generation benefit is selected; and in the dry season or ecological protection period, the solution that can meet the ecological flow and water supply demand to the maximum extent is selected.

7. The multi-objective optimal scheduling method for cascade reservoirs based on NSGA-III algorithm according to claim 6, characterized in that: Real-time hydrological data can also be combined for dynamic updating; when external hydrological conditions change suddenly, the scheduling model can quickly generate a new Pareto solution set, and the decision maker adjusts the selected scheme accordingly, thereby realizing real-time scheduling and optimization.

8. The multi-objective optimal scheduling method for cascade reservoirs based on NSGA-III algorithm according to claim 7, characterized in that: Step 4: establishing and training a prediction model based on a neural network to predict water level data and rainfall; Wherein, the historical water level data at the same time point in a period, the weather type in the period, the upper and lower limits of the weather temperature are respectively taken as the inputs of the prediction model, and the water level data and rainfall at the same time point on the second day after the period are taken as the outputs of the prediction model, so as to train the prediction model, thereby predicting the water level data and rainfall by using the trained prediction model.

9. A multi-objective optimal scheduling system for cascade reservoirs based on NSGA-III algorithm, characterized in that, It comprises: A model establishing module for establishing a multi-objective scheduling model with the objective functions of maximizing power generation benefit, minimizing ecological water shortage, minimizing water supply shortage and minimizing shipping interruption days under the constraint condition of flood control safety; A solving module for solving the multi-objective scheduling model by using an NSGA-III algorithm, thereby generating a plurality of optimal solutions under the constraint conditions to form an optimal solution set; A model establishing module for establishing a multi-objective scheduling model with the objective functions of maximizing power generation benefit, minimizing ecological water shortage, minimizing water supply shortage and minimizing shipping interruption days under the constraint condition of flood control safety; The selecting module is configured to select one optimal solution from the optimal solution set as an optimal scheduling scheme according to actual requirements.

10. The multi-objective optimal scheduling system for cascade reservoirs based on NSGA-III algorithm according to claim 9, characterized in that: Also comprising: The prediction module is configured to establish and train a neural network-based prediction model, learn historical water level data, weather types, upper and lower limits of weather temperature, and output water level data and rainfall prediction results of a future period, thereby providing input data support for the scheduling model.

11. A storage medium characterized by: The computer program is stored in the memory and is executed by the processor to implement any one of the NSGA-III-based multi-objective optimal scheduling methods of the cascade reservoir groups according to claims 1-8.

12. An electronic device, comprising: The computer program is stored in the memory and is executed by the processor to implement any one of the NSGA-III-based multi-objective optimal scheduling methods of the cascade reservoir groups according to claims 1-8.

Images