A dynamic multi-objective optimization control system and method for giant reservoir water level during flood season

Abstract

The present invention discloses a dynamic multi-objective optimization control system and method for the water level of a giant reservoir during flood season, which belongs to the technical field of reservoir optimization and dispatching. The system can quickly respond to the environment according to information such as precipitation forecast information, reservoir operation status, and real-time decision-making needs, flexibly adjust the operation plan of the reservoir, maximize the benefits of the reservoir, better coordinate the relationship between multiple target needs such as flood control and power generation of the reservoir, and improve the comprehensive benefits of the reservoir. The system has the outstanding advantages of accurate environmental change tracking, rapid response, better solution convergence and distribution, efficient dispatching plan decision-making, and a wide range of applicable scenarios, and can provide technical support for the dynamic control of the water level of a reservoir group during flood season.

Description

Technical Field

[0001] The invention belongs to the technical field of reservoir optimization and dispatching, and in particular relates to a dynamic multi-objective optimization control system and method for a giant reservoir water level in flood season. Background Art

[0002] As an important engineering measure for optimizing water resources allocation and preventing and controlling water disasters, reservoirs can play a key role in regulating and storing water resources in flood control, drought resistance, power generation and other fields. In the early stage of reservoir construction and operation, the design focused on flood prevention and control safety. Therefore, the flood limit water level of the reservoir was formulated according to the needs of flood control planning, and a certain amount of flood storage capacity was reserved before the flood season. However, in the water inflow scenario where the design flood did not occur, the reservoir water level was limited to below the flood limit water level, and a large amount of runoff entering the reservoir during the flood season was not fully utilized. With the continuous advancement of hydrological forecasting and reservoir scheduling technology, reservoir management has gradually adopted a flood limit water level scheduling method. This method takes into account the temporal distribution characteristics of floods to a certain extent and improves the utilization rate of flood resources during the flood season. However, this method does not take into account flood forecast information, resulting in a large amount of water abandonment in the reservoir during the flood season, and even the situation where the reservoir cannot store water after the flood, which seriously affects the benefits of the reservoir. In order to solve these problems, the dynamic control method of operating water level during flood season has gradually become a research hotspot. On the basis of ensuring the safe operation of reservoir flood control, the full exploration of reservoir flood control benefits and economic benefits is carried out through dynamic control of water level, which has become one of the effective means to solve the optimal allocation of water resources and the prevention and control of water disasters.

[0003] Dynamic control of reservoir flood season operating water level refers to the dispatching process of real-time dynamic control of reservoir water level according to scientifically demonstrated flood season operating water level schemes, based on water and rainfall forecasts, without lowering flood control standards. At present, many methods for dynamic control of reservoir multi-objective flood season operating water level based on modern meta-heuristic optimization technology have been proposed. These meta-heuristic optimization methods can provide decision makers with the best dispatching scheme set, called Pareto optimal solution. Although these meta-heuristic optimization methods for multi-objective reservoir dispatch have achieved some success, they ignore the uncertainty of changes in environmental factors such as forecast accuracy and reservoir status. Some methods combine flood forecasting models with deterministic optimization models by changing runoff scenarios, but static optimization models still cannot cope with real-time environmental changes, and only input deterministic data based on assumptions for model optimization without conducting decision analysis.

[0004] In order to solve these problems, the present invention proposes a dynamic multi-objective optimization control method for the flood season water level of giant reservoirs to solve the scheduling optimization model. This method can respond to the environment quickly according to information such as precipitation forecast information, reservoir operation status, and real-time decision-making needs, so as to flexibly adjust the operation plan of the reservoir, maximize the benefits of the reservoir, better coordinate the relationship between multiple target needs such as flood control and power generation of the reservoir, improve the comprehensive benefits of the reservoir, and provide strong support for the rational use of water resources and flood control. It has good application prospects in engineering practice. Summary of the invention

[0005] The purpose of the present invention is to provide a dynamic multi-objective optimization control system and method for the flood season water level of a giant reservoir, so as to solve the problem that the existing dynamic control model of the flood season water level of the reservoir is insufficient in terms of efficient solution of the scheduling scheme and dynamic response to environmental changes.

[0006] To achieve the above-mentioned purpose, the present invention provides a giant reservoir flood season water level dynamic multi-objective optimization control system, including a model initialization module, a runoff forecast scheme risk quantification assessment module, a giant reservoir flood season water level dynamic multi-objective optimization control method solution module, a GUI interface visualization module, and a calling interface module;

[0007] The giant reservoir flood season water level dynamic multi-objective optimization control method solution module includes a static scheduling optimization algorithm framework adaptation and elite set selection unit, a non-inferior scheduling scheme set decision analysis unit, a scheduling environment factor detection unit, a dynamic scheduling operation response algorithm unit, and a reservoir optimal scheduling scheme acquisition unit.

[0008] The present invention also provides a method for dynamic multi-objective optimization control of water level in a giant reservoir during flood season, comprising the following steps:

[0009] Step 1: Construct a multi-objective optimization scheduling model for giant reservoirs and generate an initial optimal population. Based on the basic information of reservoir runoff, operating status, and operating characteristics, construct a multi-objective optimization scheduling model for giant reservoirs with maximum reservoir power generation, maximum peak shaving of the control section, and maximum available water as objective functions during the scheduling period. Add relevant constraints to the multi-objective optimization scheduling model for giant reservoirs, and preset the number of objective functions M, the number of decision variables D, and the length of the forecast period T of the scheduling model. y , static optimization times MaxIter, optimization population size N, elite scheduling solution set size N c , and set the outflow during the reservoir dispatching period as the decision variable; input the relevant data of the giant reservoir, and randomly generate the optimal individual position according to the preset range of the decision variable to complete the initial optimal population generation;

[0010] Step 2: Execute the risk quantification assessment module of the runoff forecast scheme. Based on the current operation status of the giant reservoir and the Monte Carlo method, a large number of random runoff forecast scheme sets are generated within the forecast interval for risk quantification assessment. Based on the risk quantification assessment results, the risk constraints of the multi-objective optimization scheduling model of the giant reservoir are refined and modified. In the process of optimizing the multi-objective scheduling scheme, the inferior scheduling schemes with higher risk values ​​are gradually eliminated.

[0011] Step 3: Execute the static scheduling optimization algorithm framework adaptation and elite set selection unit, execute the static optimization algorithm selection strategy and elite set optimization strategy to complete the optimal optimization strategy adaptation and excellent optimization individual storage operation of the giant reservoir multi-objective scheduling optimization problem in turn;

[0012] Step 4: Execute the non-inferior scheduling solution set decision analysis unit, use the hierarchical analysis method to make decisions on the Pareto frontier non-inferior scheduling solution set obtained by the static scheduling optimization algorithm, and then analyze and match to obtain the most suitable scheduling decision solution under different scheduling requirements and environmental scenario changes;

[0013] Step 5, execute the dispatching environmental factor detection unit, and use an iterative method to continuously detect whether the precipitation runoff forecast information, reservoir operation status, and real-time decision-making demand environmental factors of the basin and reservoir area where the giant reservoir is located have changed; if the environment has changed, execute step 6; if the environment has not changed, execute step 4 to obtain the optimal dispatching plan for dispatching;

[0014] Step 6: Execute the dynamic scheduling operation response algorithm unit, adopt the memory individual adaptation optimization strategy, the diversity dynamic introduction strategy and the heterogeneous information supply collaborative prediction strategy to quickly respond to environmental changes, and quickly track and respond to the dynamic multi-objective scheduling model;

[0015] Step 7, execute the reservoir optimal scheduling plan acquisition unit to determine whether the preset number of environmental changes has been reached. If the preset number of environmental changes has been reached, execute the static optimization algorithm and decision model to output the reservoir operation status data and the current and historical optimal scheduling plan set. Otherwise, return the scheduling plan output in step 6 to step 3 for continued execution.

[0016] Preferably, the objective function for maximizing the reservoir power generation during the scheduling period in step 1 is:

[0017]

[0018] Where E is the total power generation of all reservoirs in the time period; K and t are the number of reservoirs and the current time period respectively; T y is the length of the forecast horizon; P k,t , Q k,t , H k,t are the output, power generation flow and water head of the kth reservoir in the tth period respectively; Ak is the output coefficient of the kth reservoir; Δt is the length of the calculation period;

[0019] The maximum peak shaving objective function of the control section during the dispatch period is:

[0020]

[0021] Among them, C is the peak shaving scheduling function; the decision variable is the outflow of the kth reservoir at time t; is the interval inflow forecast value between the kth reservoir and the flood control section at time t; K is the number of reservoirs; M k (·) is the flow process calculation method for routing the outflow process of the kth reservoir to the downstream reservoir or downstream flood control section. When there is a time lag, the Muskingum method is used for routing, otherwise it is directly superimposed;

[0022] The maximum objective function of water supply during the dispatch period is:

[0023]

[0024] Among them, G is the maximum water supply dispatching function during the dispatching period; the decision variable is the water supply flow of the kth reservoir at time t; and are the upper and lower limits of the water supply capacity of the kth reservoir at time t; K is the number of reservoirs.

[0025] Preferably, the relevant constraints added to the giant reservoir multi-objective optimization scheduling model in step 1 include water balance constraints, output constraints, reservoir capacity constraints, outflow constraints, power generation flow constraints and scheduling scheme risk constraints, and the specific expressions are as follows:

[0026] Water balance constraints:

[0027]

[0028] Among them, I k,t , Q k,t are the interval inflow and outflow of the kth reservoir in the tth period; V k,t is the storage capacity of the kth reservoir in the tth period, V k,t-1 is the storage capacity of the kth reservoir in the t-1th period, Δt is the length of the calculation period, Q l,t is the outflow of the lth reservoir in the tth period, T y is the length of the forecast horizon, Ω k are all upstream reservoirs directly connected to reservoir k, Remove lost water from outflows from upstream reservoirs;

[0029] Output constraints:

[0030]

[0031] in, are the maximum and minimum outputs of the kth reservoir in the tth period, respectively;

[0032] Storage capacity constraints:

[0033]

[0034] in, are the maximum and minimum storage capacity of the kth reservoir in the tth period respectively;

[0035] Outbound traffic constraints:

[0036]

[0037] in, are the maximum and minimum outflows of the kth reservoir in the tth period, respectively;

[0038] Generation flow constraints;

[0039]

[0040] in, are the maximum and minimum power generation flows of the kth reservoir in the tth period, respectively;

[0041] Scheduling plan risk constraints:

[0042]

[0043] Among them, R k , are the risk value and minimum risk threshold of the kth scheduling plan respectively.

[0044] Preferably, the specific process of executing the static scheduling optimization algorithm framework adaptation and elite set selection unit in step 3 is as follows:

[0045] S31, execute the static dispatch optimization algorithm framework adaptation strategy, and adapt the corresponding algorithm framework according to the number M of objective functions of the giant reservoir multi-objective optimization dispatch model: when the number M of objective functions is 2, the non-dominated sorting genetic algorithm II framework is used for static optimization; when the number M of objective functions is not less than 3, the non-dominated sorting genetic algorithm III framework is used for static optimization;

[0046] S32, execute the elite set selection strategy for static scheduling optimization, use the elite set to store the set of non-inferior scheduling solutions in the optimization process, the size of the elite scheduling solution set is N c Keeping consistent with the optimal population size N, when executing the elite set storage, the current optimal population Xt and the elite set Arc before the environmental change t Merge to perform efficient non-dominated sorting, select the Pareto first frontier individuals of the merged set to update the elite set Arc t , if the number of individuals on the first Pareto frontier does not exceed the size N of the elite scheduling solution set c All are stored, otherwise the elite set individuals that meet the archive number requirements are selected according to the framework of the selected static scheduling optimization algorithm. The non-dominated sorting genetic algorithm II is screened based on the crowding distance, and the non-dominated sorting genetic algorithm III is screened based on the reference point. If the scheduling environment factors change, the elite set is not reset, and the elite set Arc before the environment change is selected. t Randomly select 0.5×N c The elite set Arc after individuals enter a changing environment t+1 .

[0047] Preferably, in step 5, the optimal population X is randomly selected in each iteration. t The objective functions and constraints of the 0.1×N scheduling schemes are tested before and after iteration, as well as the matching degree of the actual operating state of the reservoir and the operating state detection of the changes in real-time decision-making conditions. If it is found that the objective function or constraint of the scheduling scheme being tested is destroyed or the operating state changes, it is determined that the current environment has changed.

[0048] Preferably, the specific process of executing the dynamic scheduling operation response algorithm unit is as follows:

[0049] S61, execute the memory individual adaptation optimization strategy, by saving the excellent solutions and valuable information found in the past, after the environment changes, a relatively excellent solution can be quickly found and the impact of environmental uncertainty can be reduced. This strategy stores the elite set Arc before the historical environmental changes to the memory archive set MS, and when the environment changes, it matches the memory archive set and selects the historical optimal solution with the highest matching degree in the current environment to respond;

[0050] S62. Execute the dynamic introduction strategy of diversity. Through the dynamic introduction strategy of diversity, a diverse search is conducted in the search space to ensure that the diversity of the population is maintained when the environment changes greatly, is difficult to predict or is irregular. This strategy considers the density of the population through a loose population crowding distance calculation method, and generates new individuals through mutation and initialization to replace the densely distributed individuals in the population, thereby maintaining the diversity of the population;

[0051] S63. Execute a collaborative prediction strategy for heterogeneous information supply. By combining historical information on environmental changes, try to foresee or estimate the changing trend of future optimization problems in order to better adapt to environmental changes and quickly find feasible solutions after environmental changes. This strategy calculates the position of the elite set before and after environmental changes, predicts the direction and degree of environmental changes, and generates new individuals through linear prediction and micro-disturbance to quickly respond to environmental changes.

[0052] Preferably, the specific process of executing the memory individual adaptation optimization strategy in S61 is as follows:

[0053] S611, first of all, the elite set Arc before each environment change t Store it in the memory archive set MS, and set the memory storage period length to 25% of the optimal population size. If the memory archive is full, the archive is updated according to the first-in-first-out principle; if the storage period length is 25, when the archive MS 1-25 If the storage is done, the next time the environment changes, the MS 26 Replace MS1, that is, the memory save is MS 2-26 ;

[0054] S612: If the environment changes, each memory archive MS is updated from the memory archive MS. t According to the ascending order of the first objective function, M+3 individuals are selected at equal intervals based on the change of the number of objective functions M to ensure the globality of the selected individuals;

[0055] S613, for each archived MS t The selected individuals use the Euclidean distance to calculate the distance of the objective function value before and after the environmental change and divide it by the maximum value of the objective function value before and after the corresponding change, eliminate the dimension, and add the obtained distances to obtain the distance and DMS of each archived individual. t , if there is 0 in the denominator, the data will be removed;

[0056] S614, finally found DMS t The smallest value among them, takes the corresponding memory archive number m as the adaptation archive, and selects the adaptation archive MS m A scheduling scheme with a size of 0.2α times the optimal population is randomly selected as the memory scheduling scheme

[0057]

[0058] in, They represent the objective function values ​​of the individuals selected in the memory archive before and after the environmental changes; E_distance represents the calculation using Euclidean distance; randperm(I,I1) is a random selection function, randomly selecting I1 individuals from I, α represents the adjustment parameter, and N represents the optimal population size.

[0059] Preferably, the specific process of executing the diversity dynamic introduction strategy in S62 is as follows:

[0060] S621. First determine the number of individuals introduced by diversity The optimal population size N is α times. Considering that the response direction after environmental changes is uncertain, the individual numbers of the introduced diversity are selected through the loose crowding distance optimization method to select individuals with close distances for replacement screening. By calculating the current optimal individual X t The crowding distance is then sorted in ascending order according to the crowding distance, and a random selection is made from the optimal population that is (1-α)×N times larger than the distance sorting. Introduce diversity into individuals;

[0061] S622, secondly, a multi-dimensional dynamic optimization method considering the optimization population size and the number of static optimizations is used to introduce diverse individuals, which include mutant individuals and initialized individuals. The quantity is set to Initialize individual The quantity is set to The mutation adopts polynomial mutation, and the mutation probability and distribution index are set to 1 and Initialization uses random generation within the upper and lower limits of the decision variable. The specific expression is as follows:

[0062]

[0063] Among them, Crowding(·) is the distance value of the corresponding individual obtained by the crowding distance method; Sort(·) is the ascending sorting function; randperm(I,I1) is the random selection function, which randomly selects I1 individuals from I; Respectively represent the use of polynomial mutation and initialization for the selected mutant individuals and the initialized individuals; P_mutation(a,b) represents the use of polynomial mutation method, a and b are the mutation probability and distribution index respectively; Express The individual's decision variables are randomly generated functions within the upper and lower limits.

[0064] Preferably, the specific process of executing the heterogeneous information supply collaborative prediction strategy in S63 is as follows:

[0065] S631, first of all, the elite set Arc before the environment changes tSort the non-dominated solutions, store the Pareto first frontier individuals and update the elite set, and then calculate the elite set mean MArc t and the mean MepArc of the endpoint individuals with the best fitness values ​​in each objective function direction t And save it, set the predicted individual N f The number is 0.5α times the optimal population size N, from the current optimal population X t According to the first objective function, the individuals to be predicted under historical environmental changes are selected equidistantly. And the individual to be predicted The optimal individuals in each target dimension are set as endpoint individuals The individuals other than the endpoint individuals are

[0066] S632, calculate the elite set mean MArc before the last environmental change t-1 Compared with the elite set mean MArc before the contemporary environmental change t The Euclidean distance DmArc t To determine the prediction distance, the endpoints of the contemporary elite set are averaged over MepArc t and the mean value of the contemporary elite set MArc t Respectively and Subtract the mean of to determine the prediction direction;

[0067] S633. Treating predicted individuals The updated prediction individual is obtained by using linear prediction and slight perturbation to predict The specific formula is as follows;

[0068]

[0069] Where Ave(·) represents the mean of all individuals; D is the dimension of the decision variable; and rand is a random number function in [0,1].

[0070] Therefore, the present invention adopts the above-mentioned giant reservoir flood season water level dynamic multi-objective optimization control system and method, which has the following beneficial effects:

[0071] (1) The present invention can automatically monitor environmental changes based on information such as precipitation forecast information, reservoir operation status, and real-time decision-making needs, and can quickly respond to environmental changes and execute intelligent decisions and quickly track optimization problems, thereby flexibly adjusting the reservoir operation plan, maximizing the benefits of the reservoir, and better coordinating the relationship between multiple target needs such as flood control and power generation of the reservoir, thereby improving the comprehensive benefits of the reservoir, providing strong support for the rational use of water resources and flood control, and has good application prospects in engineering.

[0072] (2) The static optimization algorithm in the present invention can select the corresponding algorithm framework according to the number of objective functions in the optimization problem, so as to efficiently and quickly obtain a set of scheduling solutions with good convergence and strong distribution, and reasonably store the excellent scheduling solutions to provide data support for rapid response when the environment changes.

[0073] (3) The environmental detection strategy of the present invention can detect the operating status of the objective function, constraints, and the matching degree of the actual operating status of the reservoir, the change of real-time decision conditions, etc. If the environment changes, it will be transferred to the dynamic response algorithm for processing to achieve automatic processing. In addition, the dynamic response algorithm of the present invention can quickly respond to environmental changes, improve the convergence speed, and achieve rapid tracking response to dynamic multi-objective problems.

[0074] (4) The memory individual adaptation optimization strategy of the present invention makes full use of historical information to guide the evolution of the algorithm, balances the convergence speed and population diversity of the dynamic optimization process algorithm, and has excellent effects in solving dynamic multi-objective optimization problems with periodic characteristics, such as flood optimization scheduling in short-term scheduling and medium- and long-term optimization scheduling problems with periodicity.

[0075] (5) The diversity dynamic introduction strategy of the present invention can improve the global optimization capability while preventing the algorithm from over-exploiting a certain area and falling into the local optimum. In addition, the heterogeneous information supply collaborative prediction strategy of the present invention can predict the dynamic changes of the Pareto frontier and improve the algorithm's tracking efficiency of the Pareto frontier after environmental changes.

[0076] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 A system schematic diagram of a method for dynamic multi-objective optimization control of water level in a giant reservoir during flood season according to the present invention;

[0078] Figure 2 It is a schematic diagram of the effect of the present invention on the FDA dynamic multi-objective test function;

[0079] Figure 3 It is a schematic diagram for comparing storage capacity changes in the actual scheduling problem of the present invention;

[0080] Figure 4 It is a display diagram of the optimization scheduling results of the present invention when the number of environmental changes is different in the actual scheduling problem;

[0081] Figure 5 It is a flow chart of the static scheduling optimization algorithm framework adaptation and elite set selection unit of the present invention;

[0082] Figure 6It is a flow chart of the dynamic scheduling operation response algorithm executed by the present invention. DETAILED DESCRIPTION

[0083] The following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0084] A giant reservoir flood season water level dynamic multi-objective optimization control system, comprising a model initialization module, a runoff forecasting scheme risk quantification assessment module, a giant reservoir flood season water level dynamic multi-objective optimization control method solution module, a GUI interface visualization module, and a calling interface module;

[0085] The solution module of the dynamic multi-objective optimization control method for the water level of giant reservoirs during flood season includes a static scheduling optimization algorithm framework adaptation and elite set selection unit, a non-inferior scheduling scheme set decision analysis unit, a scheduling environment factor detection unit, a dynamic scheduling operation response algorithm unit, and a reservoir optimal scheduling scheme acquisition unit.

[0086] The calling interface module is configured to perform the following actions: the encapsulated giant reservoir flood season water level dynamic control intelligent system configuration integrated service calling interface can be combined with other calling models to generate a multi-object, multi-task, multi-target combined model to realize the on-demand configuration analysis, calculation and calling of the model.

[0087] See also Figure 1-6 , a dynamic multi-objective optimization control method for water level of a giant reservoir during flood season, comprising the following steps:

[0088] Step 1: Construct a multi-objective optimization scheduling model for giant reservoirs and generate an initial optimal population. Based on the basic information of reservoir runoff, operating status, and operating characteristics, construct a multi-objective optimization scheduling model for giant reservoirs with maximum reservoir power generation, maximum peak shaving of the control section, and maximum available water as objective functions during the scheduling period. Add relevant constraints to the multi-objective optimization scheduling model for giant reservoirs, and preset the number of objective functions M, the number of decision variables D, and the length of the forecast period T of the scheduling model. y , static optimization times MaxIter, optimization population size N, elite scheduling solution set size N c , and set the outflow during the reservoir dispatching period as the decision variable; input the relevant data of the giant reservoir, and randomly generate the optimal individual position according to the preset range of the decision variable to complete the initial optimal population generation;

[0089] The objective function for maximizing reservoir power generation during the dispatch period is:

[0090]

[0091] Where E is the total power generation of all reservoirs in the time period; K and t are the number of reservoirs and the current time period respectively; T y is the length of the forecast horizon; P k,t , Q k,t , H k,t are the output, power generation flow and water head of the kth reservoir in the tth period respectively; A k is the output coefficient of the kth reservoir; Δt is the length of the calculation period;

[0092] The maximum peak shaving objective function of the control section during the dispatch period is:

[0093]

[0094] Among them, C is the peak shaving scheduling function; the decision variable is the outflow of the kth reservoir at time t; is the interval inflow forecast value between the kth reservoir and the flood control section at time t; K is the number of reservoirs; M k (·) is the flow process calculation method for routing the outflow process of the kth reservoir to the downstream reservoir or downstream flood control section. When there is a time lag, the Muskingum method is used for routing, otherwise it is directly superimposed;

[0095] The maximum objective function of water supply during the dispatch period is:

[0096]

[0097] Among them, G is the maximum water supply dispatching function during the dispatching period; the decision variable is the water supply flow of the kth reservoir at time t; and are the upper and lower limits of the water supply capacity of the kth reservoir at time t; K is the number of reservoirs.

[0098] Among them, the relevant constraints added to the multi-objective optimization scheduling model of giant reservoirs include water balance constraints, output constraints, reservoir capacity constraints, outflow constraints, power generation flow constraints and scheduling scheme risk constraints. The specific expressions are as follows:

[0099] Water balance constraints:

[0100]

[0101] Among them, I k,t , Q k,t are the interval inflow and outflow of the kth reservoir in the tth period; V k,t is the storage capacity of the kth reservoir in the tth period, V k,t-1is the storage capacity of the kth reservoir in the t-1th period, Δt is the length of the calculation period, Q l,t is the outflow of the lth reservoir in the tth period, T y is the length of the forecast horizon, Ω k are all upstream reservoirs directly connected to reservoir k, Remove lost water from outflows from upstream reservoirs;

[0102] Output constraints:

[0103]

[0104] in, are the maximum and minimum outputs of the kth reservoir in the tth period, respectively;

[0105] Storage capacity constraints:

[0106]

[0107] in, are the maximum and minimum storage capacity of the kth reservoir in the tth period respectively;

[0108] Outbound traffic constraints:

[0109]

[0110] in, are the maximum and minimum outflows of the kth reservoir in the tth period, respectively;

[0111] Generation flow constraints;

[0112]

[0113] in, are the maximum and minimum power generation flows of the kth reservoir in the tth period, respectively;

[0114] Scheduling plan risk constraints:

[0115]

[0116] Among them, R k , are the risk value and minimum risk threshold of the kth scheduling plan respectively.

[0117] Step 2: Execute the risk quantification assessment module of the runoff forecast scheme. Based on the current operation status of the giant reservoir and the Monte Carlo method, a large number of random runoff forecast scheme sets are generated within the forecast interval for risk quantification assessment. Based on the risk quantification assessment results, the risk constraints of the multi-objective optimization scheduling model of the giant reservoir are refined and modified. In the process of optimizing the multi-objective scheduling scheme, the inferior scheduling schemes with higher risk values ​​are gradually eliminated.

[0118] Step 3: Execute the static scheduling optimization algorithm framework adaptation and elite set selection unit, execute the static optimization algorithm selection strategy and elite set selection strategy to complete the optimal optimization strategy adaptation and excellent optimization individual storage operation of the giant reservoir multi-objective scheduling optimization problem in turn; the specific process is as follows:

[0119] S31, execute the static dispatch optimization algorithm framework adaptation strategy, and adapt the corresponding algorithm framework according to the number M of objective functions of the giant reservoir multi-objective optimization dispatch model: when the number M of objective functions is 2, the non-dominated sorting genetic algorithm II framework is used for static optimization; when the number M of objective functions is not less than 3, the non-dominated sorting genetic algorithm III framework is used for static optimization;

[0120] S32, execute the elite set selection strategy for static scheduling optimization, use the elite set to store the set of non-inferior scheduling solutions in the optimization process, the size of the elite scheduling solution set is N c Keeping consistent with the optimal population size N, when executing the elite set storage, the current optimal population X t and the elite set Arc before the environmental change t Merge to perform efficient non-dominated sorting, select the Pareto first frontier individuals of the merged set to update the elite set Arc t , if the number of individuals on the first Pareto frontier does not exceed the size N of the elite scheduling solution set c All are stored, otherwise the elite set individuals that meet the archive number requirements are selected according to the framework of the selected static scheduling optimization algorithm. The non-dominated sorting genetic algorithm II is screened based on the crowding distance, and the non-dominated sorting genetic algorithm III is screened based on the reference point. If the scheduling environment factors change, the elite set is not reset, and the elite set Arc before the environment change is selected. t Randomly select 0.5×N c The elite set Arc after individuals enter a changing environment t+1 .

[0121] Step 4: Execute the non-inferior scheduling solution set decision analysis unit, use the hierarchical analysis method to make decisions on the Pareto frontier non-inferior scheduling solution set obtained by the static scheduling optimization algorithm, and then analyze and match to obtain the most suitable scheduling decision solution under different scheduling requirements and environmental scenario changes;

[0122] Step 5: Execute the dispatching environmental factor detection unit, and use an iterative method to continuously detect whether the precipitation runoff forecast information, reservoir operation status, and real-time decision-making demand environmental factors of the basin and reservoir area where the giant reservoir is located have changed; if the environment has changed, execute step 6; if the environment has not changed, execute step 4 to obtain the optimal dispatching plan for dispatching; in each iteration, the optimal population X is randomly selected. t The objective functions and constraints of the 0.1×N scheduling schemes are tested before and after iteration, as well as the matching degree of the actual operating state of the reservoir and the operating state detection of the changes in real-time decision-making conditions. If it is found that the objective function or constraint of the scheduling scheme being tested is destroyed or the operating state changes, it is determined that the current environment has changed.

[0123] Step 6: Execute the dynamic scheduling operation response algorithm unit, adopt the memory individual adaptation optimization strategy, the diversity dynamic introduction strategy and the heterogeneous information supply collaborative prediction strategy to quickly respond to environmental changes, and quickly track and respond to the dynamic multi-objective scheduling model; the specific process is as follows:

[0124] S61, execute the memory individual adaptation optimization strategy, by saving the excellent solutions and valuable information found in the past, after the environment changes, a relatively excellent solution can be quickly found and the impact of environmental uncertainty can be reduced. This strategy stores the elite set Arc before the historical environmental changes to the memory archive set MS, and when the environment changes, it matches the memory archive set and selects the historical optimal solution with the highest matching degree in the current environment to respond; the specific process is as follows:

[0125] S611, first of all, the elite set Arc before each environment change t Store it in the memory archive set MS, and set the memory storage period length to 25% of the optimal population size. If the memory archive is full, the archive is updated according to the first-in-first-out principle; if the storage period length is 25, when the archive MS 1-25 If the storage is done, the next time the environment changes, the MS 26 Replace MS1, that is, the memory save is MS 2-26 ;

[0126] S612: If the environment changes, each memory archive MS is updated from the memory archive MS. t According to the ascending order of the first objective function, M+3 individuals are selected at equal intervals based on the change of the number of objective functions M to ensure the globality of the selected individuals;

[0127] S613, for each archived MS tThe selected individuals use the Euclidean distance to calculate the distance of the objective function value before and after the environmental change and divide it by the maximum value of the objective function value before and after the corresponding change, eliminate the dimension, and add the obtained distances to obtain the distance and DMS of each archived individual. t , if there is 0 in the denominator, the data will be removed;

[0128] S614, finally found DMS t The smallest value among them, takes the corresponding memory archive number m as the adaptation archive, and selects the adaptation archive MS m A scheduling scheme with a size of 0.2α times the optimal population is randomly selected as the memory scheduling scheme

[0129]

[0130] in, They represent the objective function values ​​of the individuals selected in the memory archive before and after the environmental changes; E_distance represents the calculation using Euclidean distance; randperm(I,I1) is a random selection function, randomly selecting I1 individuals from I, α represents the adjustment parameter, and N represents the optimal population size.

[0131] S62. Execute the dynamic introduction strategy of diversity. Through the dynamic introduction strategy of diversity, a diverse search is performed in the search space to ensure that the diversity of the population is maintained when the environment changes greatly, is difficult to predict or is irregular. This strategy considers the density of the population through a loose population crowding distance calculation method, and generates new individuals through mutation and initialization to replace the densely distributed individuals in the population, thereby maintaining the diversity of the population. The specific process is as follows:

[0132] S621. First determine the number of individuals introduced by diversity The optimal population size N is α times. Considering that the response direction after environmental changes is uncertain, the individual numbers of the introduced diversity are selected through the loose crowding distance optimization method to select individuals with close distances for replacement screening. By calculating the current optimal individual X t The crowding distance is then sorted in ascending order according to the crowding distance, and a random selection is made from the optimal population that is (1-α)×N times larger than the distance sorting. Introduce diversity into individuals;

[0133] S622, secondly, a multi-dimensional dynamic optimization method considering the optimization population size and the number of static optimizations is used to introduce diverse individuals, which include mutant individuals and initialized individuals. The quantity is set to Initialize individual The quantity is set to The mutation adopts polynomial mutation, and the mutation probability and distribution index are set to 1 and Initialization uses random generation within the upper and lower limits of the decision variable. The specific expression is as follows:

[0134]

[0135] Among them, Crowding(·) is the distance value of the corresponding individual obtained by the crowding distance method; Sort(·) is the ascending sorting function; randperm(I,I1) is the random selection function, which randomly selects I1 individuals from I; Respectively represent the use of polynomial mutation and initialization for the selected mutant individuals and the initialized individuals; P_mutation(a,b) represents the use of polynomial mutation method, a and b are the mutation probability and distribution index respectively; Express The individual's decision variables are randomly generated functions within the upper and lower limits.

[0136] S63, implement the collaborative prediction strategy of heterogeneous information supply, try to foresee or estimate the changing trend of future optimization problems by combining historical information of environmental changes, so as to better adapt to environmental changes and quickly find feasible solution sets after environmental changes. This strategy calculates the position of the elite set before and after environmental changes, predicts the direction and degree of environmental changes, and generates new individuals through linear prediction and slight disturbance to quickly respond to environmental changes. The specific process is as follows:

[0137] S631, first of all, the elite set Arc before the environment changes t Sort the non-dominated solutions, store the Pareto first frontier individuals and update the elite set, and then calculate the elite set mean MArc t and the mean MepArc of the endpoint individuals with the best fitness values ​​in each objective function direction t And save it, set the predicted individual N f The number is 0.5α times the optimal population size N, from the current optimal population X t According to the first objective function, the individuals to be predicted under historical environmental changes are selected equidistantly. And the individual to be predicted The optimal individuals in each target dimension are set as endpoint individuals Individuals other than the endpoint individuals are

[0138] S632, calculate the elite set mean MArc before the last environmental change t-1 Compared with the elite set mean MArc before the contemporary environmental change t The Euclidean distance DmArc t To determine the prediction distance, the endpoints of the contemporary elite set are averaged over MepArc tand the mean value of the contemporary elite set MArc t Respectively and Subtract the mean of to determine the prediction direction;

[0139] S633. Treating predicted individuals The updated prediction individual is obtained by using linear prediction and slight perturbation to predict The specific formula is as follows;

[0140]

[0141] Where Ave(·) represents the mean of all individuals; D is the dimension of the decision variable; and rand is a random number function in [0,1].

[0142] Step 7, execute the reservoir optimal scheduling plan acquisition unit to determine whether the preset number of environmental changes has been reached. If the preset number of environmental changes has been reached, execute the static optimization algorithm and decision model to output the reservoir operation status data and the current and historical optimal scheduling plan set. Otherwise, return the scheduling plan output in step 6 to step 3 for continued execution.

[0143] Example

[0144] In order to verify the effectiveness and feasibility of the algorithm of the present invention, the classic dynamic multi-objective test problem FDA is first tested, where the parameters N, n t , τ t and D represent the population size, the intensity of environmental change, the number of static optimizations and the dimension of decision variables, respectively. Each test problem undergoes 50 environmental changes and 20 independent calculations. The mean inverse generation distance (MIGD) and the mean hypervolume (MHV) are used to evaluate the convergence and distribution performance indicators of the final solution set. The rank sum test and the Friedman test are used to test the difference between the present invention and other comparison algorithms, where "+", "-", "=" indicate that the performance of other algorithms is statistically better than, lower than or equal to the present invention. The test results of the present invention are shown in Table 1, and the visualization of the test results is shown in Figure 2 .

[0145] Table 1 MIGD and MHV test results of the algorithm on the FDA benchmark function

[0146]

[0147] Result analysis:

[0148] First, by performing 20 independent calculations on the FDA test function for the three methods, as shown in Table 1, it can be seen that the MIGD index and MHV index of the present invention on the FDA problem are better than those of other solutions, and both the rank sum test and the Friedman test show significant advantages. In addition, it also shows excellent STD. Then, the effect of the present invention on the FDA test function is demonstrated, as shown in Table 1. Figure 2 It can be seen that the Pareto front of the present invention is very close to the ideal Pareto front. Therefore, the results on the FDA test function can show that the dynamic response effect, static optimization effect, convergence, distribution and stability of the present invention are all excellent, which verifies the effectiveness and feasibility of the algorithm of the present invention.

[0149] Secondly, the dynamic control model of water level in flood season is solved by the three methods under the extreme low water dispatching scenario of a certain basin reservoir, and the real-time decision flow when the environment changes is extracted to calculate the objective function and reservoir capacity change results of the decision variables in 11 change periods. The specific solution results are shown in Table 2 and Figure 3 ;

[0150] Table 2 Test results of the algorithm on actual scheduling problems

[0151]

[0152] It can be seen that: in the scheduling scenario of extreme low water, although the performance of the present invention in power generation efficiency is relatively poor, it has significant advantages in peak shaving, water supply and final storage capacity. This shows that the present invention can better combine the actual reservoir status and scheduling scenario when the forecast period is 7 days, and appropriately sacrifice the real-time power generation efficiency when optimizing scheduling, and convert the power generation efficiency into water supply efficiency and reservoir water replenishment efficiency, which better meets the needs of actual scheduling. Finally, by displaying the optimization scheduling results when the number of environmental changes is 1, 6 and 11, in order to facilitate optimization and display, each objective function is optimized as small as possible, so a negative sign is added to the optimization results of power generation and water supply, as shown in detail. Figure 4 It can be seen that the optimization effect of the present invention on power generation and peak shaving function under various environmental changes is not much different from that of other algorithms, but with the increase in the number of environmental changes, the present invention is more and more excellent in optimizing water supply. Therefore, in actual scheduling problems, the optimization results of the present invention are reliable and practical, can provide the best decision-making plan and meet the needs of actual scheduling.

[0153] Therefore, the present invention adopts the above-mentioned giant reservoir flood season water level dynamic multi-objective optimization control system and method, and responds to the environment quickly according to precipitation forecast information, reservoir operation status, real-time decision-making needs and other information, so as to flexibly adjust the reservoir operation plan, maximize the benefits of the reservoir, better coordinate the relationship between multiple target needs such as flood control and power generation of the reservoir, and improve the comprehensive benefits of the reservoir. It has the outstanding advantages of accurate environmental change tracking, rapid response, better solution convergence and distribution, efficient scheduling plan decision-making, and a wide range of applicable scenarios, and can provide technical support for the dynamic control of the water level of a reservoir group during the flood season.

[0154] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A dynamic multi-objective optimization control method for water level in giant reservoirs during flood season, characterized in that: The following steps are involved: Step 1: Construct a multi-objective optimization scheduling model for giant reservoirs and generate an initial optimal population. Based on the basic information of reservoir runoff, operating status, and operating characteristics, construct a multi-objective optimization scheduling model for giant reservoirs with maximum reservoir power generation, maximum peak shaving of the control section, and maximum available water as objective functions during the scheduling period. Add relevant constraints to the multi-objective optimization scheduling model for giant reservoirs and preset the number of objective functions of the scheduling model. M , the number of decision variables D , Long forecast period T y , static optimization times MaxIter Optimal population size N , the size of the elite scheduling solution set N c , and set the outflow during the reservoir dispatching period as the decision variable; input the relevant data of the giant reservoir, and randomly generate the optimal individual position according to the preset range of the decision variable to complete the initial optimal population generation; Step 2: Execute the risk quantification assessment module of the runoff forecast scheme. Based on the current operation status of the giant reservoir and the Monte Carlo method, a large number of random runoff forecast scheme sets are generated within the forecast interval for risk quantification assessment. Based on the risk quantification assessment results, the risk constraints of the multi-objective optimization scheduling model of the giant reservoir are refined and modified. In the process of optimizing the multi-objective scheduling scheme, the inferior scheduling schemes with higher risk values ​​are gradually eliminated. Step 3: Execute the static scheduling optimization algorithm framework adaptation and elite set selection unit, execute the static optimization algorithm selection strategy and elite set optimization strategy to complete the optimal optimization strategy adaptation and excellent optimization individual storage operation of the giant reservoir multi-objective scheduling optimization problem in turn; Step 4: Execute the non-inferior scheduling solution set decision analysis unit, use the hierarchical analysis method to make decisions on the Pareto frontier non-inferior scheduling solution set obtained by the static scheduling optimization algorithm, and then analyze and match to obtain the most suitable scheduling decision solution under different scheduling requirements and environmental scenario changes; Step 5, execute the dispatching environmental factor detection unit, and use an iterative method to continuously detect whether the precipitation runoff forecast information, reservoir operation status, and real-time decision-making demand environmental factors of the basin and reservoir area where the giant reservoir is located have changed; if the environment has changed, execute step 6; if the environment has not changed, execute step 4 to obtain the optimal dispatching plan for dispatching; Step 6: Execute the dynamic scheduling operation response algorithm unit, adopt the memory individual adaptation optimization strategy, the diversity dynamic introduction strategy and the heterogeneous information supply collaborative prediction strategy to quickly respond to environmental changes, and quickly track and respond to the dynamic multi-objective scheduling model; Step 7, execute the reservoir optimal dispatching scheme acquisition unit to determine whether the preset number of environmental changes is reached. If the preset number of environmental changes is reached, execute the static optimization algorithm and decision model to output the reservoir operation status data and the current and historical optimal dispatching scheme set. Otherwise, return the dispatching scheme output in step 6 to step 3 for further execution; The objective function for maximizing reservoir power generation during the dispatch period in step 1 is: in, E is the total power generation of all reservoirs during the period; K, t are the number of reservoirs and the current period, respectively; T y is the length of the forecast horizon; , , are the output, power generation flow and hydraulic head of the kth reservoir in the tth period respectively; is the output coefficient of the kth reservoir; To calculate the time period; The maximum peak shaving objective function of the control section during the dispatch period is: in, C is the peak shaving scheduling function; the decision variable is the outflow of the kth reservoir at time t; is the interval inflow forecast value between the kth reservoir and the flood control section at time t; K is the number of reservoirs; The flow process calculation method for calculating the outflow process of the kth reservoir to the downstream reservoir or the downstream flood control section is to use the Muskingum method for calculation when there is a time lag, otherwise it is directly superimposed; The maximum objective function of water supply during the dispatch period is: in, G is the maximum water supply dispatching function during the dispatching period; the decision variable is the water supply flow of the kth reservoir at time t; and are the upper and lower limits of the water supply capacity of the kth reservoir at time t; K is the number of reservoirs; The specific process of executing the dynamic scheduling operation response algorithm unit is as follows: S61, execute the memory individual adaptation optimization strategy, by optimizing the elite set before each environmental change. Arc t Save to Memory File Set MS , when the environment changes, it matches the memory archive set and selects the historical optimal solution with the highest degree of match in the current environment to respond; S62. Implement a dynamic diversity introduction strategy, taking into account the density of the population through a population crowding distance calculation method, and generating new individuals through mutation and initialization to replace the densely distributed individuals in the population to maintain the diversity of the population; S63, implement the collaborative prediction strategy of heterogeneous information supply, predict the direction and degree of environmental change by calculating the position of the elite set before and after environmental changes, and generate new individuals through linear prediction and slight disturbance to quickly respond to environmental changes; The specific process of executing the memory individual adaptation optimization strategy in S61 is as follows: S611, first of all, before each environmental change, the elite group Arc t Save to memory archive MS The memory storage period length is set to 25% of the optimal population size. If the memory archive is full, the archive is updated according to the first-in-first-out principle. S612, if the environment changes, save from memory MS Archive each memory MS t Sort the first objective function in ascending order and select M+3 individuals at equal intervals based on the change in the number of objective functions M; S613, for each archive MS t The selected individuals use the Euclidean distance to calculate the distance of the objective function value before and after the environmental change and divide it by the maximum value of the objective function value before and after the corresponding change, eliminate the dimension, and add the distances obtained to obtain the distance and DMS t , if there is 0 in the denominator, the data will be removed; S614, finally found DMS t The smallest value among them, takes the corresponding memory archive number m as the adaptation archive, and selects the adaptation archive from MS m Randomly select The scheduling scheme of the optimal population size is determined to be the memory scheduling scheme ; in, , They represent the objective function values ​​of the individuals selected in the memory archive before and after the environmental changes; The representative uses Euclidean distance for calculation; is a random selection function, from I Random Selection I 1 Individuals, represents the adjustment parameters, represents the optimal population size; The specific process of implementing the diversity dynamic introduction strategy in S62 is as follows: S621. First determine the number of individuals introduced by diversity In order to find the optimal population size N α times, the individual numbers of the diversity are introduced through the crowding distance optimization method, and the individuals with close distance are selected for replacement screening. X t The crowding distance is then sorted in ascending order according to the crowding distance, and randomly selected from the optimal population that is (1-α)×N times larger than the distance sorting. Introduce diversity into individuals; S622, secondly, a multi-dimensional dynamic optimization method considering the optimization population size and the number of static optimizations is used to introduce diverse individuals, which include mutant individuals and initialized individuals. The quantity is set to , initialize individual The quantity is set to , where polynomial mutation is used, and the mutation probability and distribution index are set to 1 and , the initialization is randomly generated within the upper and lower limits of the decision variable, and the specific expression is as follows: in, The crowding distance method is used to obtain the distance value of the corresponding individual; It is a function for sorting in ascending order; is a random selection function, from I Random Selection I 1 individual; , Respectively represent the use of polynomial mutation and initialization for the selected mutation individuals and initialization individuals; It represents the use of polynomial mutation method, where a and b are mutation probability and distribution index respectively; Express The individual's decision variables randomly generate functions within the upper and lower limits; The specific process of executing the collaborative prediction strategy of heterogeneous information supply in S63 is as follows: S631, first of all, before each environmental change, the elite group Arc t Sort the non-dominated solutions, store the Pareto first frontier individuals and update the elite set, then calculate the elite set mean MArc t and the mean of the endpoint individuals with the best fitness values ​​in each objective function direction MepArc t And archive, set the prediction individual N f The number is 0.5α times the size of the optimal population N, from the current optimal population According to the first objective function, the individuals to be predicted before each environmental change are selected equidistantly. , and the individual to be predicted The optimal individuals in each target dimension are set as endpoint individuals , except for the endpoint individuals, ; S632, calculate the elite set mean before the last environmental change Compared with the mean of the elite set before the contemporary environmental changes MArc t The Euclidean distance DmArc t To determine the prediction distance, the endpoints of the contemporary elite set are averaged MepArc t and the mean of the contemporary elite MArc t Respectively and Subtract the mean of to determine the prediction direction; S633. Treating predicted individuals The updated prediction individual is obtained by using linear prediction and slight perturbation to predict , the specific formula is as follows; in, It means taking the mean of all individuals; D is the dimension of decision variable; rand is the random number function in [0,1].

2. A method for dynamic multi-objective optimization control of water level in giant reservoirs during flood season according to claim 1, characterized in that: The relevant constraints added to the giant reservoir multi-objective optimization scheduling model in step 1 include water balance constraints, output constraints, reservoir capacity constraints, outflow constraints, power generation flow constraints and scheduling scheme risk constraints. The specific expressions are as follows: Water balance constraints: in, , are the interval inflow and outflow of the kth reservoir in the tth period, respectively; is the storage capacity of the kth reservoir in the tth period, For the t-1 The storage capacity of the kth reservoir in the time period is To calculate the length of the period, is the tth period l The outflow of a reservoir, T y is the length of the forecast horizon, are all upstream reservoirs directly connected to reservoir k, Remove lost water from outflows from upstream reservoirs; Output constraints: in, , are the maximum and minimum outputs of the kth reservoir in the tth period, respectively; Storage capacity constraints: in, , are the maximum and minimum storage capacity of the kth reservoir in the tth period respectively; Outbound traffic constraints: in, , are the maximum and minimum outflows of the kth reservoir in the tth period, respectively; Generation flow constraints; in, , are the maximum and minimum power generation flows of the kth reservoir in the tth period, respectively; Scheduling plan risk constraints: in, , are the risk value and minimum risk threshold of the kth scheduling plan respectively.

3. A method for dynamic multi-objective optimization control of water level in giant reservoirs during flood season according to claim 2, characterized in that: The specific process of executing the static scheduling optimization algorithm framework adaptation and elite set selection unit in step 3 is as follows: S31. Execute the static dispatch optimization algorithm framework adaptation strategy, based on the number of objective functions of the giant reservoir multi-objective optimization dispatch model M Adapt the corresponding algorithm framework: when the number of objective functions M When the number of objective functions is 2, the non-dominated sorting genetic algorithm II framework is used for static optimization; when the number of objective functions is M When it is not less than 3, the non-dominated sorting genetic algorithm III framework is used for static optimization; S32, execute the elite set selection strategy for static scheduling optimization, use the elite set to store the set of non-inferior scheduling solutions in the optimization process, and the size of the elite scheduling solution set N c and optimal population size N Keep consistent, when executing elite set storage, the current optimal population and the elite group before the environmental change Arc t Merge to perform efficient non-dominated sorting, select the Pareto first frontier individuals of the merged set to update the elite set Arc t , if the number of individuals on the first Pareto frontier does not exceed the size of the elite scheduling solution set N c All are stored, otherwise the elite set individuals that meet the number of archive requirements are selected according to the framework of the selected static scheduling optimization algorithm. The non-dominated sorting genetic algorithm II is screened based on the crowding distance, and the non-dominated sorting genetic algorithm III is screened based on the reference point. If the scheduling environment factors change, the elite set is not reset, and the elite set before the environment change is selected. Arc t Randomly select 0.5× N c The elite group after individuals enter the changing environment Arc t+1 .

4. A method for dynamic multi-objective optimization control of water level in giant reservoirs during flood season according to claim 3, characterized in that: In step 5, the optimal population is randomly selected in each iteration Medium 0.1× N The objective function and constraints of each scheduling scheme are tested before and after iteration, as well as the matching degree of the actual operating state of the reservoir and the operating state detection of real-time decision-making conditions. If it is found that the objective function or constraint of the scheduling scheme being tested is destroyed or the operating state changes, it is determined that the current environment has changed.

5. A system for a dynamic multi-objective optimization control method for a giant reservoir water level during flood season according to any one of claims 1 to 4, characterized in that: It includes model initialization module, runoff forecasting scheme risk quantification assessment module, giant reservoir flood season water level dynamic multi-objective optimization control method solution module, GUI interface visualization module, and calling interface module; The giant reservoir flood season water level dynamic multi-objective optimization control method solution module includes a static scheduling optimization algorithm framework adaptation and elite set selection unit, a non-inferior scheduling scheme set decision analysis unit, a scheduling environment factor detection unit, a dynamic scheduling operation response algorithm unit, and a reservoir optimal scheduling scheme acquisition unit.

Images