Cascade reservoir dispatching method based on multi-population self-adaption

Abstract

The invention discloses a cascade reservoir scheduling method based on multi-population self-adaption, which comprises the following steps: establishing a cascade reservoir scheduling (CRS) model, taking maximized power generation, maximized desilting and maximized ecological rate as optimization targets of the CRS model, and determining constraint conditions according to the optimization targets of the CRS model and operation requirements of a cascade reservoir system; and then performing operation on the optimization target by using a cascade reservoir scheduling method based on multi-population self-adaption. According to the invention, by designing the single-target population and the double-target population, the conflict difficulty between target optimization and constraint satisfaction is reduced; waste of computing resources caused by low-efficiency populations is reduced through a population self-adaptive activation mechanism; by designing an environment selection mechanism based on bidirectional information sharing, the effectiveness of knowledge migration is improved, and the calculation complexity is reduced; the method is high in robustness and can be suitable for reservoir data in different years.

Description

Technical Field

[0001] This invention relates to a reservoir system scheduling technology, and more particularly to a multi-group adaptive cascade reservoir scheduling method. Background Technology

[0002] Reservoir systems, as a key component of modern water resource management, play a crucial role in hydropower generation, flood control, irrigation, and ecological protection. These systems are even more important in regions like China where water resources are unevenly distributed in time and space. Due to varying reservoir demands at different locations, reservoirs along rivers are often operated in conjunction, forming cascade reservoir systems. Cascade reservoir scheduling (CRS) requires the rational allocation of water storage capacity in each reservoir while satisfying its own inherent characteristics, thereby maximizing socio-economic benefits. However, the strong constraints imposed by water balance constraints, the large-scale optimization variables resulting from time-based scheduling, and the strong coupling relationships between reservoirs make the CRS problem extremely challenging.

[0003] Currently, existing multi-objective evolutionary algorithms are commonly used to solve this problem, but new targeted solution strategies have not been designed based on the characteristics of the problem, resulting in ineffective solutions to the CRS problem. For example, widely used evolutionary algorithms, including NSGAII and ARMOEA, are directly used in most studies to solve the constructed cascade reservoir model. However, these methods are designed based on the general common characteristics of multi-objective optimization problems and do not consider the specific difficulties of the CRS problem, thus limiting their performance. In addition, these methods adopt simple feasibility-oriented constraint handling techniques, which can easily lead to the algorithm getting trapped in the local feasible region prematurely. Other studies have designed new solution algorithms, but most focus on improving existing offspring generation strategies or constraint handling techniques. Although these strategies improve algorithm performance, a deeper analysis of the characteristics and difficulties of the CRS problem, such as the conflict and synergistic relationship between objective optimization and constraint satisfaction, means that there is still significant room for improvement in algorithm performance. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes a multi-group adaptive cascade reservoir scheduling method, providing a more diverse and convergent solution for cascade reservoir scheduling.

[0005] The technical solution adopted in this invention is as follows:

[0006] A multi-group adaptive cascade reservoir scheduling method includes the following steps:

[0007] S1, Establish the CRS model; such as Figure 1As shown, the reservoirs in the model originate from the Yellow River, and from upstream to downstream they are Longyangxia, Liujiaxia, Haibowan, Wanjiazhai, Sanmenxia, ​​and Xiaolangdi. In addition, the model also includes some hydrological stations, such as Lanzhou, Toudaoguai, Aishan, Luokou, and Lijin.

[0008] S2 takes maximizing power generation, maximizing sediment discharge, and maximizing ecological efficiency as the optimization objectives of the CRS model;

[0009] S3, Determine the constraints based on the optimization objective of CRS and the operational requirements of the cascade reservoir system;

[0010] S4 utilizes a multi-group adaptive cascade reservoir scheduling method to perform operations on the optimization objective.

[0011] Furthermore, the optimization objectives include maximizing power generation, maximizing sediment discharge, and maximizing ecological satisfaction rate;

[0012] The goal of maximizing power generation:

[0013]

[0014] In the formula, N1 and T represent the number of reservoirs and the total number of scheduling periods related to f1, respectively, and k i,t Q is the output coefficient. i,t Let m be the power generation flow of the i-th reservoir in time period t (m). 3 / s), H represents the hydroelectric head (m), Δσ t Let t be the duration (s) of the t-th time interval;

[0015] The goal of maximizing sediment removal:

[0016]

[0017] In the formula, N2 represents the number of relevant reservoirs. Indicates the amount of sand transported (kg);

[0018] The goal of maximizing ecological satisfaction rate:

[0019]

[0020] In the formula, N3 represents the number of relevant reservoirs, and W i,t,E Suitable ecological flow (m 3 / s), It is the actual outbound flow (m 3 / s).

[0021] Furthermore, the constraints include water balance constraints, water level constraints, flow constraints, and water level constraints at the end of the scheduling period;

[0022] The water balance constraint:

[0023]

[0024] In the formula, q i,t Represents the actual water supply (m³) 3 / s);

[0025] The water level constraint:

[0026]

[0027] In the formula, Z i,t This represents the water level (m) of the i-th reservoir at the end of the t-th scheduling cycle. and These represent the corresponding minimum and maximum water level limits, respectively.

[0028] The flow constraint:

[0029]

[0030] In the formula, Q i,t Represents the input / output flow (m³) of the i-th reservoir in the t-th scheduling cycle. 3 / s), and These represent the corresponding minimum and maximum input / output flow limits, respectively.

[0031] The water level constraint at the end of the scheduling period:

[0032]

[0033] In the formula: V i,0 and V i,T These represent the water storage capacity of the i-th reservoir at the start and end of the scheduling process (10). 8 m 3 V i bejin and V i end This represents the actual water storage volume.

[0034] Furthermore, S4 includes the following steps:

[0035] S401, initialize parameter u = 0, parameter k = 2, population size NP = 100;

[0036] S402, such as Figure 2 As shown, population P3 is randomly initialized to optimize three objectives and all constraints, and its optimization includes the primal problem with three optimization objectives and all constraints; population P2 is randomly initialized to optimize two objectives and all constraints, including... (The superscript represents the optimization objective index, for example, 1-2 represents the first and second objectives that the population optimizes), which optimizes two objectives and all constraints respectively; a population P1 is randomly initialized to optimize one objective and all constraints, including P1 1 P1 2 P1 3 Each optimizes a single objective and all constraints respectively; initializes an archive A and copies P3 to A;

[0037] S403 records the computing resources used: FE = 7 × NP;

[0038] S404, determine if FE is greater than MaxFE×u÷k. ​​If it is, jump to S405; otherwise, jump to S407.

[0039] S405, executes the population adaptive activation mechanism;

[0040] S406, set u = u + 1, and set A = {P1, P2, P3};

[0041] S407, P3 The differential evolution algorithm is used to generate NP / 2 offspring individuals respectively;

[0042] S408, for P1 1 P1 2 P1 3 The activated population uses a differential evolution algorithm to generate NP / 2 offspring individuals.

[0043] S409, Update Record A = A∪{O1,O2,O3};

[0044] S410 executes an environment selection mechanism based on two-way information sharing;

[0045] S411: Determine if FE is greater than the set maximum computing resource MaxFE. If it is, end and output P3; otherwise, jump to S404.

[0046] Furthermore, the specific steps of S405 include:

[0047] S4051, extract the constraint value of an individual in A, denoted as G;

[0048] S4052, set L = {0,0,0}, i = 1;

[0049] S4053, extract the i-th target value from individual A, denoted as Q;

[0050] S4054, calculate the correlation coefficient between Q and G using the Spearman rank correlation coefficient method:

[0051]

[0052] S4055, if ρ is less than 0, then set L. i =1;

[0053] S4056, set i = i + 1. If i is greater than 3, end and output L; otherwise, jump to S4053.

[0054] Furthermore, the specific steps of S410 include:

[0055] S4101, Settings

[0056] S4102,

[0057]

[0058] S4103,

[0059] S4104, P1 1 Use the epsilon method from the collection Select NP individuals to form a new P1 1 The second and third target values ​​are set to 0;

[0060] S4105, P1 2 Use the epsilon method from the collection Select NP individuals to form a new P1 2 The first and third target values ​​are set to 0;

[0061] S4106, P1 3 Use the epsilon method from the collection Select NP individuals to form a new P1 3 The first and second target values ​​are set to 0;

[0062] S4107, Use the epsilon method from the collection Select NP individuals to form a new group. The third target value is set to 0;

[0063] S4108, Use the epsilon method from the collection Select NP individuals to form a new group. The second target value is set to 0;

[0064] S4109, Use the epsilon method from the collection Select NP individuals to form a new group. The first target value is set to 0;

[0065] S41010, P3 uses the epsilon method to select NP individuals from the set {P3, C3} to form a new P3.

[0066] The beneficial effects of this invention are:

[0067] 1. This invention analyzes the problem from the perspective of constrained multi-objectives, including the relationship between constraint difficulty, constraint satisfaction and optimization of different objectives, and designs targeted optimization strategies, including population adaptive evolution strategy, layer-based population structure, and layer-based information sharing mechanism, to improve the robustness and accuracy of the algorithm on the CRS problem.

[0068] 2. This invention reduces the difficulty of conflict between objective optimization and constraint satisfaction by designing single-objective and dual-objective populations; reduces the waste of computing resources by inefficient populations by using a population adaptive activation mechanism; and improves the effectiveness of knowledge transfer and reduces computational complexity by designing an environment selection mechanism based on bidirectional information sharing.

[0069] 3. The method of the present invention is robust and can be applied to reservoir data in different years.

[0070] 4. Practical applications show that the present invention achieves better results than the comparison method in terms of diversity and convergence values, proving the superiority and stability of the method, and thus providing technical support for practical scheduling. Attached Figure Description

[0071] Figure 1 This is a watershed map of the reservoir in the CRS model of this invention;

[0072] Figure 2 This is a schematic diagram of the framework structure of the optimization method of the present invention. Detailed Implementation

[0073] To make the technical concept and advantages of the invention clearer, the technical solution of the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the following embodiments are merely illustrative of preferred embodiments of the present invention and are not intended to limit the scope of patent protection claimed by the present invention.

[0074] Example 1

[0075] A multi-group adaptive cascade reservoir scheduling method includes the following steps:

[0076] S1, Establish a CRS model based on the reservoir;

[0077] like Figure 1As shown, the reservoirs in the model originate from the Yellow River, and from upstream to downstream they are Longyangxia, Liujiaxia, Haibowan, Wanjiazhai, Sanmenxia, ​​and Xiaolangdi. In addition, the model also includes some hydrological stations, such as Lanzhou, Toudaoguai, Aishan, Luokou, and Lijin.

[0078] S2 takes maximizing power generation, maximizing sediment discharge, and maximizing ecological efficiency as the optimization objectives of the CRS model;

[0079] S3, Determine the constraints based on the optimization objective of CRS and the operational requirements of the cascade reservoir system;

[0080] S4 utilizes a multi-group adaptive cascade reservoir scheduling method to perform operations on the optimization objective.

[0081] The optimization objectives include maximizing power generation, maximizing sediment discharge, and maximizing ecological satisfaction.

[0082] Formula for maximizing power generation target:

[0083]

[0084] In the formula, N1 and T represent the number of reservoirs and the total number of scheduling periods related to f1, respectively, and k i,t Q is the output coefficient. i,t Let m be the power generation flow of the i-th reservoir in time period t (m). 3 / s), H represents the hydroelectric head (m), Δσ t Let t be the duration (s) of the t-th time interval;

[0085] Formula for maximizing sediment removal target:

[0086]

[0087] In the formula, N2 represents the number of relevant reservoirs. Indicates the amount of sand transported (kg);

[0088] The formula for maximizing ecological satisfaction rate is as follows:

[0089]

[0090] In the formula, N3 represents the number of relevant reservoirs, and W i,t,E Suitable ecological flow (m 3 / s), It is the actual outbound flow (m 3 / s).

[0091] The constraints include water balance constraints, water level constraints, flow constraints, and water level constraints at the end of the scheduling period;

[0092] The formula for calculating water balance constraints is as follows:

[0093]

[0094] In the formula, q i,t Represents the actual water supply (m³) 3 / s);

[0095] The water level constraint must meet the following conditions:

[0096]

[0097] In the formula, Z i,t This represents the water level (m) of the i-th reservoir at the end of the t-th scheduling cycle. and These represent the corresponding minimum and maximum water level limits, respectively.

[0098] The flow constraint must satisfy the following conditions:

[0099]

[0100] In the formula, Q i,t Represents the input / output flow (m³) of the i-th reservoir in the t-th scheduling cycle. 3 / s), and These represent the corresponding minimum and maximum input / output flow limits, respectively.

[0101] The water level constraints at the end of the scheduling period must meet the following conditions:

[0102]

[0103] In the formula: V i,0 V i,T These represent the water storage capacity of the i-th reservoir at the start and end of the scheduling process (10). 8 m 3 V i bejin and V i end Represents the actual water storage capacity (10 8 m 3 ).

[0104] Furthermore, S4 specifically includes the following steps:

[0105] S401, initialize parameter u = 0, parameter k = 2, population size NP = 100;

[0106] S402, such as Figure 2As shown, population P3 is randomly initialized to optimize three objectives and all constraints, and its optimization includes the primal problem with three optimization objectives and all constraints; population P2 is randomly initialized to optimize two objectives and all constraints, including... (The superscript represents the optimization objective index, for example, 1-2 represents the first and second objectives that the population optimizes), which optimizes two objectives and all constraints respectively; a population P1 is randomly initialized to optimize one objective and all constraints, including P1 1 P1 2 P1 3 Each optimizes a single objective and all constraints respectively; initializes an archive A and copies P3 to A;

[0107] S403 records the computing resources used: FE = 7 × NP;

[0108] S404, determine if FE is greater than MaxFE×u÷k. ​​If it is, jump to S405; otherwise, jump to S407.

[0109] S405, executes the population adaptive activation mechanism;

[0110] S406, set u = u + 1, and set A = {P1, P2, P3};

[0111] S407, P3 Using the differential evolution algorithm, NP / 2 offspring individuals are generated respectively;

[0112] S408, for P1 1 P1 2 P1 3 The activated population uses a differential evolution algorithm to produce NP / 2 offspring individuals.

[0113] S409, Update Record A = A∪{O1,O2,O3};

[0114] S410 executes an environment selection mechanism based on two-way information sharing;

[0115] S411: Determine if FE is greater than the set maximum computing resource MaxFE. If it reaches this value, end and output P3; otherwise, jump to S404.

[0116] Example 2

[0117] This is basically the same as Example 1, except that it includes the following steps in addition to Example 1:

[0118] Furthermore, the specific steps of S405 include:

[0119] S4051, extract the constraint value of an individual in A, denoted as G;

[0120] S4052, set L = {0,0,0}, i = 1;

[0121] S4053, extract the i-th target value from individual A, denoted as Q;

[0122] S4054, using the Spearman rank correlation coefficient method, calculate the correlation coefficient between Q and G:

[0123]

[0124] S4055, if ρ is less than 0, then set L. i =1;

[0125] S4056, set i = i + 1. If i is greater than 3, end and output L; otherwise, jump to S4053.

[0126] Furthermore, the specific steps of S410 include:

[0127] S4101, Settings

[0128] S4102,

[0129]

[0130] S4103,

[0131] S4104, P1 1 Use the epsilon method from the collection Select NP individuals to form a new P1 1 The second and third target values ​​are set to 0;

[0132] S4105, P1 2 Use the epsilon method from the collection Select NP individuals to form a new P1 2 The first and third target values ​​are set to 0;

[0133] S4106, P1 3 Use the epsilon method from the collection Select NP individuals to form a new P1 3 The first and second target values ​​are set to 0;

[0134] S4107, Use the epsilon method from the collection Select NP individuals to form a new group. The third target value is set to 0;

[0135] S4108, Use the epsilon method from the collection Select NP individuals to form a new group. The second target value is set to 0;

[0136] S4109, Use the epsilon method from the collection Select NP individuals to form a new group. The first target value is set to 0;

[0137] S41010, P3 uses the epsilon method from the set Select NP individuals to form a new P3.

[0138] Example 3

[0139] To verify the effectiveness of the method of this invention, it was applied to data from nine real Yellow River reservoirs, sourced from 2014 to 2022, corresponding to problems denoted as CRS_2014-CRS_2019. The optimization objective and constraints for each problem are given in formulas (1-7). The reservoirs include Longyangxia, Liujiaxia, Haibowan, Wanjiazhai, Sanmenxia, ​​and Xiaolangdi. The scheduling cycle is one year, with a scheduling interval of ten days. Historical hydrological data were used.

[0140] The optimization method of this invention is compared with six existing algorithms: ARMOEA, NSGAII, CCMO, cDPEA, ICMA, and IMTCMO. ARMOEA adaptively adjusts the position of the reference vector to solve problems with irregular Pareto fronts; NSGAII, based on dominance relations and crowding distance, is the most commonly used multi-objective optimization algorithm; CCMO evolves two populations simultaneously, one considering constraints and the other not, and works together to improve the population's search capability through information sharing; cDPEA is also a two-population algorithm, but its two populations use adaptive penalty functions and feasibility-oriented methods to handle constraints respectively, and cooperate to produce offspring to share information. ICMA proposes an index-based constraint handling technique, which divides the defined potential region into multiple sub-regions and assigns different priorities to balance diversity, convergence, and feasibility. IMTCMO uses feasibility-oriented methods and an improved epsilon method to drive the evolution of the two populations respectively, and designs global and local search strategies to further improve the algorithm's search capability.

[0141] The parameters were set as follows: population size NP was set to 100, and maximum computational resource MaxFE was set to 1,000,000. All algorithms were run independently 30 times, and statistical results were obtained. The algorithm was evaluated based on the Inverse Generation Distance (IGD) metric, which calculates the average distance between the Pareto optimal solution and the algorithm's output population to reflect the algorithm's convergence and diversity. A smaller IGD value indicates a better algorithm performance. NaN indicates that the corresponding algorithm failed to find any feasible solution in 30 runs. It should be noted that because the Pareto optimal solution to the CRS problem is unknown, feasible non-dominated solutions in all algorithm output populations are considered Pareto optimal solutions.

[0142] The results in Tables 1 and 2 show that the method of this invention significantly outperforms the six comparative algorithms on 8, 9, 8, 7, 8, and 9 CRS problems, respectively, and achieves the best average IGD value on all 9 CRS problems. cDPEA and ICMA failed to find any feasible solutions on CRS_2015. Furthermore, some methods have a standard deviation of 0, indicating that the method only finds a feasible solution in one run.

[0143] The numbers in parentheses represent the standard deviation: The numbers in parentheses (e.g., 6.10e-02) are the standard deviations of the IGD values ​​obtained from multiple runs of the algorithm. The standard deviation reflects the stability of the algorithm. The smaller the standard deviation, the less the results fluctuate with each run, indicating greater stability and reliability. + / - / = are symbols for statistical significance tests.

[0144] "+" indicates that the comparison algorithm (such as ARMOEA) is significantly better than the method of the present invention, "-" indicates that the comparison algorithm is significantly worse than the method of the present invention, and "=" indicates that the comparison algorithm and the method of the present invention have comparable performance and the difference is not statistically significant.

[0145] Taking the first row of data in Table 1 as an example, for the CRS_2014 problem, the IGD value of ARMOEA is 0.32556, and the standard deviation is 0.061. The minus sign indicates that the performance of ARMOEAI is significantly worse than that of the method of this invention. The IGD value of NSGAII is 0.23874, and the standard deviation is 0.0155; the minus sign indicates that the performance of NSGAII is significantly worse than that of the method of this invention.

[0146] The IGD value of CCMO is 0.26572, with a standard deviation of 0.0542; the minus sign indicates that the performance of CCMO is significantly worse than that of the method of this invention. The IGD value of the method of this invention is 0.12627, with a standard deviation of 0.0247.

[0147] As can be seen from the IGD values ​​of all problems in Tables 1 and 2, the IGD value of the method of this invention is much smaller than that of the other three comparison algorithms, and the standard deviation is smaller. This shows that the algorithm of this invention has a significant advantage in solving the CRS problem, and the solution set found is closest to the true optimal solution.

[0148] In practical applications, users can fully combine these solutions to plan a satisfactory reservoir scheduling scheme according to their own circumstances.

[0149] Table 1. Comparison of IGD index between the present invention and three other evolutionary algorithms on the CRS problem.

[0150] question ARMOEA NSGAII CCMO Method of the present invention CRS_2014 3.2556e-01(6.10e-02)- 2.3874e-01(1.55e-02)- 2.6572e-01(5.42e-02)- 1.2627e-01(2.47e-02) CRS_2015 3.7708e-01(6.69e-02)- 2.9456e-01(00)- 3.0647e-01(2.80e-02)- 2.1437e-01(6.10e-02) CRS_2016 2.2140e-01(3.35e-02)- 1.6225e-01(00)- 1.8918e-01(1.71e-02)- 1.1407e-01(1.70e-02) CRS_2017 3.4682e-01(1.06e-02)= 2.5940e-01(2.89e-02)- 2.7973e-01(1.02e-02)= 1.7694e-01(1.06e-01) CRS_2018 3.8651e-01(7.59e-02)- 3.9081e-01(6.00e-02)- 3.7459e-01(1.06e-02)- 6.9756e-02(1.02e-02) CRS_2019 3.0913e-01(9.05e-02)- 2.0986e-01(1.33e-01)- 2.1194e-01(3.56e-02)- 5.5466e-02(4.00e-03) CRS_2020 2.3472e-01(3.42e-02)- 1.7835e-01(3.48e-02)- 2.1145e-01(5.26e-02)- 5.4996e-02(7.02e-03) CRS_2021 2.7993e-01(9.32e-02)- 2.0943e-01(7.85e-02)- 1.9601e-01(3.75e-02)- 8.9795e-02(7.84e-03) CRS_2022 2.7477e-01(3.02e-02)- 2.3612e-01(3.98e-02)- 2.6205e-01(3.12e-02)- 1.0978e-01(2.78e-02) + / - / ＝ 0 / 8 / 1 0 / 9 / 0 0 / 8 / 1

[0151] Table 2. Comparison of IGD index between the present invention and three other evolutionary algorithms on the CRS problem.

[0152] question cDPEA ICMA IMTCMO Method of the present invention CRS_2014 2.3304e-01(3.02e-02)- 6.0756e-01(00)- 1.8980e-01(6.99e-02)- 1.2627e-01(2.47e-02) CRS_2015 NaN- NaN- 2.6954e-01(2.31e-02)- 2.1437e-01(6.10e-02) CRS_2016 1.8044e-01(1.35e-02)- 5.3417e-01(4.54e-02)- 1.4922e-01(1.40e-02)- 1.1407e-01(1.70e-02) CRS_2017 2.6360e-01(1.83e-02)= 5.0610e-01(00)- 2.7853e-01(1.50e-02)- 1.7694e-01(1.06e-01) CRS_2018 4.4532e-01(00)- 1.9084e-01(5.49e-02)- 1.4508e-01(3.75e-02)- 6.9756e-02(1.02e-02) CRS_2019 2.1697e-01(3.81e-02)- 1.4943e-01(6.19e-02)- 1.2483e-01(1.82e-02)- 5.5466e-02(4.00e-03) CRS_2020 2.0547e-01(3.90e-02)- 2.3776e-01(1.68e-01)- 1.0864e-01(2.02e-02)- 5.4996e-02(7.02e-03) CRS_2021 2.0708e-01(3.69e-02)- 5.1251e-01(7.90e-02)- 1.1915e-01(1.29e-02)- 8.9795e-02(7.84e-03) CRS_2022 2.7151e-01(4.30e-02)- 5.5816e-01(3.70e-02)- 1.7905e-01(1.91e-02)- 1.0978e-01(2.78e-02) + / - / ＝ 0 / 7 / 2 0 / 8 / 1 0 / 9 / 0

[0153] As can be seen from Tables 1 and 2 above:

[0154] 1. Superior performance: The IGD value of the method of this invention is the lowest on the vast majority of test problems, which indicates that its overall performance (convergence and distribution of solution set) is good.

[0155] 2. Stability: By comparing the standard deviation, it can be seen that the method of the present invention has good stability in most cases.

[0156] The only exception: On the CRS_2017 problem, the performance of ARMOEA, CCMO, and cDPEA was statistically considered "comparable" to the method of this invention (but note that the IGD value of the method of this invention, 0.17694, is still much smaller than that of ARMOEA (0.34682), CCMO (0.27973), and cDPEA (0.2636). The reason for being judged as "comparable" may be because the standard deviation (1.06e-01) of the method of this invention on this problem is relatively large, that is, the results are more volatile, resulting in statistically insignificant differences).

[0157] The + / - / = symbols in the table represent the following: For example, for ARMOEA, its + / - / = values ​​are 0 / 8 / 1, indicating that ARMOEA is significantly worse than the method of this invention in 8 issues, and has no significant difference from this invention in 1 issue. Which specific issue it is depends on whether the sign after the specific IGD value of ARMOEA is + or -.

Claims

1. A multi-group adaptive cascade reservoir scheduling method, characterized in that, Includes the following steps: S1, Establish a CRS model based on the reservoir; S2 takes maximizing power generation, maximizing sediment discharge, and maximizing ecological efficiency as the optimization objectives of the CRS model; S3, Determine the constraints based on the optimization objective of CRS and the operational requirements of the cascade reservoir system; S4 utilizes a multi-group adaptive cascade reservoir scheduling method to perform operations on the optimization objective.

2. The multi-group adaptive cascade reservoir scheduling method according to claim 1, characterized in that: In step S1, The reservoirs in the CRS model originate from the Yellow River, and from upstream to downstream they are Longyangxia, Liujiaxia, Haibowan, Wanjiazhai, Sanmenxia, ​​and Xiaolangdi.

3. The multi-group adaptive cascade reservoir scheduling method according to claim 1, characterized in that: In step S2, the formula for calculating the maximization target of power generation is: In the formula, N1 and T represent the number of reservoirs and the total number of scheduling periods related to f1, respectively, and k i,t Q is the output coefficient. i,t Let m be the power generation flow of the i-th reservoir in time period t (m 3 / s), H represents the hydroelectric head (m), Δσ t Let t be the duration (s) of the t-th time interval; The formula for calculating the maximum sediment discharge target is as follows: In the formula, N2 represents the number of relevant reservoirs. Indicates the amount of sand transported (kg); The formula for calculating the target of maximizing ecological satisfaction rate is as follows: In the formula, N3 represents the number of relevant reservoirs, and W i,t,E It is a suitable ecological flow. This is the actual outbound flow rate, with the unit being meters (m). 3 / s.

4. The multi-group adaptive cascade reservoir scheduling method according to claim 1 or 2, characterized in that: In step S3, the constraints include water balance constraints, water level constraints, flow constraints, and water level constraints at the end of the scheduling period. The calculation formula for the water balance constraint is as follows: In the formula, q i,t This represents the actual water supply volume, in cubic meters (m³). 3 / s; The water level constraint satisfies the following conditions: In the formula, Z i,t This represents the water level of the i-th reservoir at the end of the t-th scheduling cycle. These represent the corresponding minimum and maximum water level limits, with the water level unit being meters (m). The flow constraint satisfies the following conditions: In the formula, Q i,t This represents the input or output flow of the i-th reservoir in the t-th scheduling cycle. and These represent the corresponding minimum and maximum input or output flow limits, respectively, in meters (m). 3 / s; The water level constraint at the end of the scheduling period must meet the following conditions: In the formula: V i,0 and V i,T V represents the water storage capacity of the i-th reservoir at the start and end of the scheduling process, respectively. i bejin and V i end This represents the actual water storage capacity, with units of 10. 8 m 3 .

5. The multi-group adaptive cascade reservoir scheduling method according to claim 1, characterized in that: In step S4, the optimization process includes the following steps: S401, initialize parameter u = 0, parameter k = 2, population size NP = 100; S402, randomly initialize population P3 for optimizing three objectives and all constraints; then randomly initialize population P2 for optimizing two objectives and all constraints, including... Then randomly initialize a population P1 to optimize an objective and all constraints, including P1 1 P1 2 P1 3 Initialize a save file A and copy P3 to A; S403 records the computing resources used: FE = 7 × NP; S404, determine whether FE is greater than MaxFE×u÷k. ​​If the condition is met, jump to S405; otherwise, jump to S407. S405, executes the population adaptive activation mechanism; S406, set u = u + 1, and set A = {P1, P2, P3}; S407, P3, Using the differential evolution algorithm, NP / 2 offspring individuals are generated respectively; S408, for P1 1 P1 2 P1 3 The activated population uses a differential evolution algorithm to produce NP / 2 offspring individuals. S409, Update Record A = A∪{O1,O2,O3}; S410 executes an environment selection mechanism based on two-way information sharing; S411: Determine if FE is greater than the set maximum computing resource MaxFE. If the condition is met, end and output P3; otherwise, jump to S404.

6. The multi-group adaptive cascade reservoir scheduling method according to claim 5, characterized in that: The specific steps of step S405 include: S4051, extract the constraint value of an individual in A, denoted as G; S4052, set L = {0,0,0}, i = 1; S4053, extract the i-th target value from individual A, denoted as Q; S4054, using the Spearman rank correlation coefficient method, calculate the correlation coefficient between Q and G: S4055, if ρ is less than 0, then set L. i =1; S4056, set i = i + 1. If i is greater than 3, end and output L; otherwise, jump to S4053.

7. The multi-group adaptive cascade reservoir scheduling method according to claim 4 or 5, characterized in that: The specific steps of S410 include: S4101, Settings S4102， S4103， S4104, P1 1 Use the epsilon method from the collection Select NP individuals to form a new P1 1 The second and third target values ​​are set to 0; S4105, P1 2 Use the epsilon method from the collection Select NP individuals to form a new P1 2 The first and third target values ​​are set to 0. S4106, P1 3 Use the epsilon method from the collection Select NP individuals to form a new P1 3 The first and second target values ​​are set to 0; S4107, Use the epsilon method from the collection Select NP individuals to form a new group. The third target value is set to 0; S4108, Use the epsilon method from the collection Select NP individuals to form a new group. The second target value is set to 0; S4109, Use the epsilon method from the collection Select NP individuals to form a new group. The first target value is set to 0; S41010, P3 uses the epsilon method to select NP individuals from the set {P3, C3} to form a new P3.

Images