A variable scale method and system for cascade reservoir group power generation dispatching

Abstract

The application discloses a cascade reservoir group power generation scheduling variable scale method and system, comprising the following steps: firstly, determining the reservoirs participating in calculation and their decision variables, determining the calculation sequence according to the hydraulic connection, and then generating an initial population in the initial scheduling trajectory neighborhood by adopting a random search strategy; secondly, updating the positions of all individuals by adopting three different scale search strategies, respectively, and comparing the fitness values of the individuals before and after the update, and reserving the individuals with better fitness; finally, repeatedly executing the above process until the termination condition is met, and taking the optimal individual in the current population as the optimal scheduling scheme. The application has the advantages of clear principle, easy execution, fast convergence speed, high search precision and the like, can significantly improve the cascade operation benefit, and provides a practical and innovative technical means for cascade reservoir group power generation scheduling.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of reservoir regulation, and particularly relates to a variable scale method and system for power generation regulation of cascade reservoir groups. BACKGROUND

[0002] Optimal regulation of a cascade reservoir group in a river basin is conducive to more reasonable allocation and more efficient utilization of water energy resources, and plays an extremely important role in flood control, drought prevention, ecology and benefit promotion. The optimal regulation of a cascade reservoir group has the characteristics of high dimension, nonlinearity, multi-stage, strong coupling of constraints, and complex hydraulic and electric power connections between upstream and downstream, and is very difficult to optimize and solve. Methods such as dynamic programming algorithm require discrete state variables, occupy a lot of memory, have large computational workload, consume long time, and are prone to cause "dimension disaster"; swarm intelligence algorithms such as particle swarm algorithm, genetic algorithm and differential evolution algorithm are prone to local convergence, and usually cannot provide stable and reliable regulation results. SUMMARY

[0003] The application aims to provide a variable scale method and system for power generation regulation of a cascade reservoir group to solve the problems in the prior art.

[0004] To achieve the above-mentioned purpose, the application provides the following technical scheme: a variable scale method for power generation regulation of a cascade reservoir group, comprising the following steps:

[0005] S1. Based on each target reservoir in the reservoir group, determining a calculation order according to the hydraulic connection of the target reservoir, and taking the reservoir capacity of each target reservoir as a decision variable;

[0006] S2. Based on the decision variable, generating an initial population consisting of a plurality of regulation schemes in the initial regulation trajectory neighborhood of each target reservoir in the reservoir group by using a random search strategy; each regulation scheme in the population corresponds to an individual;

[0007] S3. Correcting the decision value of each individual in the population, and then calculating the fitness value of all individuals in turn by using a dynamic penalty coefficient;

[0008] S4. Iteratively updating the position of each individual in the population in turn by using large-scale, medium-scale and small-scale search strategies to obtain the corresponding excellent individual under the scale search strategy; wherein the process of each scale search strategy is as follows: updating the position of the excellent individual obtained by the previous scale search strategy, then calculating the fitness value of the updated individual under the search strategy by executing step S3, comparing the fitness values of the individual before and after updating under the search strategy, and determining the individual with the higher fitness value as the excellent individual under the scale search strategy;

[0009] S5, judging whether the number of iterations of the iteration updating in S4 reaches a preset maximum iteration number, if yes, taking the excellent individual of the small-scale search strategy corresponding to the preset maximum iteration number as the best scheduling scheme.

[0010] Further, the aforementioned step S2 specifically comprises: initializing a counter k=1, and an individual m in the kth iteration is a scheduling scheme including I hydropower stations and J time periods, as follows:

[0011] Z(k,m)=[Z(k,m,1,1),…,Z(k,m,1,T),…,Z(k,m,i,j),…,Z(k,m,N,1),…,Z(k,m,N,T)],

[0012]

[0013] wherein Z(k,m,i,j), r(k,m,i,j) respectively represent a decision value of the reservoir i in the time period j and a corresponding random number in the individual m in the kth iteration; K represents the number of cycles; M represents the number of individuals; I represents the number of reservoirs; J represents the number of time periods; represents an initial decision of the reservoir i in the time period j.

[0014] Further, the aforementioned step S3 specifically comprises the following sub-steps:

[0015] S3.1, correcting the decision value of each individual to a feasible range as follows:

[0016]

[0017] wherein, respectively represent the maximum value and the minimum value of the decision value of the reservoir i in the time period j;

[0018] S3.2, calculating the fitness value of all individuals in turn by using a dynamic penalty coefficient method as follows:

[0019]

[0020]

[0021] wherein P(k,m,i,j) represents the output value of the reservoir i in the time period j in the individual m in the kth iteration; w(i,j) represents the electricity price coefficient of the reservoir i in the time period j; t(j) represents the number of hours of the time period j; Δ(k,m,l) represents a violation term of the lth constraint in the individual m in the kth iteration; L represents the number of constraints; a(k,m,l) represents the penalty coefficient of the lth constraint in the individual m in the kth iteration; a(k-1,m,l) represents the penalty coefficient of the lth constraint in the individual m in the k-1th iteration; and b represents an adjustment coefficient.

[0022] Further, the aforementioned step S4 is specifically as follows:

[0023] S401, update the position of each individual in the population by using a large-scale search strategy, then execute step S3 to calculate the updated individual fitness value, compare the individual fitness value before and after the large-scale search strategy is updated, determine the individual with higher fitness value as the excellent individual of the large-scale search strategy, and enter step S402;

[0024] S402, update the position of the excellent individual of the large-scale search strategy by using a mesoscale search strategy, then execute step S3 to calculate the updated individual fitness value, compare the individual fitness value before and after the mesoscale search strategy is updated, determine the individual with higher fitness value as the excellent individual of the mesoscale search strategy, and enter step S403;

[0025] S403, update the position of the excellent individual of the mesoscale search strategy by using a small-scale search strategy, then execute step S3 to calculate the updated individual fitness value, compare the individual fitness value before and after the small-scale search strategy is updated, and determine the individual with higher fitness value as the excellent individual of the small-scale search strategy;

[0026] Further, in the aforementioned step S401, the position of each individual in the population is updated by using a large-scale search strategy according to the following formula:

[0027]

[0028] Z(k,m)=argmax{F[Z(k,m)],F[X(k,m)]},

[0029] wherein X(k,m) represents the large-scale updated position of the individual m at the kth iteration; F[X(k,m)] represents the fitness of X(k,m); C1 and C2 represent random numbers in the interval [0, 1] obeying uniform distribution; C3 represents a random number in the interval [0, 2] obeying uniform distribution; and p1 represents a set large-scale search threshold.

[0030] Further, in the aforementioned step S402, the position of the excellent individual of the large-scale search strategy is updated by using a mesoscale search strategy according to the following formula:

[0031]

[0032] Z(k,m)=argmax{F[Z(k,m)],F[Y(k,m)]},

[0033] Wherein, Y(k,m) represents the mesoscale update position of individual m at the kth iteration; Z(k,n) represents the position of individual n at the kth iteration, and n≠m; F[Y(k,m)] represents the fitness of Y(k,m); C4 represents a random number in the interval [0, π] obeying uniform distribution.

[0034] Further, in the aforementioned step S403, the position of the excellent individual of the mesoscale search strategy is updated by using a small-scale search strategy according to the following formula:

[0035]

[0036] Z(k,m)=argmax{F[Z(k,m)],F[V(k,m)]},

[0037] Wherein, represents an adjustment coefficient; C5 represents a random number in the interval [0, 1] obeying uniform distribution; ρ2 represents a small-scale search threshold value set; Z max , Z min respectively represent the maximum value and the minimum value of the decision value of the cascade reservoir group.

[0038] Another aspect of the present application provides a variable scale system for cascade reservoir group power generation scheduling, comprising:

[0039] The decision variable acquisition module is configured to perform the following actions: based on each target reservoir in the reservoir group, determining a calculation order according to the hydraulic connection of the target reservoir, and taking the reservoir capacity of each target reservoir as a decision variable; the initialization module is configured to perform the following actions: based on the decision variable, generating an initial population consisting of a plurality of scheduling schemes within the initial scheduling trajectory neighborhood of each target reservoir in the reservoir group by using a random search strategy; each scheduling scheme in the population corresponds to an individual;

[0040] The individual fitness acquisition module is configured to perform the following actions: correcting the decision value of each individual in the population, and then calculating the fitness value of all individuals in turn by using a dynamic penalty coefficient;

[0041] The scale search module is configured to perform the following actions: iteratively updating the position of each individual in the population by using large-scale, mesoscale and small-scale search strategies in turn to obtain the corresponding excellent individual under the scale search strategy; wherein the process of each scale search strategy is as follows: updating the position of the excellent individual obtained by the previous scale search strategy, and then returning to the individual fitness module to calculate and obtain the individual fitness value after updating under the search strategy, comparing the individual fitness values before and after updating under the search strategy, and determining the individual with the higher fitness value as the excellent individual under the scale search strategy;

[0042] The best scheduling scheme decision module is configured to perform the following actions: judging whether the number of iteration updates in the scale search module reaches a preset maximum iteration number, and if yes, taking the excellent individual of the small scale search strategy corresponding to the preset maximum iteration number as the best scheduling scheme.

[0043] Compared with the prior art, the cascade operation benefit can be significantly improved, and a practical and innovative technical means is provided for cascade reservoir group power generation scheduling. BRIEF DESCRIPTION OF DRAWINGS

[0044] Figure 1 The flowchart of the application.

[0045] Figure 2 The convergence result comparison chart of the application and the classical method.

[0046] Figure 3 The cascade reservoir group scheduling scheme obtained by the application. DETAILED DESCRIPTION

[0047] In order to better understand the technical content of the application, specific embodiments are described below with reference to the accompanying drawings.

[0048] Aspects of the application are described in this detailed description and illustrated in the accompanying drawings by a number of illustrative embodiments. The embodiments of the application are not limited to the drawings described. It should be understood that the application is realized by any one of the above-mentioned various concepts and embodiments, and the concepts and embodiments described in detail below, since the disclosed concepts and embodiments of the application are not limited to any embodiment. In addition, some aspects of the application can be used alone or in any suitable combination with other aspects of the application.

[0049] As Figure 1 shown, the cascade reservoir group power generation scheduling variable scale method proposed by the application has the following process: S1, based on each target reservoir in the reservoir group, determining the calculation order according to the hydraulic connection of the target reservoir, and taking the reservoir capacity of each target reservoir as a decision variable.

[0050] S2, based on the decision variable, using a random search strategy to generate an initial population consisting of a plurality of scheduling schemes in the initial scheduling trajectory neighborhood of each target reservoir in the reservoir group; each scheduling scheme in the population corresponds to an individual. Specifically, initialize the counter k=1, and the individual m is a scheduling scheme including I hydropower stations and J periods at the kth iteration, as follows:

[0051] Z(k,m) = [Z(k,m,1,1),..., Z(k,m,1,T),..., Z(k,m,i,j),..., Z(k,m,N,1),..., Z(k,m,N,T)],

[0052]

[0053] wherein Z(k,m,i,j), r(k,m,i,j) represent the decision value and the corresponding random number of reservoir i in period j in individual m in the kth iteration, respectively; K represents the number of cycles; M represents the number of individuals; I represents the number of reservoirs; J represents the number of periods; represents the initial decision of reservoir i in period j.

[0054] S3, first correct the decision value of each individual in the population, and correct the decision value of each individual to the feasible range as follows:

[0055]

[0056] wherein, represent the maximum value and the minimum value of the decision value of reservoir i in period j, respectively.

[0057] Then the fitness value of all individuals is calculated in turn by using the dynamic penalty coefficient method as follows:

[0058]

[0059]

[0060] wherein P(k,m,i,j) represents the output value of reservoir i in period j in individual m in the kth iteration; w(i,j) represents the price coefficient of reservoir i in period j; t(j) represents the number of hours in period j; Δ(k,m,l) represents the violation term of the lth constraint in individual m in the kth iteration; L represents the number of constraints; a(k,m,l) represents the penalty coefficient of the lth constraint in individual m in the kth iteration; a(k-1,m,l) represents the penalty coefficient of the lth constraint in individual m in the (k-1)th iteration; b represents the adjustment coefficient.

[0061] S4, the positions of each individual in the population are updated in turn by using large-scale, medium-scale and small-scale search strategies to obtain the corresponding excellent individuals under the scale search strategy; wherein the process of each scale search strategy is: the position of the excellent individual obtained by the previous scale search strategy is updated, and then step S3 is executed to calculate the fitness value of the updated individual under the search strategy, the fitness value of the individual before and after updating under the search strategy is compared, and the individual with higher fitness value is determined as the excellent individual under the scale search strategy, specifically:

[0062] Firstly, the position of each individual in the population is updated by using the large-scale search strategy according to the following formula:

[0063]

[0064] Z(k,m) = argmax{F[Z(k,m)], F[X(k,m)]},

[0065] wherein X(k,m) represents the large-scale updated position of the individual m at the kth iteration; F[X(k,m)] represents the fitness of X(k,m); C1 and C2 represent random numbers in the interval [0, 1] obeying uniform distribution; C3 represents a random number in the interval [0, 2] obeying uniform distribution; and p1 represents a set large-scale search threshold.

[0066] Secondly, the updated individual fitness value is calculated in step S3, the individual fitness value before the large-scale search strategy is updated is compared with the updated individual fitness value, and the individual with the higher fitness value is determined as the excellent individual of the large-scale search strategy,

[0067] Then, the position of the excellent individual of the large-scale search strategy is updated by using the mesoscale search strategy according to the following formula:

[0068]

[0069] Z(k,m) = argmax{F[Z(k,m)], F[Y(k,m)]},

[0070] wherein Y(k,m) represents the mesoscale updated position of the individual m at the kth iteration; Z(k,n) represents the position of the individual n at the kth iteration, and n≠m; F[Y(k,m)] represents the fitness of Y(k,m); and C4 represents a random number in the interval [0, π] obeying uniform distribution.

[0071] Then, the updated individual fitness value is calculated in step S3, the individual fitness value before the mesoscale search strategy is updated is compared with the updated individual fitness value, and the individual with the higher fitness value is determined as the excellent individual of the mesoscale search strategy.

[0072] Then, the position of the excellent individual of the mesoscale search strategy is updated by using the small-scale search strategy according to the following formula:

[0073]

[0074] Z(k,m) = argmax{F[Z(k,m)], F[V(k,m)]},

[0075] wherein V(k,m) represents the small-scale updated position of the individual m at the kth iteration; Z(k,n) represents the position of the individual n at the kth iteration, and n≠m; F[V(k,m)] represents the fitness of V(k,m); and C5 represents a random number in the interval [0, 2π] obeying uniform distribution. represents an adjustment coefficient; C5 represents a random number in the interval [0,1] obeying a uniform distribution; ρ2 represents a set small-scale search threshold value; Z max , Z min respectively represent the maximum value and the minimum value of the decision value of the cascade reservoir group.

[0076] Finally, step S3 is executed to calculate the updated individual fitness value, the individual fitness value before the small-scale search strategy is updated is compared with the updated individual fitness value, and the individual with the higher fitness value is determined as the excellent individual of the small-scale search strategy;

[0077] S5, whether the number of iterations of the iteration update in S4 reaches the preset maximum iteration number is judged, and if yes, the excellent individual of the small-scale search strategy corresponding to the preset maximum iteration number is taken as the best scheduling scheme.

[0078] To show the effect achieved by the present application, the present application is selected to study a certain basin cascade reservoir group. Figure 2 As shown in the figure, compared with the convergence results of two kinds of classical methods, differential evolution algorithm and particle swarm algorithm, it can be seen that the convergence speed of the present application is obviously better than that of the classical methods, and the optimal solution can be converged in about 200 generations, while the results obtained by the differential evolution algorithm and the particle swarm algorithm at the same period are obviously worse than those of the present application. Table 1 is the statistical index of the present application and the classical methods, and it can be seen that the average value, variance and other indexes of the present application are obviously better than those of the other two methods, and the worst value of the present application is better than the optimal value of the other two methods. Therefore, the present application has good convergence speed, search performance and robustness, and can quickly provide a stable and effective cascade reservoir group power generation scheduling scheme.

[0079] Table 1

[0080]

[0081] Figure 3 The figure of the cascade reservoir group scheduling scheme obtained by the present application is shown. It can be seen that the water power station 1 is the leading water power station, and the water level is fully stored before the flood season and is raised during the flood season, and compensation adjustment is carried out during the dry season; the other power stations (water power stations 2-6) try to keep high water head operation and increase power generation; the water level and output of each water power station meet the corresponding constraints at the end of the scheduling period. Therefore, the present application provides a practical and effective method for cascade reservoir group power generation scheduling.

[0082] The cascade reservoir group power generation scheduling variable scale system provided by the present application comprises:

[0083] The decision variable acquisition module is configured to perform the following actions: based on each target reservoir in the reservoir group, determining a calculation sequence according to the hydraulic connection of the target reservoir, and taking the reservoir capacity of each target reservoir as a decision variable; the initialization module is configured to perform the following actions: based on the decision variable, generating an initial population consisting of a plurality of scheduling schemes in the initial scheduling trajectory neighborhood of each target reservoir in the reservoir group by using a random search strategy; each scheduling scheme in the population corresponds to an individual;

[0084] The individual fitness acquisition module is configured to perform the following actions: correcting the decision value of each individual in the population, and then calculating the fitness value of all individuals in turn by using a dynamic penalty coefficient;

[0085] The scale search module is configured to perform the following actions: iteratively updating the position of each individual in the population by using large-scale, medium-scale and small-scale search strategies in turn to obtain the corresponding excellent individual under the scale search strategy; wherein the process of each scale search strategy is: updating the position of the excellent individual obtained by the previous scale search strategy, then returning to the individual fitness module to calculate and obtain the individual fitness value after updating under the search strategy, comparing the individual fitness values before and after updating under the search strategy, and determining the individual with a higher fitness value as the excellent individual under the scale search strategy;

[0086] The best scheduling scheme decision module is configured to perform the following actions: judging whether the number of iterative updates in the scale search module reaches a preset maximum iteration number, and if so, taking the excellent individual under the small-scale search strategy corresponding to the preset maximum iteration number as the best scheduling scheme.

[0087] Although the present application has been described as above with reference to the preferred embodiments, it is not intended to limit the present application. Those skilled in the art, without departing from the spirit and scope of the present application, can make various modifications and decorations. Therefore, the protection scope of the present application shall be subject to the definition of the claims.

Claims

1. A variable scale method for power generation scheduling of a cascade reservoir group, characterized in that, The method comprises the following steps: S1, based on each target reservoir in the reservoir group, determining a calculation sequence according to the hydraulic connection of the target reservoir, and taking the reservoir capacity of each target reservoir as a decision variable; S2, based on the decision variable, generating an initial population consisting of a plurality of scheduling schemes in the initial scheduling trajectory neighborhood of each target reservoir in the reservoir group by using a random search strategy; each scheduling scheme in the population corresponds to an individual; S3, correcting the decision value of each individual in the population, and then calculating the fitness value of all individuals in turn by using a dynamic penalty coefficient; S4, iteratively updating the position of each individual in the population in turn by using large-scale, medium-scale and small-scale search strategies to obtain the corresponding excellent individual under the search strategy at the scale; wherein the process of each scale search strategy is as follows: updating the position of the excellent individual obtained by the previous scale search strategy, then executing step S3 to calculate the fitness value of the updated individual under the search strategy, comparing the fitness values of the individuals before and after updating under the search strategy, and determining the individual with the higher fitness value as the excellent individual under the search strategy at the scale; Specifically comprising the following sub-steps: S401, updating the position of each individual in the population by using a large-scale search strategy as follows: ， ， wherein, denotes the large-scale updated position of the individual at the th iteration; denotes the fitness of , denotes a random number uniformly distributed in the interval [0, 1]; denotes a random number uniformly distributed in the interval [0, 2]; denotes a set large-scale search threshold value; Then execute step S3 to calculate the updated individual fitness value, compare the individual fitness values before and after updating under the large-scale search strategy, determine the individual with the higher fitness value as the large-scale search strategy excellent individual, and enter step S402; S402, the position of the excellent individual of the large-scale search strategy is updated by using the mesoscale search strategy, as follows: , ， wherein, denotes the meso-scale update position of the individual at the i-th iteration; denotes the position of the individual at the i-th iteration, and has ; ; denotes the fitness of ; denotes a random number uniformly distributed in the interval ;​ Then execute step S3 to calculate the updated individual fitness value, compare the individual fitness values before and after updating under the medium-scale search strategy, determine the individual with the higher fitness value as the medium-scale search strategy excellent individual, and enter step S403; S403, update the position of the excellent individual of the middle-scale search strategy by using the small-scale search strategy; , ， wherein, denotes an adjustment coefficient; denotes a random number in the interval [0, 1] subject to a uniform distribution; denotes a set small-scale search threshold value; , denote the maximum value and the minimum value of the decision value of the cascade reservoir group, respectively. Then execute step S3 to calculate the updated individual fitness value, compare the individual fitness values before and after updating under the small-scale search strategy, determine the individual with the higher fitness value as the small-scale search strategy excellent individual; S5, judging whether the number of iterative updates in S4 reaches a preset maximum iteration number, and if yes, taking the small-scale search strategy excellent individual corresponding to the preset maximum iteration number as the best scheduling scheme.

2. The variable scale method for cascade reservoirs power generation dispatching according to claim 1, characterized in that, Step S2 specifically includes: initializing a counter , the first iteration , the individual is a scheduling scheme including hydroelectric power stations time periods, as follows: ， ， in, , They represent the first In the next iteration, the individual In the middle, reservoir During the period The decision value and the corresponding random number; Indicates the number of loop iterations; Indicates the number of individuals; Indicates the number of reservoirs; Indicates the number of time periods; Reservoir During the period The initial decision.

3. The variable scale method for cascade reservoirs generation scheduling according to claim 2, characterized in that, Step S3 specifically comprises the following sub-steps: S3.1, correcting the decision value of each individual to the feasible range as follows: ， wherein, , respectively represent a reservoir at a time period maximum and minimum values of the decision value; S3.2, calculating the fitness value of all individuals in turn by using a dynamic penalty coefficient method as follows: ， ， in, Indicates the first In the next iteration, the individual In the middle, reservoir During the period The output value; Reservoir During the period Electricity price coefficient; Indicates time period Hours; Indicates the first In the next iteration, the individual The Middle Term that violates a constraint; Indicates the number of constraints; Indicates the first In the next iteration, the individual The Middle Penalty coefficient for the constraint; Indicates the first In the next iteration, the individual The Middle Penalty coefficient for the constraint; This represents the adjustment factor.

4. A variable scale system for power generation dispatching of a cascade reservoir group, characterized in that, It comprises: A decision variable acquisition module configured to perform the following actions: based on each target reservoir in the reservoir group, determining a calculation sequence according to the hydraulic connection of the target reservoir, and taking the reservoir capacity of each target reservoir as a decision variable; An initialization module configured to perform the following actions: based on the decision variable, generating an initial population consisting of a plurality of scheduling schemes in the initial scheduling trajectory neighborhood of each target reservoir in the reservoir group by using a random search strategy; each scheduling scheme in the population corresponds to an individual; The individual fitness obtaining module is configured to perform the following actions: correcting the decision value of each individual in the population, and then calculating the fitness value of each individual in turn by using a dynamic penalty coefficient; The scale search module is configured to perform the following actions: iteratively updating the position of each individual in the population by using a large-scale, a medium-scale and a small-scale search strategy in turn to obtain the excellent individual corresponding to the search strategy; wherein the process of each scale search strategy is as follows: updating the position of the excellent individual obtained by the previous scale search strategy, and then returning to the individual fitness module to calculate the updated individual fitness value under the search strategy, comparing the individual fitness values before and after the update under the search strategy, and determining the individual with the higher fitness value as the excellent individual under the scale search strategy; and the specific execution is as follows: S401, updating the position of each individual in the population by using a large-scale search strategy, as follows: ， ， wherein, denotes the position of the individual at the large scale update of the th iteration; denotes the fitness of , denotes a random number from the interval [0, 1] uniformly distributed; denotes a random number from the interval [0, 2] uniformly distributed; denotes a set large scale search threshold value; Then, step S3 is performed to calculate the updated individual fitness value, the individual fitness value before and after the update under the large-scale search strategy is compared, the individual with the higher fitness value is determined as the excellent individual under the large-scale search strategy, and step S402 is entered; S402, the position of the excellent individual of the large-scale search strategy is updated by using the mesoscale search strategy, as follows: , ， wherein, denotes the meso-scale update position of the individual at the i-th iteration; denotes the position of the individual at the i-th iteration, and has ; denotes the fitness of ; denotes a random number uniformly distributed in the interval ;​​ Then, step S3 is performed to calculate the updated individual fitness value, the individual fitness value before and after the update under the medium-scale search strategy is compared, the individual with the higher fitness value is determined as the excellent individual under the medium-scale search strategy, and step S403 is entered; S403, update the position of the excellent individual of the middle-scale search strategy by using a small-scale search strategy; , ， wherein, denotes an adjustment coefficient; denotes a random number in the interval [0, 1] subject to a uniform distribution; denotes a set small-scale search threshold value; , denote a vector corresponding to the maximum value and the minimum value of the decision value of the cascade reservoir group, respectively; Then, step S3 is performed to calculate the updated individual fitness value, the individual fitness value before and after the update under the small-scale search strategy is compared, the individual with the higher fitness value is determined as the excellent individual under the small-scale search strategy; The best scheduling scheme decision module is configured to perform the following actions: judging whether the number of iterative updates in the scale search module reaches a preset maximum iteration number, and if yes, determining the excellent individual under the small-scale search strategy corresponding to the preset maximum iteration number as the best scheduling scheme.

Images