A reservoir group system joint dispatching layer optimization method and system

Abstract

The present invention discloses a layered optimization method and system for joint scheduling of a reservoir group system, comprising the following steps: Level 1 uses two parallel layers based on a weighted balance update strategy and an adaptive learning strategy to perform extensive search. Level 2 uses a nonlinear learning factor update strategy to perform balanced search. Level 3 uses a two-level joint search strategy: the upper level includes an overall range search strategy for improving global optimization capabilities, and a local range search strategy for achieving adaptive dynamic optimization; the lower level prevents falling into local optimality by determining the search area. The present invention accelerates the convergence speed and enhances the global optimization and local search capabilities through dynamic joint searches at different levels. It has the advantages of simple programming implementation, wide application scenarios, and stable results. It can quickly provide a reservoir group scheduling plan and provides an effective technology for solving reservoir group scheduling problems.

Description

Technical Field

[0001] The present invention belongs to the field of reservoir regulation, and relates to a reservoir group system joint regulation layer optimization method and system. Background Art

[0002] The optimal operation of a reservoir cluster is a typical large-scale, multi-dimensional, multi-stage, nonlinearly constrained optimization problem, requiring the use of modern computer technology and optimization techniques to find reasonable operation methods and work plans. However, due to the close relationship between reservoir operation and numerous complex factors, such as social, economic, natural, environmental, and engineering factors, as well as the strong coupling and complex constraints between the states and decision variables of cascades, the efficient solution of the joint operation of a reservoir cluster is extremely difficult and has been a hot topic, difficulty, and focus of research in the field of hydropower both domestically and internationally. Current methods for solving the joint operation of reservoir clusters mainly fall into two categories: 1) classical mathematical programming methods, represented by linear programming, nonlinear programming, and dynamic programming; and 2) heuristic algorithms, represented by particle swarm optimization (PSO), genetic algorithms, and ant colony optimization. Classical mathematical programming methods suffer from the curse of dimensionality and long computational time, while heuristic algorithms are prone to getting stuck in local optima and have poor global search capabilities. Summary of the Invention

[0003] The purpose of the present invention is to provide a hierarchical optimization method for joint scheduling of a reservoir group system, and to provide a new and effective technology for solving complex reservoir group scheduling problems. To achieve the above purpose, the present invention provides the following technical solutions: A hierarchical optimization method for joint scheduling of a reservoir group system, comprising the following steps: Step S1, based on reservoir group data, a reservoir group system joint scheduling model is constructed with the goal of maximizing power generation, and constraints are added to the model; at the same time, a maximum number of iterations and a population size are preset; and the water level at the end of the reservoir scheduling period is determined as an optimization variable;

[0004] Step S2: adopt the joint dispatching layer optimization method to solve the optimal joint dispatching scheme of the reservoir group system, specifically:

[0005] Step 201, constructing the initial layer: generating an initial population within a preset range of decision variables, obtaining the population's velocity variables, storing individuals in the population as individual optimal positions, calculating the fitness of individuals in the population, and updating the dominant individuals based on the fitness of the individuals;

[0006] Step 202: Construct the first level: use the adaptive learning strategy and weight balancing update strategy to obtain dynamic weight factors;

[0007] Step 203: construct the second level: obtain dynamic learning factors using a nonlinear learning factor update strategy;

[0008] Step 204: construct the third level: update the population using a two-level joint search strategy;

[0009] Step 205, construct the fourth level: calculate the fitness of individuals in the current population, update the optimal position of individuals in the current population, and the position of dominant individuals; and determine whether the preset maximum number of iterations has been reached. If so, obtain the optimal joint scheduling plan for the reservoir group system and output the fitness value; otherwise, return to step S202.

[0010] Furthermore, in the aforementioned step S1, the maximum power generation objective function of the reservoir group system joint scheduling model is:

[0011]

[0012] Where E is the total power generation of all reservoirs in the period; K is the number of reservoirs; J is the total number of periods; P k,j , Q k,j 、H k,j are the output, power generation flow and water head of the kth reservoir in the jth period respectively; A k is the output coefficient of the kth reservoir; t j is the number of hours in the jth period.

[0013] Furthermore, in the aforementioned step S1, the constraint conditions include:

[0014] Output constraints:

[0015]

[0016] in, are the minimum and maximum outputs of the kth reservoir in the jth period respectively; storage capacity constraint:

[0017]

[0018] in, are the minimum and maximum storage capacities of the kth reservoir in the jth period; water balance constraints

[0019]

[0020] Among them, I k,j , O k,j are the interval inflow and outflow of the kth reservoir in the jth period respectively; Remove lost water from outflows from upstream reservoirs;

[0021] Outbound flow constraints:

[0022]

[0023] in, are the minimum and maximum outflows of the kth reservoir in the jth period respectively; power generation flow constraint:

[0024]

[0025] in, are the minimum and maximum power generation flows of the kth reservoir in the jth period respectively; the initial and final storage capacity constraints are:

[0026]

[0027] Among them, V k beg 、V k end are the initial and final storage capacities of the kth reservoir, respectively.

[0028] Furthermore, the aforementioned step 201 includes the following sub-steps:

[0029] S2.1. Randomly generate an initial population x within the preset range of the decision variable. Set the upper and lower limits of the velocity population v as γ times the upper and lower limits of the decision variable of the initial population x, as follows:

[0030]

[0031]

[0032] Among them, x i,j and v i,j are the values ​​of the jth decision variable and speed variable of the i-th individual,

[0033] Indicates that in the interval Function to generate uniformly distributed random numbers;

[0034] S2.2. Calculate the fitness of individuals in the initial population and store them as the individual optimal position pbest. At the same time, determine whether there is a feasible solution. If so, store the individual with the smallest fitness in the pbest population as the dominant individual gbest. Otherwise, store the individual with the smallest constraint violation value in the infeasible solution as the dominant individual gbest.

[0035] Furthermore, the aforementioned step 202 specifically includes:

[0036] First, the first parallel layer is constructed: the weight W2 uses a weighted balanced update strategy to conduct a wide search in the entire solution space; then the second parallel layer is constructed: the weight W1 uses an adaptive learning strategy to bidirectionally link with the first parallel layer, while preventing it from falling into the local optimum during the balanced search, as shown in the following formula:

[0037]

[0038] in

[0039]

[0040]

[0041] Among them, W1 and W2 are the dynamic weight factors obtained by the adaptive learning strategy and the weight balance update strategy respectively, t and T are the current number of iterations and the maximum number of iterations of the algorithm respectively, and a1, a2, a3 and a4 are weight calculation parameters.

[0042] Furthermore, the aforementioned step 203 specifically includes: updating the learning factors r1 and r2 through the nonlinear distribution of the cosine function, as shown in the following formula:

[0043]

[0044] Among them, r1 and r2 are dynamic learning factors obtained by the nonlinear learning factor update strategy; rand is a random number uniformly distributed in the interval [0,1]; and b is a calculation parameter.

[0045] Furthermore, the aforementioned step S204 uses a two-stage joint search strategy to update the population, as shown in the following formula:

[0046] Superior local search strategy:

[0047] A=-abs((x i -gbest) / gbest) (35)

[0048]

[0049] Superior overall search strategy:

[0050] C=x i +(-1+2·round(rand))·x i ·((1-t / T) / 2·e (A) +(t / T)·e (B) ) (37)

[0051] Subordinate dynamic search strategy:

[0052] D=W·v+r1·(pbest-x)+r2·(gbest-x) (38)

[0053]

[0054] E=floor(N·c·(t / T))) (40)

[0055] Among them, A and B are factors corresponding to the upper-level local search strategy; C is the variable generated by the upper-level overall search strategy; D is the variable generated by the lower-level dynamic search strategy; abs is the absolute value function; floor is the floor function; round is the rounding function; N is the number of individuals in the population; rand is a random number uniformly distributed in the interval [0,1]; E is the number of individuals executing the upper-level overall search strategy; randi1, randi e ,randi E is a random integer function uniformly distributed in the interval [1, N]; i is the individual serial number; c is the proportional selection calculation parameter.

[0056] Furthermore, in the aforementioned step S205, the fitness of the individuals in the current population is calculated, and the optimal position of the individuals in the current population and the position of the dominant individual are updated as follows:

[0057]

[0058] gbest=pbest k , k=min(F(pbest)) (42)

[0059] Among them, F is the fitness value calculation function, and k is the serial number of the individual with the smallest fitness value in the optimal position.

[0060] Another aspect of the present invention provides a reservoir group system joint scheduling hierarchical optimization system, comprising: a model building module configured to perform the following actions: based on reservoir group data, construct a reservoir group system joint scheduling model with the goal of maximizing power generation, and add constraints to the model; simultaneously preset a maximum number of iterations and a population size; determine the water level at the end of the reservoir scheduling period as an optimization variable;

[0061] The optimal reservoir group system joint scheduling model solving module includes an initialization unit, a dynamic weight acquisition unit, a dynamic learning factor acquisition unit, a population update unit, and an optimal scheduling solution acquisition unit; the initialization unit is configured to perform the following actions: construct an initial layer: generate an initial population within a preset range of decision variables, obtain the population's speed variable, store individuals in the population as individual optimal positions, calculate the fitness of individuals in the population, and update dominant individuals based on the individual fitness;

[0062] The dynamic weight acquisition unit is configured to perform the following actions: construct the first level: obtain the dynamic weight factor using the adaptive learning strategy and the weight balancing update strategy;

[0063] The dynamic learning factor acquisition unit is configured to perform the following actions: construct the second level: obtain the dynamic learning factor using a nonlinear learning factor update strategy;

[0064] The population update unit,is configured to perform the following actions: construct the third level: update the population,using a two-level joint search strategy;

[0065] The optimal scheduling scheme acquisition unit is configured to perform the following actions: construct the fourth level: calculate the fitness of individuals in the current population, update the optimal position of individuals in the current population, and the position of dominant individuals; and determine whether the preset maximum number of iterations has been reached. If so, the optimal joint scheduling scheme for the reservoir group system is obtained and the fitness value is output; otherwise, it returns to the dynamic weight acquisition unit.

[0066] Compared with the existing methods, the present invention has the following beneficial effects:

[0067] 1. The present invention avoids the traditional complex parameter adjustment process, and has the advantages of fast convergence speed, strong global optimization capability, wide application scenarios, and simple programming implementation, providing an effective technology for solving the problem of optimal scheduling of reservoir groups.

[0068] 2. The present invention uses a weighted balance update strategy to traverse the entire solution space and adopts an adaptive learning strategy to enhance the algorithm's ability to escape local extreme values. The nonlinear learning factor update strategy takes into account both search efficiency and convergence speed.

[0069] 3. The present invention adopts a two-level joint search strategy: ① The overall range search in the upper level changes with the number of iterations, which improves the global optimization ability of the scheduling scheme; ② The local range search in the upper level determines the local search range by dynamically comparing the population position and the global extreme value. When the population individual approaches the global extreme value, the search range is increased, and vice versa, the search range is dynamically reduced, thus achieving a dynamic balance between global exploration and local exploration; ③ The lower level determines the dynamic search area to ensure the global convergence speed while preventing premature falling into the local optimum; ④ The individual selection probability is dynamically updated according to the number of iterations, realizing large-scale dynamic optimization of the scheduling scheme and adaptive updating of decision variables.

[0070] 4. The present invention can achieve global optimization of reservoir group scheduling problems, further improve computational efficiency and accuracy, and quickly provide a set of reservoir group scheduling solutions, thereby providing a more effective technology for solving complex reservoir group optimization scheduling engineering problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 Schematic diagram of the method of the present invention.

[0072] Figure 2 It is a flowchart of the algorithm of the present invention.

[0073] Figure 3 It is a schematic diagram of the comparison results of the maximum power generation model of a single reservoir according to the present invention.

[0074] Figure 4 It is a schematic diagram of the comparison results of the maximum power generation model of the present invention in the cascade reservoir.

[0075] Figure 5 The water level change diagram is drawn based on the optimal results of the two algorithms for optimizing the scheduling of cascade reservoirs. DETAILED DESCRIPTION

[0076] In order to better understand the technical content of the present invention, specific embodiments are given and described below with reference to the accompanying drawings.

[0077] Various aspects of the present invention are described herein with reference to the accompanying drawings, which show a number of illustrative embodiments. The embodiments of the present invention are not limited to those described in the accompanying drawings. It should be understood that the present invention can be implemented by any of the various concepts and embodiments described above, as well as the concepts and implementations described in detail below, because the concepts and embodiments disclosed herein are not limited to any particular implementation. In addition, some aspects disclosed herein may be used alone or in any appropriate combination with other aspects disclosed herein.

[0078] like Figure 1 The flowchart of the present invention is a reservoir group system joint scheduling layer optimization method, comprising the following steps:

[0079] Step S1: Based on the reservoir group data, a reservoir group system joint scheduling model is constructed with the goal of maximizing power generation, and constraints are added to the model; at the same time, the maximum number of iterations and the population size are preset; and the water level at the end of the reservoir scheduling period is determined as the optimization variable;

[0080] The maximum power generation objective function of the reservoir group system joint scheduling model is:

[0081]

[0082] Where E is the total power generation of all reservoirs in the period; K is the number of reservoirs; J is the total number of periods; P k,j , Q k,j 、H k,j are the output, power generation flow and water head of the kth reservoir in the jth period respectively; A k is the output coefficient of the kth reservoir; t j is the number of hours in the jth period.

[0083] The constraints include:

[0084] Output constraints:

[0085]

[0086] in, are the minimum and maximum outputs of the kth reservoir in the jth period respectively;

[0087] Storage capacity constraints:

[0088]

[0089] in, are the minimum and maximum storage capacities of the kth reservoir in the jth period respectively;

[0090] Water balance constraints

[0091]

[0092] Among them, I k,j , O k,j are the interval inflow and outflow of the kth reservoir in the jth period respectively; Remove lost water from outflows from upstream reservoirs;

[0093] Outbound flow constraints:

[0094]

[0095] in, are the minimum and maximum outflows of the kth reservoir in the jth period respectively; power generation flow constraint:

[0096]

[0097] in, are the minimum and maximum power generation flows of the kth reservoir in the jth period respectively; the initial and final storage capacity constraints are:

[0098]

[0099] Among them, V k beg 、V k end are the initial and final storage capacities of the kth reservoir, respectively.

[0100] Step S2 adopts a reservoir group system joint dispatching layer optimization method to solve the optimal reservoir dispatching scheme, specifically as follows: Figure 2 As shown, it includes steps S201 to S205:

[0101] Step 201, constructing the initial layer: generating an initial population within a preset range of decision variables, obtaining the population's velocity variables, storing individuals in the population as individual optimal positions, calculating the fitness of individuals in the population, and updating the dominant individuals based on the fitness of the individuals; specifically, steps S2.1 to S2.2 are as follows:

[0102] S2.1. Randomly generate an initial population x within the preset range of the decision variable. Set the upper and lower limits of the velocity population v as γ times the upper and lower limits of the decision variable of the initial population x, as follows:

[0103]

[0104]

[0105] Among them, x i,j and v i,j are the values ​​of the jth decision variable and speed variable of the i-th individual, Indicates that in the interval Function to generate uniformly distributed random numbers;

[0106] S2.2. Calculate the fitness of individuals in the initial population and store them as the individual optimal position pbest. At the same time, determine whether there is a feasible solution. If so, store the individual with the smallest fitness in the pbest population as the dominant individual gbest. Otherwise, store the individual with the smallest constraint violation value in the infeasible solution as the dominant individual gbest.

[0107] Step 202: Construct the first level: Use the adaptive learning strategy and weight balancing update strategy to obtain the dynamic weight factor. Specifically, first construct the first parallel layer: the weight W2 uses the weight balancing update strategy to conduct a wide search in the entire solution space; then construct the second parallel layer: the weight W1 uses the adaptive learning strategy to bidirectionally link with the first parallel layer, and prevents falling into the local optimum while performing a balanced search, as shown in the following formula:

[0108]

[0109] in

[0110]

[0111]

[0112] Among them, W1 and W2 are the dynamic weight factors obtained by the adaptive learning strategy and the weight balancing update strategy respectively, t and T are the current number of iterations and the maximum number of iterations of the algorithm respectively, a1, a2, a3 and a4 are weight calculation parameters; in the present invention, they are taken as 0.2, 0.7, 0.7 and 0.3 respectively.

[0113] S203. Construct the second level: Use the nonlinear learning factor update strategy to obtain the dynamic learning factor; the learning factors r1 and r2 are updated through the nonlinear distribution of the cosine function, as shown in the following formula:

[0114]

[0115] Among them, r1 and r2 are dynamic learning factors obtained by the nonlinear learning factor update strategy; rand is a random number uniformly distributed in the interval [0, 1]; b is a calculation parameter, and its value is 2 in the present invention.

[0116] S204. Construct the third level: Use a two-level joint search strategy to update the population: the upper level includes an overall range search strategy to improve global optimization capabilities and a local range search strategy to achieve adaptive dynamic optimization; the lower level prevents falling into local optimality by determining the search area, as follows:

[0117] Superior local search strategy:

[0118] A=-abs((x i - gbest ) / gbest ) (56)

[0119]

[0120] Superior overall search strategy:

[0121] C=x i +(-1+2·round(rand))·x i ·((1-t / T) / 2·e (A) +(t / T)·e (B) ) (58)

[0122] Subordinate dynamic search strategy:

[0123] D=W·v + r1·(pbest-x) + r2·(gbest-x) (59)

[0124]

[0125] E=floor(N·c·(t / T))) (61)

[0126] Among them, A and B are factors corresponding to the upper-level local search strategy; C is the variable generated by the upper-level overall search strategy; D is the variable generated by the lower-level dynamic search strategy; abs is the absolute value function; floor is the floor function; round is the rounding function; N is the number of individuals in the population; rand is a random number uniformly distributed in the interval [0,1]; E is the number of individuals executing the upper-level overall search strategy; randi1, randi e ,randi E is a random integer function uniformly distributed in the interval [1, N]; i is the individual serial number; c is the proportion selection calculation parameter, which is 0.5 in the present invention.

[0127] S205. Construct the fourth level: Update the optimal position of individuals and the position of dominant individuals: Calculate the fitness of the current population and update the optimal position of individuals and the position of dominant individuals, as follows:

[0128] Calculate the fitness of individuals in the current population, and update the optimal position of individuals in the current population, as well as the position of dominant individuals, as follows:

[0129]

[0130] gbest=pbest k , k=min(F(pbest)) (63)

[0131] Among them, F is the fitness value calculation function, and k is the serial number of the individual with the smallest fitness value in the optimal position.

[0132] Then, it is determined whether the preset maximum number of iterations has been reached. If so, the iteration is terminated, and the optimal joint scheduling plan for the reservoir group system is obtained, and the fitness value of the current iteration is output. Otherwise, it returns to step S202.

[0133] Another aspect of the present invention provides a reservoir group system joint scheduling hierarchical optimization system, comprising: a model building module configured to perform the following actions: based on reservoir group data, construct a reservoir group system joint scheduling model with the goal of maximizing power generation, and add constraints to the model; simultaneously preset a maximum number of iterations and a population size; determine the water level at the end of the reservoir scheduling period as an optimization variable;

[0134] The optimal reservoir group system joint scheduling model solving module includes an initialization unit, a dynamic weight acquisition unit, a dynamic learning factor acquisition unit, a population update unit, and an optimal scheduling solution acquisition unit; the initialization unit is configured to perform the following actions: construct an initial layer: generate an initial population within a preset range of decision variables, obtain the population's speed variable, store individuals in the population as individual optimal positions, calculate the fitness of individuals in the population, and update dominant individuals based on the individual fitness;

[0135] The dynamic weight acquisition unit is configured to perform the following actions: construct the first level: obtain the dynamic weight factor using the adaptive learning strategy and the weight balancing update strategy;

[0136] The dynamic learning factor acquisition unit is configured to perform the following actions: construct the second level: obtain the dynamic learning factor using a nonlinear learning factor update strategy;

[0137] The population update unit,is configured to perform the following actions: construct the third level: update the population,using a two-level joint search strategy;

[0138] The optimal scheduling scheme acquisition unit is configured to perform the following actions: construct the fourth level: calculate the fitness of individuals in the current population, update the optimal position of individuals in the current population, and the position of dominant individuals; and determine whether the preset maximum number of iterations has been reached. If so, the optimal joint scheduling scheme for the reservoir group system is obtained and the fitness value is output; otherwise, it returns to the dynamic weight acquisition unit.

[0139] The above scheduling model is solved by the present invention and the classical particle swarm optimization (PSO). In order to fully demonstrate the robustness of the present invention, 15 independent calculations are performed on the scheduling problems of a single reservoir and a cascade reservoir. The specific statistical indicators are shown in Table 1. Figure 3 、 Figure 4 From Table 1, Figure 3 and Figure 4 It can be seen that the statistical indicators (such as maximum value, minimum value, range, mean and mean square error) of the present invention in the scheduling problems of single reservoir and cascade reservoir group are significantly better than those of the PSO scheme, indicating that the reservoir group joint scheduling scheme obtained by the present invention has good stability and robustness.

[0140] Table 1

[0141]

[0142] To demonstrate the rationality of the results, Figure 5 The corresponding water level processes for the optimal scheduling schemes for cascade reservoirs obtained using the present invention and PSO are plotted. It can be seen that in the scheme obtained by the present invention, the cascade reservoirs retain as much water as possible in the early stages of scheduling and operate at high heads to increase power generation in the later stages. The results obtained by PSO fluctuate wildly, making it difficult to meet scheduling stability requirements. Furthermore, the power generation is significantly low, with significant potential for improvement. This demonstrates that the present invention can ensure the acquisition of a feasible joint scheduling scheme for a group of reservoirs, effectively improving the efficiency of water resource utilization in the river basin and meeting the comprehensive utilization needs of power grids and power generation companies.

[0143] While the present invention has been described above with reference to preferred embodiments, this is not intended to limit the present invention. Persons skilled in the art will readily appreciate that various modifications and variations can be made without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A reservoir group system joint scheduling layer optimization method, characterized in that: The steps include: Step S1: Based on the reservoir group data, a reservoir group system joint scheduling model is constructed with the goal of maximizing power generation, and constraints are added to the model; at the same time, the maximum number of iterations and the population size are preset; and the water level at the end of the reservoir scheduling period is determined as the optimization variable; Step S2: adopt the joint dispatching layer optimization method to solve the optimal joint dispatching scheme of the reservoir group system, specifically: Step 201, constructing the initial layer: generating an initial population within a preset range of decision variables, obtaining the population's velocity variables, storing individuals in the population as individual optimal positions, calculating the fitness of individuals in the population, and updating the dominant individuals based on the fitness of the individuals; specifically, including the following sub-steps: S2.

1. Randomly generate an initial population x within the preset range of the decision variable. Set the upper and lower limits of the velocity population v as γ times the upper and lower limits of the decision variable of the initial population x, as follows: Among them, x i,j and v i,j are the values ​​of the jth decision variable and speed variable of the i-th individual, Indicates that in the interval Function to generate uniformly distributed random numbers; S2.

2. Calculate the fitness of individuals in the initial population and store them as the optimal individual position pbest. At the same time, determine whether there is a feasible solution. If so, store the individual with the smallest fitness in the pbest population as the dominant individual gbest. Otherwise, store the individual with the smallest constraint violation value in the infeasible solution as the dominant individual gbest. Step 202: Construct the first level: Utilize the adaptive learning strategy and weight balancing update strategy to obtain dynamic weight factors; specifically, include: First, the first parallel layer is constructed: the weight W2 uses a weighted balanced update strategy to conduct a wide search in the entire solution space; then the second parallel layer is constructed: the weight W1 uses an adaptive learning strategy to bidirectionally link with the first parallel layer, while preventing it from falling into the local optimum during the balanced search, as shown in the following formula: in Among them, W1 and W2 are the dynamic weight factors obtained by the adaptive learning strategy and the weight balance update strategy respectively, t and T are the current number of iterations and the maximum number of iterations of the algorithm respectively, and a1, a2, a3 and a4 are weight calculation parameters; Step 203: Construct the second level: Use the nonlinear learning factor update strategy to obtain the dynamic learning factor; specifically, the learning factors r1 and r2 are updated through the nonlinear distribution of the cosine function, as shown in the following formula: Among them, r1 and r2 are dynamic learning factors obtained by the nonlinear learning factor update strategy; rand is a random number uniformly distributed in the interval [0,1]; b is a calculation parameter; Step 204: Construct the third level: Use the two-level joint search strategy to update the population; the upper level local search strategy: A=-abs((x i -gbest) / gbest) Superior overall search strategy: C=x i +(-1+2·round(rand))·x i ·((1-t / T) / 2·e (A) +(t / T)·e (B) ) Subordinate dynamic search strategy: D=W·v+r1·(pbest-x)+r2·(gbest-x) E = floor(N·c·(t / T))) Among them, A and B are factors corresponding to the upper-level local search strategy; C is the variable generated by the upper-level overall search strategy; D is the variable generated by the lower-level dynamic search strategy; abs is the absolute value function; floor is the floor function; round is the rounding function; N is the number of individuals in the population; rand is a random number uniformly distributed in the interval [0,1]; E is the number of individuals executing the upper-level overall search strategy; randi1, randi e ,randi E is a random integer function uniformly distributed in the interval [1, N]; i is the individual serial number; c is the proportional selection calculation parameter; Step 205, construct the fourth level: calculate the fitness of individuals in the current population, update the optimal position of individuals in the current population, and the position of dominant individuals; and determine whether the preset maximum number of iterations has been reached. If so, obtain the optimal joint scheduling plan for the reservoir group system and output the fitness value; otherwise, return to step S202.

2. A reservoir group system joint scheduling hierarchical optimization method according to claim 1, characterized in that: In step S1, the maximum power generation objective function of the reservoir group system joint scheduling model is: Where E is the total power generation of all reservoirs in the period; K is the number of reservoirs; J is the total number of periods; P k,j , Q k,j 、H k,j are the output, power generation flow and water head of the kth reservoir in the jth period respectively; A k is the output coefficient of the kth reservoir; t j is the number of hours in the jth period.

3. A reservoir group system joint scheduling hierarchical optimization method according to claim 1, characterized in that: In step S1, the constraints include: Output constraints: in, are the minimum and maximum outputs of the kth reservoir in the jth period respectively; Storage capacity constraints: in, are the minimum and maximum storage capacities of the kth reservoir in the jth period respectively; Water balance constraints Among them, I k,j , O k,j are the interval inflow and outflow of the kth reservoir in the jth period respectively; Remove lost water from outflows from upstream reservoirs; Outbound flow constraints: in, are the minimum and maximum outflows of the kth reservoir in the jth period respectively; Power generation flow constraints: in, are the minimum and maximum power generation flows of the kth reservoir in the jth period, respectively; Initial and final storage capacity constraints: in, are the initial and final storage capacities of the kth reservoir, respectively.

4. A reservoir group system joint scheduling hierarchical optimization method according to claim 1, characterized in that: In step S205, the fitness of individuals in the current population is calculated, and the optimal position of individuals in the current population and the position of dominant individuals are updated as follows: gbest=pbest k ,k=min(F(pbest)) Among them, F is the fitness value calculation function, and k is the serial number of the individual with the smallest fitness value in the optimal position.

5. A reservoir group system joint dispatching layer optimization system, characterized by: include: The model building module is configured to perform the following actions: based on the reservoir group data, build a reservoir group system joint scheduling model with the goal of maximizing power generation, and add constraints to the model; at the same time, preset the maximum number of iterations and population size; and determine the water level at the end of the reservoir scheduling period as the optimization variable; The optimal reservoir group system joint scheduling model solving module includes an initialization unit, a dynamic weight acquisition unit, a dynamic learning factor acquisition unit, a population update unit, and an optimal scheduling solution acquisition unit; The initialization unit is configured to perform the following actions: construct an initial layer: generate an initial population within the preset range of decision variables, obtain the speed variable of the population, store the individuals in the population as the individual's optimal position, calculate the fitness of the individuals in the population and update the dominant individual according to the individual's fitness; specifically perform the following steps: S2.1, randomly generate an initial population x within the preset range of decision variables, and set the upper and lower limits of the speed population v as γ times the upper and lower limits of the decision variables of the initial population x, as shown in the following formula: Among them, x i,j and v i,j are the values ​​of the jth decision variable and speed variable of the i-th individual, Indicates that in the interval Function to generate uniformly distributed random numbers; S2.

2. Calculate the fitness of individuals in the initial population and store them as the optimal individual position pbest. At the same time, determine whether there is a feasible solution. If so, store the individual with the smallest fitness in the pbest population as the dominant individual gbest. Otherwise, store the individual with the smallest constraint violation value in the infeasible solution as the dominant individual gbest. The dynamic weight acquisition unit is configured to perform the following actions: construct the first level: use the adaptive learning strategy and the weight balance update strategy to obtain the dynamic weight factor; first construct the first parallel layer: the weight W2 uses the weight balance update strategy to conduct a wide search in the entire solution space; then construct the second parallel layer: the weight W1 uses the adaptive learning strategy to bidirectionally link with the first parallel layer, and prevents falling into the local optimum while performing a balanced search, as shown in the following formula in Among them, W1 and W2 are the dynamic weight factors obtained by the adaptive learning strategy and the weight balance update strategy respectively, t and T are the current number of iterations and the maximum number of iterations of the algorithm respectively, and a1, a2, a3 and a4 are weight calculation parameters; The dynamic learning factor acquisition unit is configured to perform the following actions: construct the second level: obtain the dynamic learning factor using the nonlinear learning factor update strategy; specifically, the learning factors r1 and r2 are updated through the nonlinear distribution of the cosine function, as shown in the following formula: Among them, r1 and r2 are dynamic learning factors obtained by the nonlinear learning factor update strategy; rand is a random number uniformly distributed in the interval [0,1]; b is a calculation parameter; The population update unit is configured to perform the following actions: construct the third level: update the population using a two-level joint search strategy; the upper level local search strategy: A=-abs((x i -gbest) / gbest) Superior overall search strategy: C=x i +(-1+2·round(rand))·x i ·((1-t / T) / 2·e (A) +(t / T)·e (B) ) Subordinate dynamic search strategy: D=W·v+r1·(pbest-x)+r2·(gbest-x) E = floor(N·c·(t / T))) Among them, A and B are factors corresponding to the upper-level local search strategy; C is the variable generated by the upper-level overall search strategy; D is the variable generated by the lower-level dynamic search strategy; abs is the absolute value function; floor is the floor function; round is the rounding function; N is the number of individuals in the population; rand is a random number uniformly distributed in the interval [0,1]; E is the number of individuals executing the upper-level overall search strategy; randi1, randi e ,randi E is a random integer function uniformly distributed in the interval [1, N]; i is the individual serial number; c is the proportional selection calculation parameter; The optimal scheduling scheme acquisition unit is configured to perform the following actions: construct the fourth level: calculate the fitness of individuals in the current population, update the optimal position of individuals in the current population, and the position of dominant individuals; and determine whether the preset maximum number of iterations has been reached. If so, the optimal joint scheduling scheme for the reservoir group system is obtained and the fitness value is output; otherwise, it returns to the dynamic weight acquisition unit.

Images