Multi-scale optimization method and system for flood control of cascade reservoirs

Abstract

The application discloses a kind of cascade reservoir group flood control scheduling multi-scale optimization method and system, the present application is with the conventional dynamic programming method as framework, first obtain the initial scheduling state of each reservoir satisfying constraint condition and neighborhood space;Then in each iteration process, in turn select a few computing stages to execute lens imaging operation, according to the upper and lower limit range on neighborhood space, the mirror image state of initial state is calculated to carry out iterative optimization, to seek optimal scheduling state, repeat the above multi-scale lens imaging operation, until different synchronous computing stages have been completed lens imaging search;Finally, the range of neighborhood space is contracted, and repeated iteration is carried out until convergence, so as to approach global optimal solution, and output optimal scheduling result.The method principle of the present application is simple, the control parameter is few, easy to realize, the amount of calculation is small, the search ability is strong, has excellent optimization ability and optimization precision, and is suitable for efficiently solving large-scale reservoir group optimization scheduling problem.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of flood control scheduling of cascade reservoir groups, in particular to a multi-scale optimization method and system for flood control scheduling of cascade reservoir groups. BACKGROUND

[0002] The joint flood control scheduling of cascade reservoir groups is an operation control method that considers the flood control needs of upstream and downstream protection objects and uses reserved flood control storage to generate flood control benefits under the premise of ensuring safety, and the purpose is to fully and effectively utilize the flood control storage to achieve flood volume storage and flood peak reduction, so as to minimize the loss of downstream flood control objects in flood disasters. With the rapid development of social economy, the optimization scheduling of reservoir groups is increasingly important, and plays a significant role in reducing flood disaster losses and maintaining the safety of national life and property.

[0003] From a mathematical point of view, the optimization scheduling of cascade reservoir groups is a complex high-dimensional optimization problem with multiple stages, multiple variables and multiple constraints, and its optimal solution is not easy to obtain. The dynamic programming method (DP) is an effective method for solving multi-stage optimization scheduling problems, and has been maturely applied to the solution of single reservoir flood control optimization scheduling problems, but for the optimization scheduling problem of cascade reservoir groups involving multiple reservoirs, especially large-scale cascade reservoir groups, due to the large scale of reservoirs, multiple optimization variables and multiple calculation stages, the calculation complexity increases sharply, resulting in "curse of dimensionality", premature convergence and low solving efficiency, and a practical and mature calculation method is urgently needed to effectively solve the optimization scheduling problem of cascade reservoir groups.

[0004] Therefore, the applicant considers proposing a multi-scale optimization method for flood control scheduling of cascade reservoir groups to effectively solve the optimization scheduling problem of cascade reservoir groups. SUMMARY

[0005] In order to overcome the shortcomings of the above-mentioned technologies, the purpose of the present application is to provide a multi-scale optimization method and system for flood control scheduling of cascade reservoir groups, which solves the problems of "curse of dimensionality" and poor search ability of DP in solving the optimization scheduling problem of reservoir groups, and presents excellent optimization ability and optimization precision.

[0006] To achieve the above-mentioned purpose, the technical solutions adopted by the present application are as follows:

[0007] A multi-scale optimization method for flood control scheduling of cascade reservoir groups, comprising the following steps:

[0008] 1) Determine the initial calculation conditions, including the objective function, constraint conditions and decision variables of the optimization scheduling of cascade reservoir groups;

[0009] 2) setting calculation parameters; the calculation parameters include reservoir number N, scheduling period stage number T, maximum iteration number M, maximum lens imaging calculation scale A, maximum lens imaging calculation number K, and convergence precision ε;

[0010] 3) obtaining initial state of each reservoir in each stage satisfying the constraint condition by using conventional method or artificial experience and neighborhood space Ω=[p i,j ,q i,j ],1≤i≤N,1≤j≤T;

[0011] wherein represents initial state of dam front water level of reservoir i in stage j; p i,j and q i,j represent upper and lower limits of water level of reservoir i in stage j, respectively;

[0012] 4) setting iteration number m=1;

[0013] 5) setting lens imaging calculation scale a=0;

[0014] 6) setting stage number j=1;

[0015] 7) selecting 1+a synchronous calculation stages from j to j+a to perform lens imaging search;

[0016] 8) comparing water level difference of each stage of adjacent two times of optimal state, if the water level difference is less than or equal to the convergence precision ε, then the neighborhood space is contracted, and step 9) is proceeded, if the water level difference is greater than the convergence precision ε, then step 5) is proceeded;

[0017] 9) m=m+1, if m>M, then step 10) is proceeded, otherwise step 5) is proceeded;

[0018] 10) stopping calculation, and outputting final optimal state.

[0019] Preferably, in the step 1), the objective function is:

[0020]

[0021] The constraint condition includes:

[0022] (1) water quantity balance constraint of each reservoir: V i,j =V i,j-1 +(I i,j-1 -O i,j-1 )×Δt,

[0023] (2) hydraulic connection constraint of each reservoir: I i,j =O i,j +R i,j,

[0024] (3) Each reservoir water level constraint:

[0025] (4) Each reservoir outflow constraint:

[0026] (5) Each reservoir discharge capacity constraint: i,j ≤O i,max (Z i,j ),

[0027] (6) Each reservoir initial and final water level constraint: i,0 =Z i,start ,Z i,T =Z i,end ;

[0028] (7) Non-negative constraint: each variable is non-negative;

[0029] The decision variable is the dam front water level value Z i,j of each reservoir at each stage.

[0030] In the formula, W is the flow process square sum of the flood control object; V i,j is the reservoir capacity value of reservoir i at the end of stage j, m 3 ; I i,j is the inflow of reservoir i at stage j, m 3 / s; O i,j is the outflow of reservoir i at stage j, m 3 / s; Δt is the stage length, h; and are the minimum and maximum outflow of reservoir i at stage j, m 3 / s; R i,j is the interval flow from reservoir i at stage j to the downstream reservoir or downstream flood control section, m 3 / s; Z i,j is the dam front water level value of reservoir i at stage j, m; and are the minimum and maximum dam front water level values of reservoir i at stage j, m; O i,max (Z i,j ) is the maximum discharge capacity of reservoir i at stage j corresponding to water level Z i,j , m 3 / s; Z i,start and Z i,end are the initial reservoir water level and the final control water level of reservoir i during the dispatching period, m.

[0031] Preferably, in the step 2), the value range of the maximum lens imaging calculation scale A is 1-4.

[0032] Preferably, the calculation step of the step 7) is as follows:

[0033] 7.1) initializing the lens imaging search times k = 1;

[0034] 7.2) combining the calculation principle of the lens imaging, generating the lens state of the i-th reservoir in the k-th lens imaging search of each stage for the 1+a synchronous calculation stages from j to j+a; The calculation formula is:

[0035]

[0036] In the formula, g is a Gaussian distributed random number; n is a lens imaging scaling factor, which is used to control the search range of the lens imaging;

[0037] 7.3) comparing the lens state of the lens imaging search with the current state If the lens state is better than the current state, the current state is replaced by the lens state value, otherwise no processing is performed; wherein, is the water level state of the m-th iteration of the i-th reservoir in each stage;

[0038] 7.4) k = k + 1, if k≤K, go to step 7.2), otherwise go to step 7.5);

[0039] 7.5) j = j + 1, if j≤T-a, go to step 7.1), otherwise go to step 7.6);

[0040] 7.6) a = a + 1, if a≤A, go to step 6), otherwise go to step 8).

[0041] Preferably, the value range of the lens imaging scaling factor n is 1000-3000.

[0042] Preferably, in the step 8), the calculation method of the water level difference value of each stage of the adjacent two optimal states is:

[0043]

[0044] Preferably, in the step 8), the contraction of the neighborhood space is specifically: q i,j = q i,j * δ, δ is a neighborhood contraction factor.

[0045] Preferably, the value range of the neighborhood contraction factor δ is 0.5-0.95.

[0046] A multi-scale optimization system for flood control scheduling of a cascade reservoir group comprises:

[0047] An initialization module is configured to determine a target function, constraint conditions and decision variables for flood control optimization scheduling of the cascade reservoir group, and generate scheduling states and a neighborhood range satisfying the constraint conditions according to artificial experience decision or a conventional dynamic programming method under the constraint conditions;

[0048] A multi-scale lens imaging optimization module is configured to repeatedly perform lens imaging search and iteratively perform scheduling state optimization.

[0049] An output module is configured to repeatedly execute the multi-scale lens imaging optimization module until a preset iteration stop condition is met, obtain an optimal scheduling state, and output the optimal scheduling state as a final reservoir group optimization scheduling result.

[0050] Preferably, the multi-scale lens imaging optimization module comprises:

[0051] A synchronous calculation stage selection submodule is configured to select different calculation stages according to a current calculation scale to perform synchronous calculation of the calculation stages.

[0052] A lens state generation submodule is configured to calculate a lens state of a current state according to a calculation principle of lens imaging.

[0053] A state comparison and replacement submodule is configured to compare the current state and the lens state to determine an optimal state.

[0054] Compared with the prior art, the present application has the following advantages:

[0055] The present application uses a conventional dynamic programming method as a framework, first obtains initial scheduling states and a neighborhood space of the reservoirs satisfying the constraint conditions, then in each iteration process, selects a few calculation stages to perform lens imaging operation, calculates the mirror state of the initial state according to the upper and lower limits of the neighborhood space to iteratively optimize the scheduling state, repeats the above multi-scale lens imaging operation until the lens imaging search of different synchronous calculation stages is completed, finally shrinks the neighborhood space range and iteratively converges to approximate the global optimal solution, and outputs the optimal scheduling result.

[0056] The present application iteratively performs scheduling state optimization through lens imaging operation, avoids full combination between all discrete states of the reservoir group at each stage, and significantly improves the calculation efficiency; the present application expands the variable scale calculation of the calculation stages on the basis of only considering one calculation stage in the past, selects different stages for synchronous optimization to increase the search ability and avoid falling into a local optimum; the present application has simple and convenient method principle, few control parameters, easy implementation, small calculation amount and strong search ability, has excellent optimization ability and optimization precision, and is suitable for efficiently solving large-scale reservoir group optimization scheduling problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is a flow chart of a multi-scale optimization method for flood control scheduling of a cascade reservoir group according to the present invention;

[0058] Figure 2 This is the principle diagram of lens imaging;

[0059] Figure 3 This is a diagram showing the flood control dispatch calculation results of Reservoir A in a cascade reservoir group under the 1965 0.5% water inflow condition in an embodiment of the present invention;

[0060] Figure 4 This is a diagram showing the flood control dispatch calculation results of Reservoir B in a cascade reservoir group under the 1965 0.5% water inflow condition in an embodiment of the present invention;

[0061] Figure 5 This is a diagram showing the flood control dispatch calculation results of Reservoir C in a cascade reservoir group under the 1965 0.5% water inflow condition in an embodiment of the present invention. DETAILED DESCRIPTION

[0062] In order to better explain the present invention, the main contents of the present invention are further illustrated below in conjunction with specific examples, but the contents of the present invention are not limited to the following examples.

[0063] To reduce the flood control burden and flood losses downstream of reservoir groups and provide technical support for real-time flood control scheduling decisions, the maximum peak-shaving criterion is currently widely used to construct the optimal scheduling objective function. To this end, based on the maximum peak-shaving criterion, the flood control optimization scheduling model for cascade reservoir groups can generally be described as follows: given the initial water level, final water level, inflow flood process, and interval flood process of each reservoir during the scheduling period, while satisfying the complex water level and flow constraints of each reservoir and upstream and downstream protection objects, the optimization objective is to minimize the sum of the squares of the flow processes of each flood control object. The specific mathematical expression is:

[0064]

[0065] Where: W is the sum of squares of the flow process of the flood control object; N is the number of reservoirs; i is the reservoir number, i = 1, 2, ..., N; T is the total number of stages in the scheduling period; j is the stage number, j = 1, 2, ..., T; O i,j is the outflow of reservoir i in stage j (m 3 / s); R i,j is the interval flow from reservoir i to the downstream reservoir or downstream flood control section at stage j (m 3 / s).

[0066] The constraints that need to be considered in the optimization calculation of flood control operation of cascade reservoirs mainly include:

[0067] (1) Each reservoir water balance constraint: V i,j = V i,j-1 + (I i,j-1 - O i,j-1 ) x Δt,

[0068] (2) Each reservoir hydraulic connection constraint: I i,j = O i,j + R i,j ,

[0069] (3) Each reservoir water level constraint:

[0070] (4) Each reservoir outflow constraint:

[0071] (5) Each reservoir discharge capacity constraint: O i,j ≤ O i,max (Z i,j ),

[0072] (6) Each reservoir initial and final water level constraint: Z i,0 = Z i,start , Z i,T = Z i,end

[0073] (7) Non-negative constraint: all variables are non-negative.

[0074] In the formula, V i,j is the reservoir i capacity value at the end of stage j (m 3 ); I i,j is the reservoir i inflow at stage j (m 3 / s); O i,j is the reservoir i outflow at stage j (m 3 / s); Δt is the stage duration (h); and are the minimum and maximum outflow of reservoir i at stage j (m 3 / s) and (m 3 / s), respectively; Z i,j is the reservoir i dam front water level value at stage j (m); and are the minimum and maximum dam front water level of reservoir i at stage j (m); O i,max (Z i,j ) is the maximum discharge capacity of reservoir i at stage j corresponding to water level Z i,j (m 3 / s); Z i,start and Zi,end are the initial reservoir water level (m) and the controlled water level (m) at the end of the scheduling period of reservoir i, respectively.

[0075] Clearly, based on the aforementioned objective function and constraints, the optimal operation of cascade reservoirs for flood control is characterized by multiple constraints, high dimensionality, and nonlinearity, making its optimal solution difficult to obtain. Traditional DP methods have limitations when addressing this complex optimization problem. Furthermore, as the number of reservoirs and computational stages increases, they often face a serious "curse of dimensionality." New methods with robust search capabilities are needed to provide new insights into the development of optimized flood control operation plans.

[0076] Lens imaging is an optical law in physics. The current solution is regarded as an individual, and the process of the individual obtaining its reverse solution in the solution space can be regarded as a lens imaging process. Studies have shown that some positions of the individual reverse solution may be closer to the optimal solution. Therefore, by obtaining the lens imaging state of the current state and finding the candidate solution to the corresponding problem, it is possible to effectively avoid falling into the local optimum. The method of the present invention considers combining multi-scale lens imaging to optimize the scheduling process, avoiding the dimensionality curse problem of the comprehensive combination of all discrete states of each reservoir at each stage during DP solution, improving the ability to jump out of the local optimum, and having good optimization ability and optimization accuracy.

[0077] The principle of lens imaging is as follows Figure 2 As shown. In a given space, there is an interval range [lb, ub] on the horizontal axis. In this interval, the height of an individual S is h, and its projection on the horizontal axis is x. At point o = (lb + ub) / 2, there is a convex lens with a focal length of f. Point S is mapped to point S through the lens. * Department, S * The height is h * , S * The projection on the horizontal axis is x * At this time, x * That is, x is mapped to the reverse entity on the horizontal axis through lens imaging.

[0078] The inverse solution x obtained by lens imaging mapping of individual x * , its mathematical formula is:

[0079]

[0080] Let h / h * =n, n is the lens imaging scaling factor, convert the above formula, x * The calculation formula is transformed into the following formula:

[0081] x * =(lb+ub) / 2+(lb+ub) / (2·n)-x / n (2)

[0082] Further, in order to make full use of the optimization information, the Gaussian perturbation operator is used to perturb x in the lens mapping process, so that x * The calculation formula is updated as:

[0083] x * =(lb+ub) / 2+(lb+ub) / (2·n)-g·x / n (3)

[0084] In the formula, g is a Gaussian distributed random number, and the purpose is to use the probability characteristics of Gaussian distribution, that is, most Gaussian random numbers are perturbed in a small range, so as to improve the ability to jump out of local optimum; a small number of Gaussian random numbers are perturbed far away from themselves, so as to maintain the diversity of calculation.

[0085] Finally, the above two-dimensional lens imaging mapping method is extended and applied to the cascade reservoir group optimization scheduling problem, and the dam front water level value Z i,j , the upper limit of water level p i,j , and the lower limit of water level q i,j are respectively replaced by x, ub and lb, so as to obtain the expression of the lens state f i,j of Z i,j :

[0086] f i,j =(p i,j +q i,j ) / 2+(p i,j +q i,j ) / (2·n)-(g·Z i,j ) / n (4).

[0087] In addition, considering different calculation scales, the corresponding synchronous calculation stage can be further selected. Thus, the calculation method of the multi-scale lens imaging state proposed in the present application is obtained.

[0088] Based on the above idea, as shown in Figure 1 , the multi-scale optimization method for flood control scheduling of cascade reservoir groups provided by the present application comprises the following steps:

[0089] 1) Determine the initial calculation conditions, including the objective function, constraint conditions and decision variables of the cascade reservoir group flood control optimization scheduling;

[0090] Among them, the objective function is:

[0091]

[0092] The constraint conditions include:

[0093] (1) Water balance constraint of each reservoir: V i,j =V i,j-1 +(Ii,j-1 -O i,j-1 )×Δt,

[0094] (2) Each reservoir hydraulic connection constraint: I i,j =O i,j +R i,j ,

[0095] (3) Each reservoir stage water level constraint:

[0096] (4) Each reservoir outflow constraint:

[0097] (5) Each reservoir discharge capacity constraint: O i,j ≤O i,max (Z i,j ),

[0098] (6) Each reservoir initial and final water level constraint: Z i,0 =Z i,start ,Z i,T =Z i,end

[0099] (7) Non-negative constraint: all variables are non-negative.

[0100] The decision variable is the dam water level value Z i,j of each reservoir at each stage.

[0101] 2) Set the calculation parameters, including the number of reservoirs N, the number of scheduling stages T, the maximum number of iterations M, the maximum lens imaging calculation scale A (generally take value between 1-4, determine the calculation stage selection range of lens imaging search), the maximum number of lens imaging calculations K; convergence precision ε.

[0102] 3) Use conventional methods or artificial experience to obtain the initial state of each reservoir at each stage that satisfies each constraint condition and the neighborhood space Ω = [p i,j , q i,j ], 1≤i≤N, 1≤j≤T, where represents the initial state of the dam water level of reservoir i at stage j; p i,j and q i,j represent the upper and lower limits of the water level of reservoir i at stage j, respectively.

[0103] 4) Set the iteration number m = 1;

[0104] 5) Set the lens imaging calculation scale a = 0;

[0105] 6) Set the stage number j = 1;

[0106] 7) Select j to j+a total 1+a synchronous computing phase to perform lens imaging search. When a = 0, then only for j lens imaging search. The specific calculation steps are as follows:

[0107] 7.1) initialization lens imaging search times k = 1;

[0108] 7.2) combined with the calculation principle of lens imaging, for j to j+a total 1+a synchronous computing phase, the lens state of each stage of the reservoir i kth lens imaging search is generated The specific calculation formula is:

[0109]

[0110] In the formula, g is a Gaussian distributed random number; n is the lens imaging scaling factor, which is used to control the search range of lens imaging, generally taking the value between 1000-3000.

[0111] 7.3) compare the lens state of lens imaging search And the current state If the lens state is better than the current state, replace the current state with the lens state; otherwise, do not process; wherein, is the water level state of the mth iteration of each stage of the reservoir i.

[0112] 7.4) k = k + 1, if k≤K, go to step 7.2); otherwise, go to step 7.5).

[0113] 7.5) j = j + 1, if j≤T-a, go to step 7.1); otherwise go to step 7.6).

[0114] 7.6) a = a + 1, if a≤A, go to step 6); otherwise go to step 8).

[0115] 8) calculate the water level difference of each stage of the adjacent two optimal states, if:

[0116]

[0117] Then shrink the neighborhood space, let q i,j = q i,j *δ, δ is the neighborhood shrinkage factor, generally taking the value between 0.5-0.95, go to step 9); otherwise go to step 5).

[0118] When it is judged that the water level difference of each stage of the adjacent two optimal states is very small, the multi-scale lens imaging search is stopped, and the next iteration is turned; otherwise, when it is judged that the water level difference of each stage of the adjacent two optimal states is quite different, indicating that there is still optimization space, the multi-scale lens imaging search is continued in the current iteration.

[0119] 9) Let m = m + 1, if m > M, turn to step 10); otherwise, turn to step 5).

[0120] 10) Stop calculation, and output the final optimal state.

[0121] A multi-scale optimization system for flood control scheduling of a cascade reservoir group implementing the above method, comprising:

[0122] An initialization module for determining a target function, constraint conditions and decision variables of the flood control optimization scheduling of the cascade reservoir group, and generating scheduling states and neighborhood ranges satisfying the constraint conditions according to artificial experience decision or a conventional dynamic programming method under various constraint conditions;

[0123] A multi-scale lens imaging optimization module for repeatedly performing lens imaging search and iteratively performing scheduling state optimization, comprising three sub-modules of synchronous calculation stage selection, lens state generation and state comparison and replacement;

[0124] The synchronous calculation stage selection sub-module is used for selecting different calculation stages according to the current calculation scale to perform synchronous calculation of a few calculation stages; wherein the number of calculation stages is determined by the maximum lens imaging calculation scale A;

[0125] The lens state generation sub-module is used for calculating the lens state of the current state according to the calculation principle of lens imaging;

[0126] The state comparison and replacement sub-module is used for comparing the current state and the lens state to constantly optimize and replace the current state;

[0127] An output module for repeatedly executing the multi-scale lens imaging optimization module until a preset iteration stop condition is met, obtaining an optimal scheduling state, and outputting the optimal scheduling state as a final reservoir group optimization scheduling result.

[0128] Next, a cascade reservoir group of a certain river is taken as an example to further describe the technical scheme of the application.

[0129] The total flood control capacity of the three large-scale reservoirs A, B and C reaches 45 billion m 3 .

[0130] Select 0.2%, 0.5% and 1% typical design flood in 1965 as the water series, respectively, using the method and DP to carry out the calculation of flood control optimization scheduling of cascade reservoir group. The specific calculation results are shown in Table 1.

[0131] From Table 1, for three frequency design floods, compared with the traditional DP method, the calculation results and calculation time of the method are obviously superior, specifically: ① from the calculation results, the optimization results of the method are smaller, and the discharge of the last reservoir C is smaller, which shows that the cascade reservoirs play a good role in blocking, storing and discharging for each flood; ② from the calculation time, the method only takes 15% of the time of DP, greatly improving the operation efficiency. It can be seen that the method has high calculation precision and fast efficiency, and is effective for solving the flood control optimization scheduling problem of cascade reservoir group.

[0132] Table 1: Comparison of calculation results of the method and DP

[0133]

[0134]

[0135] As Figures 3-5 shown is the flood control scheduling calculation result of the above cascade reservoirs of the river under the condition of 0.5% water in 1965 by using the method, the scheduling results meet the constraint conditions of water level, storage capacity and flow, and the discharge of the last reservoir C of the cascade reservoirs is very stable, which shows that the method has the functions of peak shaving, compensation regulation and disaster prevention and reduction.

[0136] In summary, the method is based on the lens imaging theory, considers the selection of different scales of synchronous calculation stages, and proposes an optimization method based on multi-scale lens imaging, which can be effectively coupled with the DP calculation framework. First, the lens imaging operation is used to iteratively execute the scheduling state optimization, which avoids the full combination between all discrete states of the reservoir group in each stage, and significantly improves the calculation efficiency. Second, on the basis of only considering one calculation stage in the past, the variable scale calculation of the calculation stage is expanded, and the search ability is increased by selecting different stages for synchronous optimization to avoid falling into local optimum. The method has simple principle, few control parameters, easy implementation, small calculation amount and strong search ability, and has excellent optimization ability and optimization precision, and is suitable for efficiently solving the optimization scheduling problem of large-scale reservoir group.

[0137] Those skilled in the art will readily understand that the above description is only a preferred embodiment of the present application, and is not intended to limit the present application, and any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0138] The other parts not specified belong to the state of the art.

Claims

1. A multi-scale optimization method for flood control scheduling of a cascade reservoir group, characterized in that: The method comprises the following steps: 1) determining initial calculation conditions, including an objective function, constraint conditions and decision variables of flood control optimization scheduling of the cascade reservoir group; 2) setting calculation parameters; the calculation parameters include the number of reservoirs N, the number of stages T in the scheduling period, the maximum number of iterations M, the maximum lens imaging calculation scale A, the maximum number of lens imaging calculations K, and the convergence precision ε; 3) using conventional methods or human experience, obtaining the initial state of each reservoir at each stage that satisfies the constraints and the neighborhood space Ω = [p i,j , q i,j ], 1≤i≤N, 1≤j≤T; wherein represents the initial state of the water level in reservoir i before the dam at stage j; p i,j and q i,j respectively represent the upper and lower limits of the water level in reservoir i at stage j; 4) setting the number of iterations m = 1; 5) setting the lens imaging calculation scale a = 0; 6) setting the stage number j = 1; 7) selecting 1 + a synchronous calculation stages from j to j + a to perform lens imaging search; 7.1) initializing the number of lens imaging search times k = 1; 7.2) combined with the computational principle of lens imaging, for j to j+a total of 1+a synchronous calculation stages, generate the lens state of the i-th reservoir k-time lens imaging search of each stage The calculation formula is: In the formula, g is a Gaussian distributed random number; and n is a lens imaging scaling factor used to control the search range of lens imaging; 7.3) Compare the lens state of the lens imaging search and the current state If the lens state is better than the current state, replace the current state with the lens state value, otherwise do nothing; wherein, is the water level state of the reservoir i for the mth iteration of the stage. 7.4) setting k = k + 1, and if k ≤ K, proceeding to step 7.2), otherwise proceeding to step 7.5); 7.5) setting j = j + 1, and if j ≤ T - a, proceeding to step 7.1), otherwise proceeding to step 7.6); 7.6) setting a = a + 1, and if a ≤ A, proceeding to step 6), otherwise proceeding to step 8); 8) comparing the water level difference of each stage of the adjacent two optimal states, if the water level difference is less than or equal to the convergence precision ε, shrinking the neighborhood space, and proceeding to step 9), if the water level difference is greater than the convergence precision ε, proceeding to step 5); 9) setting m = m + 1, and if m > M, proceeding to step 10), otherwise proceeding to step 5); 10) stopping calculation and outputting the final optimal state.

2. The multi-scale optimization method for flood control scheduling of the cascade reservoir group according to claim 1, characterized in that: In the step 1), the objective function is: The constraint conditions include: (1) Reservoir water balance constraints: (2) The hydraulic connection constraint of each reservoir: (3) Reservoir water level constraints: (4) Reservoir outflow constraints: (5) Reservoir release capacity constraints: (6) The initial and final water level constraints of each reservoir: Z i,0 = Z i,start , Z i,T = Z i,end ; (7) non-negative constraint: all variables are non-negative values; The decision variable is the dam front water level value Z of each reservoir at each stage i,j ; Wherein, W is the square sum of flow process of flood control object; V i,j is the storage capacity value of reservoir i at the end of j stage, m 3 ; I i,j is the inflow of reservoir i in j stage, m 3 / s; O i,j is the outflow of reservoir i in j stage, m 3 / s; Δt is the stage length, h; and are respectively the minimum outflow and the maximum outflow of reservoir i in j stage, m 3 / s; R i,j is the interval flow from reservoir i in j stage to downstream reservoir or downstream flood control section, m 3 / s; Z i,j is the water level value before dam of reservoir i in j stage, m; and are respectively the minimum water level before dam and the maximum water level before dam of reservoir i in j stage, m; O i,max (Z i,j ) is the maximum discharge capacity of reservoir i in j stage corresponding to water level Z i,j , m 3 / s; Z i,start and Z i,end are respectively the initial reservoir water level at the beginning of dispatching period and the control water level at the end of dispatching period of reservoir i, m.

3. The multi-scale optimization method for flood control scheduling of the cascade reservoir group according to claim 1, characterized in that: In the step 2), the value range of the maximum lens imaging calculation scale A is 1-4.

4. The multi-scale optimization method for flood control scheduling of the cascade reservoir group according to claim 1, characterized in that: The value range of the lens imaging scaling factor n is 1000-3000.

5. The multi-scale optimization method for flood control scheduling of the cascade reservoir group according to claim 1, characterized in that: In the step 8), the calculation method of the water level difference of each stage of the adjacent two optimal states is:

6. The multi-scale optimization method for flood control scheduling of cascade reservoirs according to claim 1, characterized in that: In the step 8), the neighborhood space is contracted specifically as follows: δ is a neighborhood contraction factor.

7. The multi-scale optimization method for flood control scheduling of the cascade reservoir group according to claim 6, characterized in that: The value range of the neighborhood shrinkage factor δ is 0.5-0.

95.

8. A multi-scale optimization system for flood control scheduling of a cascade reservoir group, used to implement the multi-scale optimization method for flood control scheduling of a cascade reservoir group according to claim 1, characterized in that, It comprises: An initialization module for determining the objective function, constraint conditions and decision variables of flood control optimization scheduling of the cascade reservoir group, and generating scheduling states and neighborhood ranges meeting the constraint conditions according to artificial experience decision or a conventional dynamic programming method under the constraint conditions; A multi-scale lens imaging optimization module for repeatedly performing lens imaging search and iteratively performing scheduling state optimization; An output module for repeatedly executing the multi-scale lens imaging optimization module until a preset iteration stopping condition is met, obtaining an optimal scheduling state, and outputting the optimal scheduling state as a final reservoir group optimization scheduling result.

9. The multi-scale optimization system for flood control scheduling of the cascade reservoir group according to claim 8, characterized in that: The multi-scale lens imaging optimization module comprises: A synchronous calculation stage selection submodule for selecting different calculation stages according to the current calculation scale to perform synchronous calculation of the calculation stages; A lens state generation submodule for calculating the lens state of the current state according to the calculation principle of lens imaging; A state comparison and replacement submodule for comparing the current state and the lens state to determine the optimal state.

Images