Flood classification method and system based on mixed flood optimization algorithm

Abstract

The invention discloses a flood classification method and system based on a mixed flood optimization algorithm. The method comprises the following steps: collecting sample set data of flood classification; performing normalization processing; establishing a flood classification combination weight fuzzy clustering iteration model; performing population initialization arrangement operation; updating the optimal individual through a flood routing simulation operation, a flood change simulation operation, a catfish effect updating operation and a teaching optimization updating operation in sequence; terminating condition judgment operation, and outputting an optimal individual; and obtaining a flood classification fuzzy clustering membership degree matrix according to the optimal individual, and finally judging the level of the flood sample according to a level characteristic value formula. According to the method provided by the invention, the calculation complexity of the model is effectively reduced, the flood classification clustering effect is improved, a scientific basis is provided for flood level identification and flood control situation research and judgment, flood level evaluation can be quickly carried out, and a scientific basis is provided for flood level identification and flood control situation research and judgment under complex conditions.

Description

Technical Field

[0001] The invention relates to the technical field of hydrological analysis, and in particular to a flood classification method and system based on a hybrid flood optimization algorithm. Background Art

[0002] The purpose of flood classification is to analyze the similarities and differences in flood causes, time distribution, spatial distribution, and peak-volume relationships, and to use effective methods to cluster flood samples to reasonably classify floods, providing data support for quickly and effectively identifying flood control situations in actual flood control and scheduling decisions.

[0003] At present, there are many methods for flood classification, including projection pursuit, fuzzy clustering iteration model, etc. However, due to the complexity of the projection pursuit calculation process, it is necessary to rely on particle swarm optimization, differential evolution algorithm, etc. for coupled calculation, and the subjective division based on the projection value scatter plot is relatively subjective; while the fuzzy clustering iteration model can comprehensively consider the distribution and weight of flood classification indicators, and the calculation effect is relatively reasonable. It is a reasonable and effective flood classification method. However, the calculation process of the fuzzy clustering iteration model is relatively complicated. Generally, the fuzzy clustering iteration model is combined with intelligent optimization algorithms such as differential evolution and particle swarm. The optimization algorithm is used to replace the repeated derivation of the fuzzy clustering iteration formula to improve the classification effect of fuzzy clustering. Due to the problems of some algorithms such as weak search ability, long calculation time, and easy to fall into local optimality, the practical application promotion is limited, and new intelligent optimization algorithms are urgently needed to be developed.

[0004] The Flood algorithm (FLA) is a novel swarm intelligence optimization algorithm proposed in June 2024. It is inspired by the study of the complex movement and flow patterns of floods. It has the advantages of clear principles, simple structure, and fast optimization rate. It mainly includes population initialization, flood evolution simulation, and flood change simulation operations.

[0005] Based on this, the present invention considers proposing a new flood classification method in combination with a flood optimization algorithm to quickly evaluate flood levels and provide a scientific basis for flood level identification and flood control situation analysis under complex conditions. Summary of the invention

[0006] In order to overcome the shortcomings of the above-mentioned technology, the purpose of the present invention is to provide a flood classification method and system based on a hybrid flood optimization algorithm to solve the problems in the prior art such as complex calculation process, inaccurate results, weak search ability, long calculation time, and easy to fall into local optimality.

[0007] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0008] A flood classification method based on a hybrid flood optimization algorithm comprises the following steps:

[0009] 1) Collect sample data for flood classification;

[0010] 2) The sample set data is normalized;

[0011] 3) Based on the normalized sample data set, a flood classification combined weight fuzzy clustering iterative model is established, including setting the fuzzy clustering membership matrix U = (u hi ) c×n 、Fuzzy cluster center matrix S=(s jh ) m×c and the combined weight vector ω=(ω1,ω2,...,ω m ), and construct the fuzzy clustering optimization objective function;

[0012] 4) Population initialization and arrangement operation, based on the flood classification combined weight fuzzy clustering iteration model, to obtain the optimal individual for initialization;

[0013] 5) Execute flood evolution simulation operation, flood change simulation operation, catfish effect update operation, and teaching optimization update operation in sequence to continuously update the optimal individual;

[0014] 6) Termination condition judgment operation, that is, judging whether the current number of iterations reaches the maximum number of iterations. If g≤G, set g=g+1 and repeat step 5); otherwise, stop iteration and output the optimal individual;

[0015] 7) According to the optimal individual, the fuzzy clustering center matrix and combined weight vector of flood classification are obtained, and the fuzzy clustering membership matrix of flood classification is obtained according to the fuzzy clustering iteration formula. Finally, according to the level eigenvalue formula To determine the level of flood sample xi.

[0016] Preferably, in step 1), the flood sample set is an n×m indicator feature matrix Y:

[0017]

[0018] Where n is the number of flood sample sets, m is the sample y i The number of indicators included, y i is the i-th sample, i=1, 2, ..., n, y i =(y i1 ,y i2 ,...,y im );y ij is the jth index of the ith sample, j = 1, 2, ..., m.

[0019] Preferably, the indicators are statistical data describing floods, including peak water level, peak flow, flood duration, and total flood volume.

[0020] Preferably, in step 2), the flood sample y i The normalized matrix R after normalization is (r ij ) n×m :

[0021] r ij =(y ij -y min (j)) / (y max (j)-y min (j));

[0022] Among them, r ij is the normalized value of index j of sample i, y max (j) and y min (j) are the maximum and minimum values ​​of index j, respectively.

[0023] Preferably, in step 3), the fuzzy clustering membership matrix U is as follows:

[0024]

[0025] Among them, u hi is the membership degree of sample i belonging to category h, i = 1, 2, ..., n; h = 1, 2, ..., c; c is the number of categories for cluster analysis of flood sample set, and the number of flood categories c ≥ 3.

[0026] Preferably, in step 3), the fuzzy cluster center matrix S is as follows:

[0027]

[0028] Among them, s jh is the cluster center of index j for category h.

[0029] Preferably, in step 3), the combined weight is obtained by adding the subjective weight and the objective weight of the flood sample index; the combined weight ω=(ω1,ω2,...,ω m ),,in:

[0030] ω j =β·ω Sj +(1-β)·ω Oj ,j=1,2,...,m;

[0031] The subjective and objective weights are ω S =(ω S1 ,ω S2 ,...,ωSm ) and ω O =(ω O1 ,ω O2 ,...,ω Om );

[0032] ω m is the combined weight of the mth indicator, ω Sj is the subjective weight of the jth indicator, ω Oj is the objective weight of the jth indicator; β is the sensitivity coefficient of the main and objective weights, and β∈[0,1].

[0033] Preferably, in step 3), the fuzzy clustering optimization objective function is as follows:

[0034]

[0035] According to the Lagrangian function method, solving the above formula can obtain the following fuzzy clustering iteration formula:

[0036]

[0037] Preferably, the step 4) comprises:

[0038] (1) Set the population size of the hybrid flood optimization algorithm to NP and the maximum number of iterations to G. The population consists of NP D-dimensional individuals, and each individual Recorded as have Among them, np=1,2,...,NP, g=0,1,2,...,G, d=1,2,...,D, D=m×c+1;

[0039] (2) Select the fuzzy cluster center matrix (s jh ) m×c and sensitivity coefficient β are optimization variables, then the population individuals Arrange them as follows:

[0040]

[0041] Among them, individual It is an m×c+1 dimensional variable, the first m×c dimensions are the cluster center code, and the last dimension is the sensitivity coefficient β code;

[0042] (3) Let the number of iterations g = 0, and according to the flood sample set and the range of each variable, Perform random initialization, is the d-th dimension variable of individual np in the g-th generation population, np = 1, 2, ..., NP; d = 1, 2, ..., D; at the same time, the individual with the smallest fitness value is selected as the optimal individual in the population, denoted by x best =(xbest,1 ,x best,2 ,...,x best,D ); where the fitness value is obtained by the fitness function, fitness function f(x): f(x)=F(β,ω Sj ,ω Oj ,u hi ,s jh ).

[0043] Preferably, in step 5), the flood evolution simulation operation includes:

[0044] (1) Perform iterative update operations on all individuals in the population in step 4) to obtain mutant individuals At this time, 1≤np≤NP,1≤d≤D, the calculation formula is as follows:

[0045]

[0046] Among them, randn is a normally distributed random number; r1, r2, r3 and r4 are random numbers between 0 and 1; p d and q d are the upper and lower limits of the d-th dimension variable respectively; is the d-dimensional variable of the individual θ randomly selected from the g-th generation population, with θ∈[1,NP] and θ≠np; PK is the water consumption coefficient, and the calculation formula is as follows:

[0047]

[0048] PE np is the fitness coefficient value of individual np, and the calculation formula is as follows:

[0049]

[0050] Among them, f(x) is the individual fitness function;

[0051] (2) Comparison of individuals and The fitness of Then Otherwise, no processing will be done;

[0052] (3) After all individuals have completed the flood evolution simulation operation, update the optimal individual x best :

[0053]

[0054] Preferably, in step 5), the flood change simulation operation, i.e., simulating the evaporation attenuation of the flood, comprises the following steps:

[0055] (1) By sorting the updated population after the flood evolution simulation operation, NW poor individuals are sorted. Perform iterative update, NW takes 10% of NP, and obtains new variant individuals The calculation formula is as follows:

[0056]

[0057] (2) Directly using new variant individuals Replace the poor individuals in the population after the flood evolution simulation operation is updated That is

[0058] (3) Repeat steps (1) and (2) of the flood change simulation operation until all individuals have completed the flood change simulation operation and update the optimal individual x best .

[0059] Preferably, in step 5), the catfish effect update operation is to introduce the catfish effect strategy based on the standard flood optimization algorithm. best Simulate the evolution of the flood and find the best elite individuals in the updated population Perform linear fitting, implement random perturbations for iterative updates, and set the NE value to 10% of NP; for any elite individual The maximum number of catfish effect updates is k max ; The catfish effect update operation comprises the following steps:

[0060] (1) Let the catfish effect update number k = 1, generate a random vector have

[0061] (2) Through e k right and Perform cross-learning to obtain two catfish individuals and The calculation formula is:

[0062]

[0063] in, is a random number between -1 and 1, so that and It could be and The linear interpolation between them may also exceed the range between them, thus extending the new search space to avoid falling into the local optimum;

[0064] (3) Comparison of individuals and The fitness of Then Otherwise, no action is taken; if Then Otherwise, no processing will be done;

[0065] (4) Let k = k + 1 and use Chebyshev chaos mapping to generate the random variable for the next catfish effect update The calculation formula is as follows:

[0066]

[0067] (5) If k>k max , it means that the catfish effect update has been completed; otherwise, go to step b and use Continue to focus on elite individuals Implemented catfish effect update;

[0068] (6) After all elite individuals have completed the catfish effect update operation, update the optimal individual x best .

[0069] Preferably, in step 5), the teaching optimization and updating operation includes:

[0070] (1) After the catfish effect update operation is completed, all individuals are taught to optimize learning and update. x best and the population mean position Get new individuals The calculation formula is as follows:

[0071]

[0072] Among them, r7 is a random number between 0 and 1; r8 is a positive integer with a value of 1 or 2;

[0073] (2) Comparison of individuals and The fitness of Then Otherwise, no processing will be done;

[0074] (3) Repeat steps (1) and (2) of the teaching optimization update operation until all individuals have completed the teaching optimization learning operation and update the optimal individual x best .

[0075] Preferably, in step 7), according to the level characteristic value formula To determine the flood sample x i The level is determined as follows:

[0076] When 1≤H i When x is less than 1.5,i The classification level is 1; when When x i The classification level is here When c-0.5≤H i When ≤c, x i The classification level is c.

[0077] Preferably, when the number of flood categories c=3, the flood level is divided into 3 levels, and levels 1 to 3 are medium flood, large flood and extremely large flood respectively; the judgment criteria are as follows: when 1≤H i <1.5, the flood is judged as Level 1: medium flood; when 1.5≤H i <2.5, the flood is judged as Level 2: Major flood; when 2.5≤H i When ≤3.0, the flood is judged to be level 3: extremely severe flood;

[0078] Alternatively, when the number of flood categories c=4, the flood levels are divided into 4 levels, and the 1st to 4th levels are respectively divided into small flood, medium flood, large flood and extremely large flood; the judgment criteria are as follows: when 1≤H i <1.5, the flood is judged as Level 1: small flood; when 1.5≤H i <2.5, the flood is judged as Level 2: medium flood; when 2.5≤H i <3.5, the flood is judged to be level 3: major flood; when 3.5≤H i When ≤4.0, the flood is judged to be level 4: extremely severe flood;

[0079] Alternatively, when the number of flood categories c=5, the flood levels are divided into 5 levels, and the 1st to 5th levels are respectively divided into small flood, small and medium flood, medium flood, large flood and extremely large flood; the judgment criteria are as follows: when 1≤H i <1.5, the flood is judged as Level 1: small flood; when 1.5≤H i <2.5, the flood is judged as Level 2: small or medium flood; when 2.5≤H i <3.5, the flood is judged as Level 3: moderate flood; when 3.5≤H i <4.5, the flood is judged to be Level 4: major flood; when 4.5≤H i When ≤5.0, the flood is judged to be Level 5: extremely severe flood.

[0080] The present invention also provides a flood classification system based on a hybrid flood optimization algorithm, which is used to implement the flood classification method based on the hybrid flood optimization algorithm, comprising:

[0081] The initial condition setting module is used to collect sample set data of flood classification and determine the objective function and fuzzy clustering iteration formula of the combined weight fuzzy clustering iteration model of flood classification after normalization;

[0082] The population initialization operation module is used to determine the arrangement of decision variables, randomly initialize the population individuals, and compare and select the best individuals;

[0083] The population optimization module is used to update and optimize the initialized population through flood evolution simulation operation, flood change simulation operation, catfish effect update operation, and teaching optimization update operation to obtain the updated optimal individual;

[0084] The optimal individual output module is used to repeatedly execute the population optimization module until the iteration stop condition is met to obtain the optimal individual.

[0085] The flood category determination module is used to obtain the fuzzy clustering center matrix and combined weight vector, fuzzy clustering membership matrix of flood classification, and give the flood level evaluation results according to the level eigenvalues.

[0086] Preferably, the population optimization module includes:

[0087] The flood evolution simulation operation submodule is used to simulate the regular flood movement and perform iterative updates on all individuals;

[0088] The flood change simulation operation submodule is used to simulate the evaporation attenuation of floods and iteratively update several poor individuals;

[0089] The catfish effect update operation submodule is used to introduce the catfish effect learning strategy and iteratively update several elite individuals;

[0090] The teaching optimization update operation submodule is used to introduce the teaching optimization learning strategy, calculate the average position of the population, and iteratively update all individuals.

[0091] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory is used to store at least one program, and the processor is used to load the at least one program to execute the flood classification method based on the hybrid flood optimization algorithm.

[0092] Compared with the prior art, the present invention has the following beneficial effects:

[0093] The method of the present invention attempts to introduce the flood optimization method into the flood classification problem. At the same time, in order to improve its global search ability, a new hybrid flood optimization algorithm is proposed on the basis of identifying the modeling principle of the standard flood optimization algorithm, that is, the standard flood optimization algorithm is organically integrated with the catfish effect and the teaching optimization learning strategy, including introducing the catfish effect to learn and update the better individuals, and introducing the teaching optimization learning strategy to update the positions of all individuals in the population, so as to improve the global search ability of the optimization algorithm. There are three main advantages:

[0094] The present invention introduces the flood optimization algorithm into the flood classification problem for the first time, and has the advantages of clear principle, simple structure, fast optimization rate, etc.

[0095] The present invention organically integrates the standard flood optimization algorithm with the catfish effect and the teaching optimization learning strategy, including introducing the catfish effect to perform learning updates on better individuals and introducing the teaching optimization learning strategy to perform position updates on all individuals in the population. It can encode premature convergence phenomena, has strong optimization capability, fast convergence rate and high solution accuracy.

[0096] The hybrid flood optimization algorithm proposed in the present invention is organically nested with the flood classification combined weight fuzzy clustering iterative model, which effectively reduces the computational complexity of the model and improves the flood classification clustering effect, providing a scientific basis for flood level identification and flood control situation assessment, and can quickly evaluate flood levels, providing a scientific basis for flood level identification and flood control situation assessment under complex conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Figure 1 The present invention is a flow chart of a flood classification method based on a hybrid flood optimization algorithm. DETAILED DESCRIPTION

[0098] In order to better explain the present invention, the main contents of the present invention are further explained below in conjunction with the drawings and specific embodiments, but the contents of the present invention are not limited to the following embodiments.

[0099] like Figure 1 As shown, the present invention provides a flood classification method based on a hybrid flood optimization algorithm, comprising the following steps:

[0100] 1) Collect sample set data for flood classification.

[0101] There are n flood sample sets, and the i-th sample yi contains m indicators, denoted by y i =(y i1 ,y i2 ,...,y im ), i = 1, 2, ..., n, then the flood sample set is an n × m indicator feature matrix Y:

[0102]

[0103] Among them, y ij is the jth index of the ith sample, j = 1, 2, ..., m.

[0104] 2) Normalization of flood samples.

[0105] In order to eliminate the dimension effect, ij The normalization process is performed to obtain the normalized matrix R = (r ij ) n×m .

[0106] r ij =(y ij -y min (j)) / (y max (j)-y min (j));

[0107] Among them, r ij is the normalized value of index j of sample i, y max (j) and y min (j) are the maximum and minimum values ​​of index j, respectively.

[0108] 3) Establish a flood classification combined weight fuzzy clustering iterative model;

[0109] Set the fuzzy clustering membership matrix U = (u hi ) c×n 、Fuzzy cluster center matrix S=(s jh ) m×c and the combined weight vector ω=(ω1,ω2,...,ω m ), and construct the fuzzy clustering optimization objective function, including:

[0110] 3.1) Set up the fuzzy clustering membership matrix. Cluster analysis of the flood sample set is performed according to c categories to obtain the membership matrix:

[0111]

[0112] Among them, u hi is the membership degree of sample i to category h, i = 1, 2, ..., n; h = 1, 2, ..., c.

[0113] 3.2) Set the fuzzy cluster center matrix. Let the fuzzy cluster center of c categories be S:

[0114]

[0115] Among them, s jhis the cluster center of index j for category h, and the number of flood categories c ≥ 3.

[0116] 3.3) Set the combined weight vector. Add the subjective weight and objective weight of the flood sample index to obtain the combined weight ω=(ω1,ω2,...,ω m ):

[0117] ω j =β·ω Sj +(1-β)·ω Oj ,j=1,2,...,m;

[0118] Among them, the subjective and objective weights are ω S =(ω S1 ,ω S2 ,...,ω Sm ) and ω O =(ω O1 ,ω O2 ,...,ω Om );

[0119] ω m is the combined weight of the mth indicator, ω Sj is the subjective weight of the jth indicator, ω Oj is the objective weight of the jth indicator; β is the sensitivity coefficient of the main and objective weights, and β∈[0,1].

[0120] 3.4) Establish the fuzzy clustering optimization objective function. Combine the fuzzy clustering membership matrix, fuzzy clustering center matrix and combined weight vector to minimize the sum of squares of weighted generalized Euclidean weighted distances of all samples to all categories:

[0121]

[0122] According to the Lagrangian function method, solving the above formula can obtain the fuzzy clustering iteration formula:

[0123]

[0124]

[0125] 4) Population initialization and arrangement operations are performed to obtain the optimal individuals for initialization based on the flood classification combined weight fuzzy clustering iterative model.

[0126] 4.1) The hybrid flood optimization algorithm is used to optimize and solve the above flood classification combined weight fuzzy clustering iterative model. The population size of the hybrid flood optimization algorithm is set to NP and the maximum number of iterations is G. The population consists of NP D-dimensional individuals, and each individual Recorded as have Among them, np=1,2,...,NP, g=0,1,2,...,G, d=1,2,...,D, D=m×c+1;

[0127] 4.2) Select the fuzzy clustering center matrix (s jh ) m×c and sensitivity coefficient β are optimization variables, then the population individuals Arrange them as follows:

[0128]

[0129] That is to say, individual It is an m×c+1 dimensional variable, where the first m×c dimension is the cluster center code and the last dimension is the sensitivity coefficient β code. In other words, an individual represents the cluster center matrix and the sensitivity coefficient.

[0130] When the hybrid flood optimization algorithm is used to solve the optimization objective function of the flood classification combined weight fuzzy clustering iterative model, the fitness function is f(x)=F(β,ω Sj ,ω Oj ,u hi ,s jh ), and find the optimal solution that satisfies the minimum f(x) through iterative optimization.

[0131] 4.3) Let the number of iterations g = 0, and according to the flood sample set and the range of each variable, Perform random initialization, is the d-th dimension variable of individual np in the g-th generation population, np = 1, 2, ..., NP; d = 1, 2, ..., D; at the same time, the individual with the smallest fitness value is selected as the optimal individual in the population, denoted by x best =(x best,1 ,x best,2 ,...,x best,D ). The fitness value is calculated by the fitness function.

[0132] 5) Perform flood evolution simulation operation, flood change simulation operation, catfish effect update operation, and teaching optimization update operation in sequence.

[0133] 5.1) Flood evolution simulation operation, that is, simulating regular flood movement.

[0134] (1) Perform iterative update operations on all individuals to obtain mutant individuals Here 1≤np≤NP,1≤d≤D, the calculation formula is as follows:

[0135]

[0136] Among them, randn is a normally distributed random number; r1, r2, r3 and r4 are random numbers between 0 and 1; p d and q d are the upper and lower limits of the d-th dimension variable respectively; is the d-dimensional variable of the individual θ randomly selected from the g-th generation population, with θ∈[1,NP] and θ≠np; PK is the water consumption coefficient, and the calculation formula is as follows:

[0137]

[0138] PE np is the fitness coefficient value of individual np, and the calculation formula is as follows:

[0139]

[0140] Among them, f() is the individual fitness function;

[0141] (2) Comparison of individuals and The fitness of Then Otherwise, no processing will be done;

[0142] (3) After all individuals have completed the flood evolution simulation operation, update the optimal individual x best :

[0143]

[0144] 5.2) Flood change simulation operation, that is, simulating the evaporation attenuation of floods.

[0145] (1) By sorting the updated population after the flood evolution simulation operation, NW poor individuals are sorted. Perform iterative updates, and the value of NW is generally 10% of NP to obtain new variant individuals The calculation formula is as follows:

[0146]

[0147] (2) Directly using new variant individuals To replace the poor individuals in the population after the flood evolution simulation operation is updated That is

[0148] (3) Repeat steps (1) and (2) of the flood change simulation operation until all individuals have completed the flood change simulation operation and update the optimal individual x best .

[0149] 5.3) Catfish effect update operation, that is, on the basis of the standard flood optimization algorithm, the catfish effect strategy is introduced based on x best Simulate the evolution of the flood and find the best elite individuals in the updated population Perform linear fitting and implement random perturbations for iterative updates. The NE value is generally taken as about 10% of NP. The maximum number of catfish effect updates is k max ; The catfish effect update operation comprises the following steps:

[0150] (1) Let the catfish effect update number k = 1, generate a random vector have

[0151] (2) Through e k right and Perform cross-learning to obtain two catfish individuals and The calculation formula is:

[0152]

[0153] in, is a random number between -1 and 1, so that and It could be and The linear interpolation between them may also exceed the range between them, thus extending the new search space to avoid falling into the local optimum;

[0154] (3) Comparison of individuals and The fitness of Then Otherwise, no action is taken; if Then Otherwise, no processing will be done;

[0155] (4) Let k = k + 1 and use Chebyshev chaos mapping to generate the random variable for the next catfish effect update The calculation formula is as follows:

[0156]

[0157] (5) If k>k max , it means that the catfish effect update has been completed; otherwise, go to step b and use Continue to focus on elite individuals Implemented catfish effect update;

[0158] (6) After all elite individuals have completed the catfish effect update operation, update the optimal individual x best .

[0159] 5.4) Teaching optimization update operation, that is, on the basis of the standard flood optimization algorithm, in order to promote the position update of the entire group, on the basis of the aforementioned updates for worse individuals and better individuals, a teaching optimization learning strategy is further introduced.

[0160] (1) After the catfish effect update operation is completed, all individuals are taught to optimize learning and update. x best and the population mean position Get new individuals The calculation formula is as follows:

[0161]

[0162] Among them, r7 is a random number between 0 and 1; r8 is a positive integer with a value of 1 or 2;

[0163] (2) Comparison of individuals and The fitness of Then Otherwise, no processing will be done;

[0164] (3) Repeat steps (1) and (2) of the teaching optimization update operation until all individuals have completed the teaching optimization learning operation and update the optimal individual x best .

[0165] 6) Termination condition judgment operation, that is, judging whether the current number of iterations reaches the maximum number of iterations. If g≤G, set g=g+1 and repeat step 5); otherwise, stop iteration and output the optimal individual.

[0166] 7) According to the optimal individual, the fuzzy clustering center matrix and combined weight vector of flood classification are obtained, and the fuzzy clustering membership matrix of flood classification is obtained according to the fuzzy clustering iteration formula. Finally, according to the level eigenvalue formula To determine the flood sample x i level.

[0167] When 1≤H i When x is less than 1.5, i The classification level is 1; when When x i The classification level is here When c-0.5≤H i When ≤c, x i The classification level is c; c is the number of flood categories, and c is generally taken as 3, 4 or 5.

[0168] When the number of flood categories c is 3, the flood levels are divided into extremely large floods, major floods and moderate floods; the judgment criteria are: when the level characteristic value formula value falls within [2.5,3.0], the flood is judged to be an extremely large flood; when the level characteristic value formula value falls within [1.5,2.5), the flood is judged to be a major flood; when the level characteristic value formula value falls within [1.0,1.5), the flood is judged to be a moderate flood.

[0169] When the number of flood categories c is 4, the flood levels are divided into extremely large floods, large floods, medium floods and small floods; the discrimination criteria are: when the level characteristic value formula value falls into [3.5,4.0], the flood is judged to be an extremely large flood; when the level characteristic value formula value falls into [2.5,3.5), the flood is judged to be a large flood; when the level characteristic value formula value falls into [1.5,2.5), the flood is judged to be a medium flood; when the level characteristic value formula value falls into [1.0,1.5), the flood is judged to be a small flood.

[0170] When the number of flood categories c is 5, the flood levels are divided into extremely large floods, large floods, moderate floods, small and medium floods, and small floods; the discrimination criteria are: when the level characteristic value formula value falls within [4.5,5.0], the flood is judged to be an extremely large flood; when the level characteristic value formula value falls within [3.5,4.5), the flood is judged to be a large flood; when the level characteristic value formula value falls within [2.5,3.5), the flood is judged to be a moderate flood; when the level characteristic value formula value falls within [1.5,2.5), the flood is judged to be a small and medium flood; when the level characteristic value formula value falls within [1.0,1.5), the flood is judged to be a small flood.

[0171] The present invention also provides a flood classification system based on a hybrid flood optimization algorithm, which is used to implement the flood classification method based on the hybrid flood optimization algorithm, comprising:

[0172] The initial condition setting module is used to collect sample set data of flood classification and determine the objective function and iteration formula of the combined weight fuzzy clustering iteration model of flood classification after normalization;

[0173] The population initialization operation module is used to determine the arrangement of decision variables, randomly initialize the population individuals, and compare and select the best individuals;

[0174] The population optimization module is used to update and optimize the initialized population to obtain the updated optimal individuals; the population optimization module includes: a flood evolution simulation operation submodule, which is used to simulate conventional flood movement and perform iterative updates on all individuals; a flood change simulation operation submodule, which is used to simulate the evaporation attenuation of floods and iteratively update several poor individuals; a catfish effect update operation submodule, which is used to introduce a catfish effect learning strategy and iteratively update several elite individuals; and a teaching optimization update operation submodule, which is used to introduce a teaching optimization learning strategy, calculate the average position of the population, and iteratively update all individuals.

[0175] The optimal individual output module is used to repeatedly execute the population optimization module until the iteration stop condition is met to obtain the optimal individual;

[0176] The flood category determination module is used to obtain the fuzzy clustering center matrix and combined weight vector, fuzzy clustering membership matrix of flood classification, and give the flood level evaluation results according to the level eigenvalues.

[0177] The present invention also provides a computer device, including a memory and a processor, wherein the memory is used to store at least one program, and the processor is used to load the at least one program to execute the above-mentioned flood classification method based on the hybrid flood optimization algorithm.

[0178] Next, taking the flood classification problem of a certain hydrological station as an example, the technical solution of the present invention is further explained.

[0179] According to historical data, the basin where the hydrological station is located has experienced many floods. The following 10 historical flood samples were collected for flood level evaluation (see Table 1), and the flood samples were divided into three categories: extremely large floods, large floods and moderate floods.

[0180] Table 1: 10 historical flood samples of a hydrological station (partial data)

[0181]

[0182]

[0183] The flood classification evaluation index includes five evaluation indicators: peak water level (A1), number of days when the flood level exceeds 9m (A2), peak flow (A3), flood volume from May to September (A4), and flow duration (A5). The subjective weights of the five evaluation indicators are 0.16, 0.25, 0.19, 0.15, and 0.25, and the objective weights are 0.17, 0.23, 0.20, 0.19, and 0.21. It can be seen that there are 10 flood sample sets, each of which contains 5 indicators, and the number of flood categories is 3, that is, n = 10, m = 5, and c = 3.

[0184] The method of the present invention, differential evolution algorithm and flood optimization algorithm are respectively used to optimize the combined weight fuzzy clustering iterative model of flood classification, and the fuzzy clustering center matrix and combined weight sensitivity coefficient are used for encoding. The optimization variable dimension is 16, and the minimum sum of squares of weighted generalized Euclidean weight distances of all samples to all categories is the optimization objective function, and iterative optimization calculation is implemented. The specific calculation results of the three algorithms are shown in Table 2.

[0185] Table 2: Optimization calculation results of different algorithms

[0186]

[0187] It can be seen from Table 2 that compared with the differential evolution algorithm and the standard flood optimization algorithm, the objective function value of the method of the present invention is the smallest, indicating that the improved strategy proposed this time is reasonable and feasible, with strong global search capability, and provides a feasible way for rapid evaluation of flood classification.

[0188] Based on the method of the present invention, the sensitivity coefficient β and the combination weight ω under the optimal condition are obtained. The optimal fuzzy clustering matrix U and the fuzzy clustering center matrix S are as follows:

[0189] β=0.5594

[0190] ω=(0.1663,0.2420,0.1900,0.1718,0.2300)

[0191]

[0192] According to the characteristic value formula, the level characteristic value H of 10 flood samples can be obtained: i :

[0193] H i =(1.07,2.94,2.75,2.90,2.05,2.54,2.90,2.11,2.02,1.19)

[0194] Finally, the flood samples were classified into medium flood, extremely large flood, extremely large flood, extremely large flood, large flood, extremely large flood, large flood, and medium flood. The method of the invention can quickly evaluate flood levels and provide a scientific basis for flood level identification and flood control situation assessment under complex conditions. It also has good application value in sorting, evaluation and decision-making problems without evaluation criteria.

[0195] In summary, the method of the present invention attempts to introduce the flood optimization method into the flood classification problem, and on the basis of the standard flood optimization algorithm, organically integrates the catfish effect and the teaching optimization learning strategy, and proposes a new hybrid flood optimization algorithm with significant search capability and solution accuracy. It has three main advantages: first, it retains the characteristics of simplicity and fast calculation of the standard flood optimization algorithm; second, it introduces the catfish effect and the teaching optimization learning strategy, which can encode the premature convergence phenomenon; third, the proposed algorithm is organically nested with the flood classification fuzzy clustering model, which reduces the computational complexity and improves the flood classification clustering effect.

Claims

1. A flood classification method based on a hybrid flood optimization algorithm, characterized by: The following steps are involved: 1) Collect sample data for flood classification; 2) The sample set data is normalized; 3) Based on the normalized sample data set, a flood classification combined weight fuzzy clustering iterative model is established, including setting the fuzzy clustering membership matrix U = (u hi ) c×n 、Fuzzy cluster center matrix S=(s jh ) m×c and the combined weight vector ω=(ω1,ω2,...,ω m ), and construct the fuzzy clustering optimization objective function; 4) Population initialization and arrangement operation, based on the flood classification combined weight fuzzy clustering iteration model, to obtain the optimal individual for initialization; 5) Execute flood evolution simulation operation, flood change simulation operation, catfish effect update operation, and teaching optimization update operation in sequence to continuously update the optimal individual; 6) Termination condition judgment operation, that is, judging whether the current number of iterations reaches the maximum number of iterations. If g≤G, set g=g+1 and repeat step 5); otherwise, stop iteration and output the optimal individual; 7) According to the optimal individual, the fuzzy clustering center matrix and combined weight vector of flood classification are obtained, and the fuzzy clustering membership matrix of flood classification is obtained according to the fuzzy clustering iteration formula. Finally, according to the level eigenvalue formula To determine the flood sample x i level.

2. The flood classification method according to claim 1, characterized in that: In step 1), the flood sample set is an n×m indicator feature matrix Y: Where n is the number of flood sample sets, m is the sample y i The number of indicators included, y i is the i-th sample, i=1, 2, ..., n, y i =(y i1 ,y i2 ,...,y im );y ij is the jth index of the ith sample, j = 1, 2, ..., m.

3. The flood classification method according to claim 1, characterized in that: The indicators are statistical data describing floods, including peak water level, peak flow, flood duration and total flood volume.

4. The flood classification method according to claim 2, characterized in that: In step 2), the flood sample y i The normalized matrix R after normalization is (r ij ) n×m : r ij =(y ij -y min (j)) / (y max (j)-y min (j)); Among them, r ij is the normalized value of index j of sample i, y max (j) and y min (j) are the maximum and minimum values ​​of index j, respectively.

5. The flood classification method according to claim 1, characterized in that: In step 3), the fuzzy clustering membership matrix U is as follows: Among them, u hi is the membership degree of sample i belonging to category h, i = 1, 2, ..., n; h = 1, 2, ..., c; c is the number of categories for cluster analysis of flood sample set, and the number of flood categories c ≥ 3.

6. The flood classification method according to claim 5, characterized in that: In step 3), the fuzzy cluster center matrix S is as follows: Among them, s jh is the cluster center of index j for category h.

7. The flood classification method according to claim 1, characterized in that: In step 3), the combined weight is obtained by adding the subjective weight and the objective weight of the flood sample index; the combined weight ω=(ω1,ω2,...,ω m ),,in: oh j =b·ω Sj +(1-β)·ω Oj ,j=1,2,...,m; The subjective and objective weights are ω S =(ω S1 ,ω S2 ,...,ω Sm ) and ω O =(ω O1 ,ω O2 ,...,ω Om ); ω m is the combined weight of the mth indicator, ω Sj is the subjective weight of the jth indicator, ω Oj is the objective weight of the jth indicator; β is the sensitivity coefficient of the main and objective weights, and β∈[0,1].

8. The flood classification method according to claim 1, characterized in that: In step 3), the fuzzy clustering optimization objective function is as follows: According to the Lagrangian function method, solving the above formula can obtain the following fuzzy clustering iteration formula:

9. The flood classification method according to claim 1, characterized in that: The step 4) comprises: (1) Set the population size of the hybrid flood optimization algorithm to NP and the maximum number of iterations to G. The population consists of NP D-dimensional individuals, and each individual Recorded as have Among them, np=1,2,...,NP, g=0,1,2,...,G, d=1,2,...,D, D=m×c+1; (2) Select the fuzzy cluster center matrix (s jh ) m×c and sensitivity coefficient β are optimization variables, then the population individuals Arrange them as follows: Among them, individual It is an m×c+1 dimensional variable, the first m×c dimensions are the cluster center code, and the last dimension is the sensitivity coefficient β code; (3) Let the number of iterations g = 0, and according to the flood sample set and the range of each variable, Perform random initialization, is the d-th dimension variable of individual np in the g-th generation population, np=1,2,...,NP;d=1,2,...,D;At the same time, the individual with the smallest fitness value is selected as the optimal individual of the population, denoted by x best =(x best,1 ,x best,2 ,...,x best,D ); where the fitness value is obtained by the fitness function, fitness function f(x): f(x)=F(β,ω Sj ,ω Oj ,u hi ,s jh ).

10. The flood classification method according to claim 9, characterized in that: In step 5), the flood evolution simulation operation includes: (1) Perform iterative update operations on all individuals in the population in step 4) to obtain mutant individuals At this time, 1≤np≤NP,1≤d≤D, the calculation formula is as follows: Among them, randn is a normally distributed random number; r1, r2, r3 and r4 are random numbers between 0 and 1; p d and q d are the upper and lower limits of the d-th dimension variable respectively; is the d-dimensional variable of the individual θ randomly selected from the g-th generation population, with θ∈[1,NP] and θ≠np; PK is the water consumption coefficient, and the calculation formula is as follows: PE np is the fitness coefficient value of individual np, and the calculation formula is as follows: Among them, f(x) is the individual fitness function; (2) Comparison of individuals and The fitness of Then Otherwise, no processing will be done; (3) After all individuals have completed the flood evolution simulation operation, update the optimal individual x best :

11. The flood classification method according to claim 1, characterized in that: In step 5), the flood change simulation operation, i.e., simulating the evaporation attenuation of the flood, comprises the following steps: (1) By sorting the updated population after the flood evolution simulation operation, NW poor individuals are sorted. Perform iterative update, NW takes 10% of NP, and obtains new variant individuals The calculation formula is as follows: (2) Directly using new variant individuals Replace the poor individuals in the population after the flood evolution simulation operation is updated That is (3) Repeat steps (1) and (2) of the flood change simulation operation until all individuals have completed the flood change simulation operation and update the optimal individual x best .

12. The flood classification method according to claim 1, characterized in that: In step 5), the catfish effect update operation is to introduce the catfish effect strategy based on the standard flood optimization algorithm. best Simulate the evolution of the flood and find the best elite individuals in the updated population Perform linear fitting, implement random perturbations for iterative updates, and set the NE value to 10% of NP; for any elite individual The maximum number of catfish effect updates is k max ; The catfish effect update operation comprises the following steps: (1) Let the catfish effect update number k = 1, generate a random vector have (2) Through e k right and Perform cross-learning to obtain two catfish individuals and The calculation formula is: in, is a random number between -1 and 1, so that and It could be and The linear interpolation between them may also exceed the range between them, thus extending the new search space to avoid falling into the local optimum; (3) Comparison of individuals and The fitness of Then Otherwise, no action is taken; if Then Otherwise, no processing will be done; (4) Let k = k + 1 and use Chebyshev chaos mapping to generate the random variable for the next catfish effect update The calculation formula is as follows: (5) If k>k max , it means that the catfish effect update has been completed; otherwise, go to step b and use Continue to focus on elite individuals Implemented catfish effect update; (6) After all elite individuals have completed the catfish effect update operation, update the optimal individual x best .

13. The flood classification method according to claim 1, characterized in that: In the step 5), the teaching optimization and updating operation includes: (1) After the catfish effect update operation is completed, all individuals are taught to optimize learning and update. x best and the population mean position Get new individuals The calculation formula is as follows: Among them, r7 is a random number between 0 and 1; r8 is a positive integer with a value of 1 or 2; (2) Comparison of individuals and The fitness of Then Otherwise, no processing will be done; (3) Repeat steps (1) and (2) of the teaching optimization update operation until all individuals have completed the teaching optimization learning operation and update the optimal individual x best .

14. The flood classification method according to claim 5, characterized in that: In step 7), according to the level characteristic value formula To determine the flood sample x i The level is determined as follows: When 1≤H i When x is less than 1.5, i The classification level is 1; when When x i The classification level is here When c-0.5≤H i When ≤c, x i The classification level is c.

15. The flood classification method according to claim 14, characterized in that: When the number of flood categories c=3, the flood level is divided into 3 levels, and levels 1 to 3 are medium flood, large flood and extremely large flood respectively; the judgment criteria are as follows: when 1≤H i <1.5, the flood is judged as Level 1: medium flood; when 1.5≤H i <2.5, the flood is judged as Level 2: Major flood; when 2.5≤H i When ≤3.0, the flood is judged to be level 3: extremely severe flood; Alternatively, when the number of flood categories c=4, the flood levels are divided into 4 levels, and the 1st to 4th levels are respectively divided into small flood, medium flood, large flood and extremely large flood; the judgment criteria are as follows: when 1≤H i When 1.5≤H i <2.5, the flood is judged as Level 2: medium flood; when 2.5≤H i <3.5, the flood is judged to be level 3: major flood; when 3.5≤H i When ≤4.0, the flood is judged to be level 4: extremely severe flood; Alternatively, when the number of flood categories c=5, the flood levels are divided into 5 levels, and the 1st to 5th levels are respectively divided into small flood, small and medium flood, medium flood, large flood and extremely large flood; the judgment criteria are as follows: when 1≤H i <1.5, the flood is judged as Level 1: small flood; when 1.5≤H i <2.5, the flood is judged as Level 2: small or medium flood; when 2.5≤H i <3.5, the flood is judged as Level 3: moderate flood; when 3.5≤H i <4.5, the flood is judged to be Level 4: major flood; when 4.5≤H i When ≤5.0, the flood is judged to be Level 5: extremely severe flood.

16. A flood classification system based on a hybrid flood optimization algorithm, characterized by: The flood classification method based on the hybrid flood optimization algorithm for implementing any one of claims 1 to 15 comprises: The initial condition setting module is used to collect sample set data of flood classification and determine the objective function and fuzzy clustering iteration formula of the combined weight fuzzy clustering iteration model of flood classification after normalization; The population initialization operation module is used to determine the arrangement of decision variables, randomly initialize the population individuals, and compare and select the best individuals; The population optimization module is used to update and optimize the initialized population through flood evolution simulation operation, flood change simulation operation, catfish effect update operation, and teaching optimization update operation to obtain the updated optimal individual; The optimal individual output module is used to repeatedly execute the population optimization module until the iteration stop condition is met to obtain the optimal individual. The flood category determination module is used to obtain the fuzzy clustering center matrix and combined weight vector, fuzzy clustering membership matrix of flood classification, and give the flood level evaluation results according to the level eigenvalues.

17. The flood classification system according to claim 16, characterized in that: The population optimization module includes: The flood evolution simulation operation submodule is used to simulate the regular flood movement and perform iterative updates on all individuals; The flood change simulation operation submodule is used to simulate the evaporation attenuation of floods and iteratively update several poor individuals; The catfish effect update operation submodule is used to introduce the catfish effect learning strategy and iteratively update several elite individuals; The teaching optimization update operation submodule is used to introduce the teaching optimization learning strategy, calculate the average position of the population, and iteratively update all individuals.

18. A computer device, characterized in that: It comprises a memory and a processor, wherein the memory is used to store at least one program, and the processor is used to load the at least one program to execute the flood classification method based on the hybrid flood optimization algorithm as described in any one of claims 1 to 15.

Images