阅读说明：本技术 基于线性判别分析和ialo-elm的中长期径流预测方法及系统 (Medium-and-long-term runoff prediction method and system based on linear discriminant analysis and IALO-ELM ) 是由 彭甜 花磊 张楚 孙娜 成佳伟 吴健 于 2021-05-28 设计创作，主要内容包括：本发明公开了一种基于线性判别分析和IALO-ELM的中长期径流预报方法及系统,所述方法包括：(1)获取某水文站的径流数据并对数据进行归一化处理,运用线性判别分析法对径流数据进行预报因子选择；(2)运用拉丁超立方对蚁狮种群初始化；(3)建立基于改进蚁狮算法(IALO)的优化极限学习机模型；(4)将蚁狮原始种群分为p等分,运用多种群方法进行交互和预选机制两两对比分析出全局最优路径；(5)将Fuch映射引入蚁狮算法中,对局部最优解的邻域进行混沌遍历搜索。本发明利用改进蚁狮算法与极限学习机相结合,大大提高了径流预测准确性和预测精度的可靠性。(The invention discloses a medium and long term runoff forecasting method and system based on linear discriminant analysis and IALO-ELM, wherein the method comprises the following steps: (1) acquiring runoff data of a hydrological station, normalizing the data, and selecting a forecasting factor for the runoff data by using a linear discriminant analysis method; (2) initializing ant lion populations by using Latin hypercube; (3) establishing an optimized extreme learning machine model based on an improved ant lion algorithm (IALO); (4) dividing the original ant lion population into p equal parts, and analyzing a global optimal path by pairwise comparison of interaction and a preselection mechanism by using a multi-population method; (5) and introducing Fuch mapping into the ant lion algorithm, and performing chaotic traversal search on the neighborhood of the local optimal solution. The method combines the improved ant lion algorithm with the extreme learning machine, and greatly improves the runoff prediction accuracy and the reliability of the prediction precision.)

1. A medium-long term runoff prediction method based on linear discriminant analysis and IALO-ELM is characterized by comprising the following steps:

(1) acquiring runoff data in advance, normalizing the data, and selecting a forecasting factor for initial runoff data by using a linear discriminant analysis method;

(2) initializing an initial ant lion population by using a Latin hypercube, randomly initializing ant lion positions, and starting iterative search;

(3) dividing the original ant lion population into p equal parts, and analyzing a global optimal path by pairwise comparison of interaction and a preselection mechanism by using a multi-population method;

(4) introducing Fuch mapping into the ant lion algorithm, and performing chaotic traversal search on the neighborhood of the local optimal solution;

(5) establishing an extreme learning machine model based on the optimization of the improved ant lion algorithm, and optimizing the connection weight and the threshold of the extreme learning machine by using the improved ant lion algorithm; and predicting the test sample by adopting the trained IALO-ELM model.

2. The linear discriminant analysis and IALO-ELM based method for medium and long term runoff prediction according to claim 1, wherein the step (1) comprises the steps of:

(11) converting the acquired flow volume data set into a scatter matrix, and performing data encapsulation:

S_i＝∑(x-m_i)(x-m_i)^T,i＝1,2

wherein S is_iTo compute the intra-class dispersion matrix, x is the sample data, m_iAs a mean vector of data samples, S_wIs a scatter matrix;

(12) computing eigenvectors of the scatter matrix and corresponding eigenvalues, wherein the formula for computing eigenvalues is (λ E-S)_W) 0, λ is eigenvalue, E is identity matrix;

(13) sorting the eigenvectors in descending order according to the magnitude of the eigenvalue, selecting the eigenvectors corresponding to the first n largest eigenvalues, and establishing a matrix, namely each row is an eigenvector;

(14) and transforming the sample to a new subspace by using the eigenvector matrix to complete the final forecast factor selection.

3. The linear discriminant analysis and IALO-ELM based method for medium and long term runoff prediction according to claim 1, wherein the step (2) comprises the steps of:

(21) constructing an n-dimensional unit cube [0,1] from the number of populations (n)]ⁿ；

(22) According to the initial population number, each one-dimensional coordinate interval [0,1]]Are all divided into N_aDividing the N into equal parts_aThe equal parts are randomly arranged, and assume that N is_aThe random arrangements are independent of each other;

(23) to obtain an N_aRandom matrix of x m dimensions, will yield N_aAs an initial population.

4. The linear discriminant analysis and IALO-ELM based method for medium and long term runoff prediction according to claim 1, wherein said step (3) comprises the steps of:

(31) ant lion populationNP represents the size of the population; let the number of ant lion individuals in each sub-population be c, and let the ant lion individuals in the sub-population express that X ═ X_iI 1,2, …, c, the position of the ith lion in the search space is X_i＝(X_i1，X_i2，…，X_ic) Dividing the population into two parts, testing one part of the two parts by using a local search mode, calculating the fitness value of the current population individual, finding out excellent individuals and updating the position; inserting another kind of population data into the population data to perform data test; initializing each sub-population, and selecting an elite population by a cross variation method;

(32) initializing the positions of ants and ant lions, calculating and sorting the corresponding fitness values, and selecting the fitness valuesThe ant lion with the maximum medium fitness value is taken as the elite ant lion individual; setting the maximum number of iterations to T_mThe dimension of the fitness function is dim, and the numbers of ants and ant lions are N respectively_B、N_YSelecting variable control ranges as ub and lb respectively, wherein ub is an upper limit of a decision variable, and lb is a lower limit of the decision variable;

(33) c, updating the ant lion by using a roulette method and a greedy selection method^t _iAnd d^t _iThe value of (c):

wherein the content of the first and second substances,represents the maximum and minimum values of the ith ant in the t iteration, d_i、c_iRepresents the maximum value and the minimum value of the displacement of the ith ant; d^t、c^tThe maximum value and the minimum value of the t-th iteration displacement of all ants are shown to be reduced along with the increase of the iteration times;represents a normalized displacement;representing the random displacement of the ith ant for the t iteration.

5. The method for predicting medium and long term runoff based on linear discriminant analysis and IALO-ELM of claim 1 wherein said step (4) comprises the steps of:

(41) comparing the ants after the migration process with the ant lion at the best position, and readjusting the position of the best ant lion; generating new individuals near the optimal ant lion through Fuch mapping search; fuch chaotic mapping function expression X_a+1＝cos(1/X_a) Fuch mapping is a novel one-dimensional discrete mapping, a chaotic sequence is mapped into a solution space to obtain a population X, and the initialized population individual of the chaos is X_ia：

X_ia＝lb+(ub-lb)·X_n+1

Wherein, X_iaThe d-dimensional value of the individual of the ith ant colony is shown, and lb and ub are the upper and lower bounds of the d-dimensional value respectively; meanwhile, an elite reverse learning strategy is introduced, population number is initialized to calculate an elite reverse population, and the population number and the elite reverse population are combined into a new population to calculate a new fitness value;

(42) the displacement updating formula of the optimized ant individual is as follows:

wherein the content of the first and second substances,the position of the ith ant at the t iteration is shown;ant lion individual-based shifts selected for the t-th iterative roulette;displacement based on the elite individual for the t-th iteration;after the ant position is updated by adding reverse regulation factor, the correspondent adaptability value is obtained, when the adaptability value of ant is superior to that of ant lion, the correspondent ant position is usedThe position for replacing the ant lion.

6. A middle-long term runoff prediction system based on linear discriminant analysis IALO-ELM adopting the method as claimed in any one of claims 1 to 5, comprising a data preprocessing module, a forecasting factor selection module, a runoff prediction module, a ant lion algorithm improvement module, an extreme learning machine model optimization module and a forecasting value output module;

the data preprocessing module is mainly used for preprocessing the obtained runoff data to finally obtain a runoff sequence;

the forecasting factor selection module is mainly used for selecting the forecasting factors of the radial data by adopting a linear discriminant analysis method;

the runoff prediction module mainly comprises an ant lion algorithm improvement unit, an extreme learning machine model optimization unit and a prediction value output unit;

the runoff ant lion algorithm improvement module initializes the ant lion population by adopting Latin hypercube; performing cross variation on the ant lion algorithm by adopting a plurality of groups; comparing the wandered ants with the ant lion with the best current position by Fuch mapping, and readjusting the position of the best ant lion;

the extreme learning machine model optimization module is used for optimizing parameters of the extreme learning machine by adopting training sample data and an improved ant lion algorithm to obtain optimal parameters and obtain an optimized extreme learning machine model;

and the prediction value output module optimizes the extreme learning machine model by using the ant lion algorithm to obtain the optimal parameters, and then inputs the optimal parameters and prediction data into the extreme learning machine model to obtain a final prediction result.

Technical Field

The invention belongs to the field of hydrological time series prediction, and relates to medium and long term runoff prediction and a system based on linear discriminant analysis IALO-ELM.

Background

The long-term runoff prediction in rivers is the main content of hydrological prediction, and the high-precision runoff prediction is beneficial to reservoir optimal scheduling, flood control and drought resistance, reasonable formulation of hydropower station power generation plans and guarantee of safety in production and life of the hydropower stations. Therefore, it is very important to make accurate hydrologic predictions, and many methods are proposed for middle-and long-term hydrologic predictions from different directions and in combination with corresponding subject knowledge in order to improve the accuracy and reliability of hydrologic predictions, especially middle-and long-term hydrologic predictions.

In recent years, artificial intelligence methods have been widely used in various fields. While the traditional artificial intelligence techniques, such as BP neural network, bayesian network, fuzzy neural network, recurrent neural network, etc., are widely used in medium and long term runoff prediction, the traditional neural network is easy to fall into local optimality during data training due to the insufficient analysis capability of the data, so that the final result does not achieve the expected effect.

The rapid development of the extreme learning machine provides new possibility for the prediction of the runoff sequence. The extreme learning machine has the characteristics of autonomous learning, high learning speed and the like, but the initial weight and threshold are randomly selected, so that the network performance cannot be optimal. In order to solve the problems, a medium-and-long-term runoff prediction model is established by the ant lion algorithm and the extreme learning machine, the global search capability is improved, and the prediction accuracy is greatly improved through parameter optimization.

Disclosure of Invention

The purpose of the invention is as follows: in order to solve the technical problems, the invention provides the medium-and-long-term runoff prediction and system based on the linear discriminant analysis IALO-ELM, which has strong generalization capability and high prediction precision.

The technical scheme is as follows: the invention provides a medium-and-long-term runoff forecasting method based on linear discriminant analysis and IALO-ELM, which specifically comprises the following steps:

(1) acquiring runoff data in advance, normalizing the data, and selecting a forecasting factor for initial runoff data by using a linear discriminant analysis method;

(2) initializing an initial ant lion population by using a Latin hypercube, randomly initializing ant lion positions, and starting iterative search;

(3) dividing the original ant lion population into p equal parts, and analyzing a global optimal path by pairwise comparison of interaction and a preselection mechanism by using a multi-population method;

(4) introducing Fuch mapping into the ant lion algorithm, and performing chaotic traversal search on the neighborhood of the local optimal solution;

(5) establishing an extreme learning machine model based on the optimization of the improved ant lion algorithm, and optimizing the connection weight and the threshold of the extreme learning machine by using the improved ant lion algorithm; and predicting the test sample by adopting the trained IALO-ELM model.

Further, the step (1) includes the steps of:

(11) converting the acquired flow volume data set into a scatter matrix, and performing data encapsulation:

S_i＝∑(x-m_i)(x-m_i)^T,i＝1,2

wherein S is_iTo compute the intra-class dispersion matrix, x is the sample data, m_iAs a mean vector of data samples, S_wIs a scatter matrix;

(12) computing a scatter matrixAnd corresponding eigenvalues, wherein the formula for computing the eigenvalues is (λ E-S)_W) 0, λ is eigenvalue, E is identity matrix;

(13) sorting the eigenvectors in descending order according to the magnitude of the eigenvalue, selecting the eigenvectors corresponding to the first n largest eigenvalues, and establishing a matrix, namely each row is an eigenvector;

(14) and transforming the sample to a new subspace by using the eigenvector matrix to complete the final forecast factor selection.

Further, the step (2) comprises the steps of:

(21) constructing an n-dimensional unit cube [0,1] from the number of populations (n)]ⁿ；

(22) According to the initial population number, each one-dimensional coordinate interval [0,1]]Are all divided into N_aDividing the N into equal parts_aThe equal parts are randomly arranged, and assume that N is_aThe random arrangements are independent of each other;

(23) to obtain an N_aRandom matrix of x m dimensions, will yield N_aAs an initial population.

Further, the step (3) includes the steps of:

(31) ant lion populationNP represents the size of the population; let the number of ant lion individuals in each sub-population be c, and let the ant lion individuals in the sub-population express that X ═ X_iI 1,2, …, c, the position of the ith lion in the search space is X_i＝(X_i1，X_i2，…，X_ic) Dividing the population into two parts, testing one part of the two parts by using a local search mode, calculating the fitness value of the current population individual, finding out excellent individuals and updating the position; inserting another kind of population data into the population data to perform data test; initializing each sub-population, and selecting an elite population by a cross variation method;

(32) initializing the positions of ants and ant lions, calculating and sequencing corresponding fitness values, and selecting the ant lions with the maximum fitness value to be used as the ant lionsIs an Elite lion individual; setting the maximum number of iterations to T_mThe dimension of the fitness function is dim, and the numbers of ants and ant lions are N respectively_B、N_YSelecting variable control ranges as ub and lb respectively, wherein ub is an upper limit of a decision variable, and lb is a lower limit of the decision variable;

(33) c, updating the ant lion by using a roulette method and a greedy selection method^t _iAnd d^t _iThe value of (c):

wherein the content of the first and second substances,represents the maximum and minimum values of the ith ant in the t iteration, d_i、c_iRepresents the maximum value and the minimum value of the displacement of the ith ant; d^t、c^tThe maximum value and the minimum value of the t-th iteration displacement of all ants are shown to be reduced along with the increase of the iteration times;represents a normalized displacement;representing the random displacement of the ith ant for the t iteration.

Further, the step (4) comprises the steps of:

(41) comparing the ants after the migration process with the ant lion at the best position, and readjusting the position of the best ant lion; generating new individuals near the optimal ant lion through Fuch mapping search; fuch chaotic mapping function expression X_a+1＝cos(1/X_a) Fuch mapping is a novel one-dimensional discrete mapping, a chaotic sequence is mapped into a solution space to obtain a population X, and the initialized population individual of the chaos is X_ia：

X_ia＝lb+(ub-lb)·X_n+1

Wherein, X_iaThe d-dimensional value of the individual of the ith ant colony is shown, and lb and ub are the upper and lower bounds of the d-dimensional value respectively; meanwhile, an elite reverse learning strategy is introduced, population number is initialized to calculate an elite reverse population, and the population number and the elite reverse population are combined into a new population to calculate a new fitness value;

(42) the displacement updating formula of the optimized ant individual is as follows:

wherein the content of the first and second substances,the position of the ith ant at the t iteration is shown;ant lion individual-based shifts selected for the t-th iterative roulette;displacement based on the elite individual for the t-th iteration;and when the adaptability value of the ant is superior to that of the ant lion, the ant lion position is replaced by the corresponding ant position.

Based on the same inventive concept, the invention also provides a medium-and-long-term runoff prediction system based on linear discriminant analysis IALO-ELM, which comprises a data preprocessing module, a forecast factor selection module, a runoff prediction module, an ant lion algorithm improvement module, an extreme learning machine model optimization module and a forecast value output module;

the data preprocessing module is mainly used for preprocessing the obtained runoff data to finally obtain a runoff sequence;

the forecasting factor selection module is mainly used for selecting the forecasting factors of the radial data by adopting a linear discriminant analysis method;

the runoff prediction module mainly comprises an ant lion algorithm improvement unit, an extreme learning machine model optimization unit and a prediction value output unit;

the runoff ant lion algorithm improvement module initializes the ant lion population by adopting Latin hypercube; performing cross variation on the ant lion algorithm by adopting a plurality of groups; comparing the wandered ants with the ant lion with the best current position by Fuch mapping, and readjusting the position of the best ant lion;

the extreme learning machine model optimization module is used for optimizing parameters of the extreme learning machine by adopting training sample data and an improved ant lion algorithm to obtain optimal parameters and obtain an optimized extreme learning machine model;

and the prediction value output module optimizes the extreme learning machine model by using the ant lion algorithm to obtain the optimal parameters, and then inputs the optimal parameters and prediction data into the extreme learning machine model to obtain a final prediction result.

Has the advantages that: compared with the prior art, the invention has the beneficial effects that: the ant lion algorithm is effectively combined with the extreme learning machine, selective variation operation is carried out on the optimal position of the ant lion, the convergence capacity is improved, the convergence speed is accelerated, and the optimal solution is easier to solve; the improved ant lion algorithm overcomes the defects that the convergence speed of the traditional ant lion algorithm is slow, the searching precision is low and the traditional ant lion algorithm is easy to fall into local optimum after the searching is run; the ant lion algorithm is improved to optimize the hybrid algorithm of the extreme learning machine, so that the method has the characteristics of high convergence speed, strong generalization capability and high prediction precision, and is more suitable for medium-and-long-term runoff prediction.

Drawings

Fig. 1 is a flowchart of a method for predicting medium-and long-term runoff according to an embodiment of the present invention;

fig. 2 is a comparison graph of the real value and the predicted value provided by the embodiment of the present invention.

Detailed Description

The present invention will be described in further detail with reference to the accompanying drawings.

The invention provides a medium-and-long-term runoff forecasting method based on an Improved Ant Lion Algorithm optimization limit Learning Machine, which is used for forecasting medium-and-long-term runoff based on the optimization combination of an Extreme Learning Machine (ELM) and an Improved Ant Lion Algorithm (IALO), and combines the advantages of quick convergence of the ELM and strong global search capability of the Improved Ant Lion Algorithm.

The method specifically comprises the following steps:

step 1: acquiring runoff data in advance, normalizing the data, and selecting a forecasting factor for initial runoff data by using a linear discriminant analysis method.

And acquiring runoff data of the Lijiawan hydrological station and carrying out normalization processing on the data. Performing forecasting factor selection on initial runoff data by using a linear discriminant analysis method, wherein the method comprises the following specific steps:

1) the runoff data of the Li Jia Bay hydrology station 2002-2019 each month is selected as a sample data set. And converting the data set into a dispersion matrix for data encapsulation. Scatter matrix S_wThe calculation formula of (a) is as follows:

S_i＝∑(x-m_i)(x-m_i)^T,i＝1,2

wherein S is_iTo compute the intra-class dispersion matrix, x is the sample data, m_iAs a mean vector of data samples, S_wIs a scatter matrix.

2) Computing eigenvectors of the scatter matrix and corresponding eigenvalues:

(λE-S_W)＝0

wherein λ is an eigenvalue and E is an identity matrix.

3) Sorting the eigenvectors in descending order according to the magnitude of the eigenvalue, then selecting the eigenvector corresponding to the first n largest eigenvalues, and constructing a matrix, namely each column is an eigenvector.

4) The samples are transformed to the new subspace using this eigenvector matrix, thus completing the final predictor selection.

Step 2: initializing an initial ant lion population by using the Latin hypercube, randomly initializing the ant lion positions, and starting iterative search.

Initializing an initial ant lion population by using a Latin hypercube, performing iteration times, performing random initialization of ant lion positions, starting iterative search, and setting a maximum iteration time T_maxThe dimensionality dim of the fitness function, the number of ants and the number of ant lions are N respectively_B、N_YThe control ranges of the variables are selected to be ub and lb, respectively. The Latin hypercube step is as follows:

1) firstly, an n-dimensional unit cube [0,1] is constructed according to the number (n) of the population]ⁿ；

2) Then, each one-dimensional coordinate interval [0,1] is determined according to the initial population number]Are all divided into N_aDividing the N into equal parts_aThe equal parts are randomly arranged, and assume that N is_aThe random arrangements are independent of each other;

3) finally obtaining an N_aRandom matrix of x m dimension, and obtaining N by the above process_aAs an initial population.

And step 3: calculating and sequencing corresponding fitness values, and selecting the ant lion with the largest fitness value as an elite ant lion individual; dividing the ant lion original population into p equal parts, and analyzing the global optimal path by pairwise comparison of interaction and a preselection mechanism by using a multi-population method.

Determining an initial weight value and a threshold value of the extreme learning machine network, activating a function, and training the extreme learning machine under a given sample. Nerve of extreme learning machineThe connection weight beta between the hidden layer and the output layer of a forward-propagation neural network does not need to be iteratively adjusted on the network structure, and N samples (X) are assumed to exist for a single hidden layer neural network_i,t_i) Wherein X is_i＝[x_i1,x_i2,…x_in],t_i＝[t_i1,t_i2,…t_im]. For a single hidden layer neural network with L hidden layer nodes, the formula is expressed as:

wherein g (x) is an activation function, w_iAs input weights, β_iOutput weight, b_iIs the ith hidden layer cell bias.

Known ant lion populationsNP represents the size of the population. Let the number of ant lion individuals in each sub-population be c, and let the ant lion individuals in the sub-population express that X ═ X_iI 1,2, …, c, the position of the ith lion in the search space is X_i＝(X_i1，X_i2，…，X_ic) Dividing the population into two parts, testing one part of the two parts by using a local search mode, calculating the fitness value of the current population individual, finding out excellent individuals and updating the position. And then inserting another kind of population data into the population data for data testing. Method for cross mutation to select elite population

Initializing the positions of ants and ant lions, calculating corresponding fitness values and sequencing, and selecting the ant lions with the largest fitness value as elite ant lions individuals; setting the maximum number of iterations to T_mThe dimension of the fitness function is dim, and the numbers of ants and ant lions are N respectively_B、N_YAnd selecting the control ranges of the variables as ub and lb respectively, wherein ub is the upper limit of the decision variable, and lb is the lower limit of the decision variable.

The random displacement step length of the ant is calculated using the following formula:

X(t)＝[0,cumsum(2r(t₁)-1),cumsum(2r(t₂)-1),...,cumsum(2r(t_M)-1)]

wherein cumsum is a sum function; m is the maximum iteration number; t is the current iteration number; r (t) is a random function, discretizing the transfer function. Where rand is a random number and ranges from [0,1 ].

C, updating the ant lion by using a roulette method and a greedy selection method^t _iAnd d^t _iThe value of (c):

wherein the content of the first and second substances,represents the maximum and minimum values of the ith ant in the t iteration, d_i、c_iRepresents the maximum value and the minimum value of the displacement of the ith ant; d^t、c^tThe maximum value and the minimum value of the t-th iteration displacement of all ants are shown to be reduced along with the increase of the iteration times;represents a normalized displacement;representing the random displacement of the ith ant for the t iteration. The purpose of using roulette and greedy selection is to select ant lion positions and update ant positions.

And 4, step 4: and introducing Fuch mapping into the ant lion algorithm, and performing chaotic traversal search on the neighborhood of the local optimal solution.

The ants after the migration process are positioned at the best position at presentComparing the placed ant lions, and readjusting the position of the best ant lions; generating new individuals near the optimal ant lion through Fuch mapping search; fuch chaotic mapping function expression X_a+1＝cos(1/X_a) Fuch mapping is a novel one-dimensional discrete mapping, a chaotic sequence is mapped into a solution space to obtain a population X, and the initialized population individual of the chaos is X_ia：

X_ia＝lb+(ub-lb)·X_n+1

Wherein, X_iaThe d-dimensional value of the individual of the ith ant colony is shown, and lb and ub are the upper and lower bounds of the d-dimensional value respectively; meanwhile, a strategy of reverse learning of elite is introduced, population number is initialized to calculate an elite reverse population, and finally the population number and the elite reverse population are combined into a new population to calculate a new fitness value so as to determine the position of the best ant lion.

The displacement updating formula of the optimized ant individual is as follows:

wherein the content of the first and second substances,the position of the ith ant at the t iteration is shown;ant lion individual-based shifts selected for the t-th iterative roulette;displacement based on the elite individual for the t-th iteration;for adding reverse regulatory factor at ant positionAnd after updating, calculating a corresponding fitness value, and replacing the ant lion position with the corresponding ant position when the fitness value of the ant is superior to that of the ant lion.

And 5: establishing an extreme learning machine model based on the optimization of the improved ant lion algorithm, and optimizing the connection weight and the threshold of the extreme learning machine by using the improved ant lion algorithm; and predicting the test sample by adopting the trained IALO-ELM model.

According to the monthly runoff data of 2002-2019 of the Li Jia Bay hydrological station, the actually measured runoff and the predicted value are substituted into the following evaluation indexes, and corresponding evaluation index values are calculated. RMSE (root mean square error), MAPE (mean percent error), MSE (mean square error), MAE (mean absolute error), R²The expression of (coefficient of determination) is:

wherein, y_iIs the true output, x, of the ith training sample_iIs a predicted value of the ith sample,is the average of the samples and n is the total number of samples.

Based on the same inventive concept, the invention also provides a medium-and-long-term runoff prediction system based on linear discriminant analysis IALO-ELM, which comprises a data preprocessing module, a forecast factor selection module, a runoff prediction module, an ant lion algorithm improvement module, an extreme learning machine model optimization module and a forecast value output module. Wherein:

the data preprocessing module is mainly used for preprocessing the obtained runoff data to finally obtain a runoff sequence;

the forecasting factor selection module is mainly used for selecting forecasting factors of the radial data by adopting a linear discriminant analysis method;

the runoff prediction module mainly comprises an ant lion algorithm improvement unit, an extreme learning machine model optimization unit and a prediction value output unit;

a runoff ant lion algorithm improvement module initializes the ant lion population by adopting Latin hypercube; performing cross variation on the ant lion algorithm by adopting a plurality of groups; comparing the wandered ants with the ant lion with the best current position by Fuch mapping, and readjusting the position of the best ant lion;

the extreme learning machine model optimization module is used for optimizing parameters of the extreme learning machine by adopting training sample data and an improved ant lion algorithm to obtain optimal parameters and obtain an optimized extreme learning machine model;

and the prediction value output module is used for optimizing the extreme learning machine model by using the ant lion algorithm to obtain the optimal parameters, and then inputting the optimal parameters and prediction data into the extreme learning machine model to obtain a final prediction result.

In order to verify the superiority of the hybrid prediction model provided by the invention, PSO-BP (particle swarm optimization BP neural network), BP (BP neural network), ALO-BP (ant lion optimization BP neural network) and ALO-ELM (ant lion optimization extreme learning model) IALO-ELM (improved ant lion optimization extreme learning model) are compared. As shown in Table 1, IALO-ELM predicts the highest accuracy in all models and performs best in all evaluation metrics.

TABLE 1 error Table of results of the model of the present invention and other control models

Compared with PSO-BP, BP, ALO-ELM and IALO-ELM, the improved ant lion algorithm can effectively improve the prediction precision. Taking RMSE as an example, the PSO-BP value is 213, the BP value is 278, the ALO-BP value is 185, the ALO-ELM value is 167, and the IALO-ELM value is 154, which can prove the effectiveness of the optimization algorithm in the hybrid prediction model. Compared with other four prediction methods, the IALO-ELM has the best prediction effect. As can be seen from fig. 2, the improved ant lion algorithm is combined with the extreme learning machine, and is applied to the medium-and-long-term runoff prediction process, and the obtained real value and the predicted value are compared to obtain a comparison graph, so that the prediction effect is better.

It should be noted that the invention improves the ant lion algorithm to optimize the extreme learning machine, thereby improving the accuracy of global convergence and further improving the prediction accuracy.

The medium-and long-term runoff prediction method system and the medium-and long-term runoff prediction device provided by the invention are introduced in detail. And together with the description, serve to explain the principles and embodiments of the invention, and to facilitate an understanding of the methods and core concepts of the invention. It should be noted that, for those skilled in the art, without departing from the principle of the present invention, several improvements and modifications can be made to the present invention, and these improvements and modifications also fall into the protection scope of the claims of the present invention.