Method and system for predicting power generation capacity of hydroelectric system, electronic equipment and medium

Abstract

The invention discloses a hydropower system power generation capacity prediction method and system, electronic equipment and a medium. Relates to the technical field of hydroelectric generation. According to the scheme, the influence of the rainfall process on the power generation capacity of the hydroelectric system is considered, reference power generation output sequences of different grades are divided, then missing values in the actually-measured output sequences are filled up according to the sequence similarity, and the integrity of output data can be guaranteed; the least square twinborn extreme learning machine is used for solving the power generation capacity prediction problem of the hydroelectric system, and the defect that the prediction result of a traditional method is poor is overcome; in the prediction model construction process, based on the improved grey wolf algorithm and the adaptive mapping initial solution space, an evolutionary operator is introduced to dynamically update the wolf pack position, the uniformity and diversity of the initialized population are enhanced, the global search and local search capabilities are balanced, and the parameter convergence speed and optimization efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of hydropower generation, and in particular to a method, system, electronic equipment and medium for predicting the power generation capacity of a hydropower system. Background Art

[0002] As a model of clean energy, hydropower plays a vital role in the power system, and accurate hydropower station output forecasting is crucial.

[0003] Due to the uneven temporal and spatial distribution of water resources in my country, with the completion of the interconnected large power grid, the transmission and consumption of hydropower has become an important means of efficient utilization of hydropower energy. However, in the case of a large amount of water concentrated in different basins in the regional power grid during the flood season, the concentrated hydropower generation in each basin squeezes the limited transmission channels, leading to problems such as water abandonment and power congestion, and even affecting the safe and stable operation of the power system. Therefore, it is urgent to predict the hydropower generation capacity of different basins in the region, explore the hydraulic, storage capacity, and hydrological compensation regulation laws of regional hydropower station reservoirs and their water energy spatial and temporal replacement capabilities, and coordinate the peak generation of large and small hydropower to ensure the efficient utilization and full consumption of hydropower energy. In addition, the prediction of hydropower generation capacity can help the power grid dispatching department to coordinate and coordinate various power sources, adjust the dispatching plan in a timely and timely manner, reasonably arrange the operation mode of the power grid, and reduce the adverse effects of the access of new energy such as wind and light on the power grid.

[0004] Precipitation is the main meteorological factor affecting the power generation capacity of hydropower systems, but existing studies ignore the complexity and variability of the rainfall process, resulting in insufficient prediction accuracy. Summary of the invention

[0005] The technical problem to be solved by the present invention is that precipitation is the main meteorological factor affecting the power generation capacity of a hydropower system, but existing studies have ignored the complexity and variability of the rainfall process, resulting in insufficient prediction accuracy. The present invention provides a method, system, electronic equipment and medium for predicting the power generation capacity of a hydropower system. Based on the existing hydropower system power generation capacity prediction technology, methodological improvements are made, focusing on the nonlinear and non-stationary prediction of the hydropower output sequence under the condition of changing weather processes, comprehensively capturing the complex relationship between weather processes and the power generation capacity of a hydropower system, and improving the accuracy of hydropower output prediction.

[0006] The present invention is achieved through the following technical solutions:

[0007] This scheme provides a method for predicting the power generation capacity of a hydropower system, including:

[0008] Acquire a measured power generation output sequence of the hydropower system, and perform a first preprocessing on the measured power generation output sequence;

[0009] Based on the improved successive variational mode decomposition method, the measured power generation output sequence after the first preprocessing is decomposed into multiple subsequences;

[0010] After performing the second preprocessing on each subsequence, a training sample set and a test sample set are divided, and a power generation output prediction model of each subsequence is constructed based on the training sample set;

[0011] Based on the improved grey wolf algorithm, the power generation output prediction model of each subsequence is iteratively optimized to obtain the optimal power generation output prediction model of each subsequence;

[0012] The test sample set is input into the optimal power generation output prediction model to calculate the prediction results of each subsequence. The prediction results of each subsequence are summed and denormalized to obtain the prediction results of the hydropower system power generation capacity.

[0013] A further optimization scheme is that the measured power generation output sequence of the hydropower system is obtained, and the measured power generation output sequence is subjected to a first preprocessing; including the method:

[0014] Obtain the historical power generation output sequence of the hydropower system, and divide the historical power generation output sequence into benchmark output sequences of different levels according to the rainfall and its impact period s (s=1,2,…,7);

[0015] Obtain the measured power generation output sequence of the hydropower system and follow the benchmark output sequence Υ s (s=1,2,…,7) Identify the missing values ​​in the measured power output sequence, and fill the missing values ​​based on the sequence similarity method and linear interpolation method, so as to obtain the complete measured power output sequence.

[0016] A further optimization scheme is that the historical power generation output sequence is divided into benchmark output sequences Y at different levels according to the rainfall and its impact period. s (s=1,2,…,7), including methods:

[0017] S111, initial configuration period t=1;

[0018] S112, determine the rainfall P of the hydropower system in period t t Impact period Z t ;

[0019] S113, according to the period t to t+Z t The accumulated rainfall value for the period Determine the hydropower system from time t to t+Z t The power generation sequence N of the time period z (z=1,2,…,Z t ) belongs to the benchmark output sequence Y s(s=1,2,…,7); where Indicates the period from t to t+Z t The rainfall in the period; P t represents the rainfall in period t; P z represents the rainfall in time period z; z represents the time period increment;

[0020] Step S113 specifically includes the following method:

[0021] like It belongs to the benchmark output sequence Υ1; if It belongs to the benchmark output sequence Υ2; if It belongs to the benchmark output sequence Y3; if It belongs to the benchmark output sequence Y4; if It belongs to the benchmark output sequence Υ5; if It belongs to the benchmark output sequence Y6; if It belongs to the benchmark output sequence Y7; the hydropower system from time t to t+Z t The power generation sequence N of the time period z (z=1,2,…,Z t ) are added to the corresponding benchmark output sequence Ys in sequence;

[0022] S114, set time period t=t+Z t +1, if time period t≤T, T is the maximum time period; then return to step S112; otherwise, stop the calculation and output the benchmark output sequence Υs (s=1,2,…,7) at different levels, where Υs={Ns,1,…,Ns,a,…,Ns, As}; N s,a A represents the ath output in the sth benchmark output sequence; s Represents the length of the sth benchmark output sequence, satisfying

[0023]

[0024] A further optimization scheme is that the measured power generation output sequence after the first preprocessing is decomposed into multiple subsequences based on the improved successive variational modal decomposition method, including the following method:

[0025] Based on the successive variational modal decomposition method, the measured power generation output sequence s(o) is decomposed into the Kth order modal component μ K (o) and the residual component s r (o); s(o) = μ K (o)+s r (o); s μ (o) indicates the untreated part; μ i(o) represents the i-th order modal component; where o represents the time series to be processed;

[0026] Based on the decomposition principle, the successive variation modal decomposition is established to obtain multiple subsequences of the measured power output sequence; the decomposition principle includes: each modal component is closely around its central frequency principle, ensuring that the Kth order modal component μ K (o) Based on the effective component, the residual component s r (o) Minimization principle, the principle that both the Kth and K-1th order modes are identifiable, and the principle that the signal can be reconstructed during the decomposition process;

[0027] The modal component of each subsequence is determined based on the correlation coefficient method, and the dimensionality reduction and denoising of each subsequence is performed based on the principal component analysis method.

[0028] A further optimization scheme is that the method for implementing the principle that each modal component is closely around its central frequency includes:

[0029] Minimize the K-th order modal component according to the constraint J1:

[0030]

[0031] Where δ(o) is the Dirac function; * is the convolution operator; ω K (o) is the center frequency of the Kth mode; It means partial derivative of o; j is the unit of imaginary number; represents the square of L2 norm;

[0032] The Kth order modal component μ is guaranteed to K (o) Based on the effective component, the residual component s r (o) Methods for implementing the minimization principle include:

[0033] Select a suitable filter to minimize the spectral overlap between the residual component and the K-th mode according to the constraint J2:

[0034]

[0035] Among them, β K (o) represents the impulse response of the filter corresponding to the K-th order mode, and the response frequency is ω represents the frequency of the current signal;

[0036] Ensure that the K-th order modal component μ K (o) Based on the effective component, the residual component s r (o) Methods for implementing the minimization principle include:

[0037] According to the constraint J3, the frequency spectrum overlap from the Kth mode to the K-1th mode is minimized:

[0038]

[0039] Among them, β i (o) is the impulse response of the filter corresponding to the overlapping frequency, and the response frequency is ω i Represents the center frequency corresponding to the i-th component of the overlapping mode;

[0040] Methods for realizing the principle that both the K-order mode and the K-1-order mode are identifiable and the principle that the signal can be reconstructed during the decomposition process include:

[0041] make When the K-1th mode is known, the extraction problem of the Kth mode is expressed as a constrained minimization problem:

[0042] Where α represents the equilibrium parameter.

[0043] A further optimization scheme is to determine the modal components of each subsequence based on the correlation coefficient method, and perform dimensionality reduction and noise elimination processing on each subsequence based on the principal component analysis method, including the following method:

[0044] Create the original data set X of the low-frequency modal component P P , and remove the mean value to get the data set The low-frequency modal component P∈[1st-order modal component, K-1th-order modal component];

[0045] Computational Dataset The covariance matrix Φ P , and the covariance matrix Φ P The eigenvalue λ i (i=1,2,…,P) and eigenvector a i , according to the feature vector a i Arrange the principal component matrix A=a1,a2,…,a in descending order P ;

[0046] Determine the number of elements P of the principal component matrix A * , extract the first P of the principal component matrix * The column vectors form the mapping matrix U;

[0047] According to the projection matrix, we can find the reduced dimension data W = X P U, and finally get K P subsequences.

[0048] A further optimization scheme is that each subsequence is divided into a training sample set and a test sample set after the second preprocessing, and a power generation output prediction model of each subsequence is constructed based on the training sample set; including the method:

[0049] Based on the partial autocorrelation analysis method, the number of previous influencing factors of each subsequence is determined and normalized;

[0050] Divide the normalized subsequences into training sample sets according to the preset ratio and test sample set in Respectively represent the influencing factor set and output value of the i-th training sample; They represent the influencing factor set and output value of the jth test sample respectively; m and n represent the number of training samples and test samples respectively;

[0051] Based on the training sample set, a power generation output prediction model is constructed based on the least squares twin extreme learning machine:

[0052]

[0053] Among them, f1(x) represents the upper bound function corresponding to the influencing factor x; f2(x) represents the lower bound function corresponding to the influencing factor x; β1 represents the upper limit of the prediction result; β2 represents the lower limit of the prediction result; h(x) represents the weighting function; f(x) represents the final prediction result corresponding to the influencing factor x.

[0054] A further optimization scheme is that the power generation output prediction model of each subsequence is iteratively optimized based on the improved grey wolf algorithm to obtain the optimal power generation output prediction model of each subsequence; including the method:

[0055] Initial configuration improves the parameters of the Grey Wolf Algorithm;

[0056] Generate an initialization population based on an adaptive mapping strategy:

[0057]

[0058] In the formula, represents the i-th parameter solution generated in the l-th iteration, i=1,2,…,I; I represents the total number of parameter solutions; l=1,2,…,L; L represents the total number of iterations; coe represents the adjustment parameter; mod() represents the remainder; x max 、x min They represent the upper and lower limits of the variable respectively; C represents the intermediate variable; a represents the first auxiliary variable; b represents the second auxiliary variable; rand() represents the function of generating an N×D matrix, where each element of the matrix is ​​a random number uniformly distributed in the interval [0,1], N represents the number of rows of the matrix, and D represents the number of columns of the matrix.

[0059] Update all parameter solutions based on dynamic equations:

[0060] In the formula, represents the position vector of the i-th gray wolf at the l-th iteration; They represent the best, second best, and third best parameter solutions found from the first iteration to the lth iteration respectively; C1, C2, and C3 represent the coefficient vectors for adjusting the search direction in the process of finding the best, second best, and third best parameter solutions respectively; and Respectively The difference vector of

[0061] Update the first auxiliary variable a of the lth iteration based on the nonlinear decay strategy l :

[0062]

[0063] Wherein, exp represents the exponential function.

[0064] The present invention also provides a hydropower system power generation capacity prediction system, which is used to implement the above-mentioned hydropower system power generation capacity prediction method, and the system comprises:

[0065] A preprocessing module, used for obtaining a measured power generation output sequence of the hydropower system and performing a first preprocessing on the measured power generation output sequence;

[0066] A decomposition module, used for decomposing the measured power generation output sequence after the first preprocessing into a plurality of subsequences based on an improved successive variational mode decomposition method;

[0067] A model building module, used to divide each subsequence into a training sample set and a test sample set after performing a second preprocessing on each subsequence, and to build a power generation output prediction model for each subsequence based on the training sample set;

[0068] A model optimization module is used to iteratively optimize the power generation output prediction model of each subsequence based on the improved grey wolf algorithm to obtain the optimal power generation output prediction model of each subsequence;

[0069] The prediction module is used to input the test sample set into the optimal power generation output prediction model to calculate the prediction results of each subsequence, and to obtain the prediction results of the hydropower system power generation capacity after summing and denormalizing the prediction results of each subsequence.

[0070] The present solution also provides a computer-readable medium having a computer program stored thereon, and the computer program is executed by a processor to implement a method for predicting the power generation capacity of a hydropower system as described above.

[0071] The present invention also provides an electronic device, comprising:

[0072] Memory for storing computer programs;

[0073] The processor is used to execute the computer program stored in the memory, and when the computer program is executed, the above-mentioned method for predicting the power generation capacity of the hydropower system is implemented.

[0074] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0075] 1. The present invention provides a method, system, electronic device and medium for predicting the power generation capacity of a hydropower system; the influence of the rainfall process on the power generation capacity of the hydropower system is taken into consideration, and different levels of benchmark power generation output sequences are divided, and then the missing values ​​in the measured output sequence are filled according to the sequence similarity, so as to ensure the integrity of the output data.

[0076] 2. The present invention provides a method, system, electronic device and medium for predicting the power generation capacity of a hydropower system; the least squares twin extreme learning machine is used to solve the problem of predicting the power generation capacity of a hydropower system, which solves the problem of poor prediction results of traditional methods. Compared with the extreme learning machine, it has strong generalization ability and high prediction stability; compared with the twin extreme learning machine, it has faster implementation speed and higher prediction efficiency, and can better realize the task of hydropower output prediction.

[0077] 3. The present invention provides a method, system, electronic device and medium for predicting the power generation capacity of a hydropower system; in the process of constructing a prediction model, based on the improved gray wolf algorithm and adaptive mapping to initialize the solution space, an evolutionary operator is introduced to dynamically update the position of the wolf pack, thereby enhancing the uniformity and diversity of the initialized population, balancing the global search and local search capabilities, and improving the parameter convergence speed and optimization efficiency.

[0078] 4. The present invention provides a method, system, electronic device and medium for predicting the power generation capacity of a hydropower system; the output sequence is preprocessed based on successive variational mode decomposition combined with principal component analysis, which effectively reduces the non-stationarity and nonlinearity of the data, integrates the advantages of data decomposition method, improved gray wolf algorithm and new regression prediction model, and significantly improves the accuracy of the hydropower output prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without creative work. In the drawings:

[0080] Figure 1 It is a flow chart of the method for predicting the power generation capacity of a hydropower system;

[0081] Figure 2It is a schematic diagram of the principle of the method for predicting the power generation capacity of the hydropower system;

[0082] Figure 3 It is a schematic diagram for comparing the prediction results of various prediction methods;

[0083] Figure 4 A schematic diagram of the electronic device structure. DETAILED DESCRIPTION

[0084] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0085] Precipitation is the main meteorological factor affecting the power generation capacity of hydropower systems, but existing studies ignore the complexity and variability of the rainfall process, resulting in insufficient prediction accuracy. In view of this, this solution provides the following embodiments to solve the above technical problems:

[0086] Example 1

[0087] This embodiment provides a method for predicting the power generation capacity of a hydropower system. Figure 1 and Figure 2 As shown, including:

[0088] Step 1: obtaining a measured power generation output sequence of the hydropower system, and performing a first preprocessing on the measured power generation output sequence; Step 1 specifically includes the following method:

[0089] S111, initial configuration period t=1;

[0090] S112, determine the rainfall P of the hydropower system in period t t Impact period Z t ;

[0091] S113, according to the period t to t+Z t The accumulated rainfall value for the period Determine the hydropower system from time t to t+Z t The power generation sequence N of the time period z (z=1,2,…,Z t ) belongs to the benchmark output sequence Υ s (s=1,2,…,7); where Indicates the period from t to t+Z t The rainfall in the period; P t represents the rainfall in period t; P z represents the rainfall in time period z; z represents the time period increment;

[0092] Specific methods include:

[0093] like It belongs to the benchmark output sequence Υ1; if It belongs to the benchmark output sequence Υ2; if It belongs to the benchmark output sequence Y3; if It belongs to the benchmark output sequence Y4; if It belongs to the benchmark output sequence Υ5; if It belongs to the benchmark output sequence Y6; if It belongs to the benchmark output sequence Y7; the hydropower system from time t to t+Z t The power generation sequence N of the time period z (z=1,2,…,Z t ) are added to the corresponding benchmark output sequence Ys in sequence;

[0094] S114, set time period t=t+Z t +1, if time period t≤T, T is the maximum time period; then return to step S112; otherwise, stop the calculation and output the benchmark output sequence Y at different levels s (s=1,2,…,7), where N s,a A represents the ath output in the sth benchmark output sequence; s Represents the length of the sth benchmark output sequence, satisfying

[0095]

[0096] S12, obtaining the measured power generation output sequence of the hydropower system, and s (s=1,2,…,7) Identify the missing values ​​in the measured power output sequence, and fill the missing values ​​based on the sequence similarity method and linear interpolation method, so as to obtain the complete measured power output sequence.

[0097] Step 2: Decompose the measured power generation output sequence after the first preprocessing into multiple subsequences based on the improved successive variational mode decomposition method; Step 2 specifically includes the following method:

[0098] S21, based on the successive variational modal decomposition method, decompose the measured power output sequence s(o) into the Kth order modal component μ K (o) and the residual component s r (o); s(o) = μ K (o)+s r (o); s μ (o) indicates the untreated part; μ i (o) represents the i-th order modal component; where o represents the time series to be processed;

[0099] S22, based on the decomposition principle, establish a successive variational modal decomposition to obtain multiple subsequences of the measured power output sequence; the decomposition principle includes: each modal component is closely around its central frequency principle, ensuring that the Kth order modal component μ K (o) Based on the effective component, the residual component s r (o) Minimization principle, the principle that both the Kth and K-1th order modes are identifiable, and the principle that the signal can be reconstructed during the decomposition process;

[0100] The method for implementing the principle that each modal component is closely around its central frequency includes:

[0101] Minimize the K-th order modal component according to the constraint J1:

[0102]

[0103] Where δ(o) is the Dirac function; * is the convolution operator; ω K (o) is the center frequency of the Kth mode; It means partial derivative of o; j is the unit of imaginary number; represents the square of L2 norm;

[0104] The Kth order modal component μ is guaranteed to K (o) Based on the effective component, the residual component s r (o) Methods for implementing the minimization principle include:

[0105] Select a suitable filter to minimize the spectral overlap between the residual component and the K-th mode according to the constraint J2:

[0106]

[0107] Among them, β K (o) represents the impulse response of the filter corresponding to the K-th order mode, and the response frequency is ω represents the frequency of the current signal;

[0108] Ensure that the K-th order modal component μ K (o) Based on the effective component, the residual component s r (o) Methods for implementing the minimization principle include:

[0109] The Kth and K-1th order modes cannot be identified by using the criteria J1 and J2. Therefore, the idea of ​​criterion J2 is adopted and the impulse response is selected as β i The filter of (o) minimizes the spectral overlap from the Kth mode to the K-1th mode according to the constraint J3:

[0110]

[0111] Among them, β i (o) is the impulse response of the filter corresponding to the overlapping frequency, and the response frequency is ω i Represents the center frequency corresponding to the i-th component of the overlapping mode;

[0112] Methods for realizing the principle that both the K-order mode and the K-1-order mode are identifiable and the principle that the signal can be reconstructed during the decomposition process include:

[0113] In order to ensure that the signal can be reconstructed during the decomposition process, When the K-1th mode is known, the extraction problem of the Kth mode is expressed as a constrained minimization problem:

[0114] Where α represents the equilibrium parameter.

[0115] S23, judging the modal components of each subsequence based on the correlation coefficient method, and performing dimensionality reduction and noise elimination processing on each subsequence based on the principal component analysis method. Step S23 specifically includes the following method:

[0116] Create the original data set X of the low-frequency modal component P P , and remove the mean value to get the data set The low-frequency modal component P∈[1st-order modal component, K-1th-order modal component];

[0117] Computational Dataset The covariance matrix Φ P , and the covariance matrix Φ P The eigenvalue λ i (i=1,2,…,P) and eigenvector a i , according to the feature vector a i Arrange the principal component matrix A=a1,a2,…,a in descending order P ;

[0118] Determine the number of elements P of the principal component matrix A * , extract the first P of the principal component matrix * The column vectors form the mapping matrix U;

[0119] According to the projection matrix, we can find the reduced dimension data W = X P U, and finally get K P subsequences.

[0120] Step three, after performing the second preprocessing on each subsequence, a training sample set and a test sample set are divided, and a power generation output prediction model of each subsequence is constructed based on the training sample set; Step three specifically includes the following method:

[0121] S31, determining the number of previous influencing factors of each subsequence based on partial autocorrelation analysis method, and performing normalization processing;

[0122] S32, dividing the normalized subsequences into training sample sets according to a preset ratio and test sample set in Respectively represent the influencing factor set and output value of the i-th training sample; They represent the influencing factor set and output value of the jth test sample respectively; m and n represent the number of training samples and test samples respectively;

[0123] S33, based on the training sample set, builds a power generation output prediction model based on the least squares twin extreme learning machine:

[0124]

[0125] Among them, f1(x) represents the upper bound function corresponding to the influencing factor x; f2(x) represents the lower bound function corresponding to the influencing factor x; β1 represents the upper limit of the prediction result; β2 represents the lower limit of the prediction result; h(x) represents the weighting function; f(x) represents the final prediction result corresponding to the influencing factor x.

[0126] The least squares twin extreme learning machine solves the original problem of the twin extreme learning machine using the least squares method, including using the L2 norm of the slack variable to replace the L1 norm and replacing the inequality constraint with an equality constraint. In the least squares twin extreme learning machine, the original problem of the twin extreme learning machine is converted into the following form:

[0127]

[0128] In the formula, H=(h(x1) T ,…,h(x n ) T ) represents the hidden layer of the training set; h(x)=(h(x1),…,h(x n )) T represents the nonlinear random feature mapping of the hidden layer relative to x, L represents the number of hidden layers; ξ and η represent the first slack variable and the second slack variable; C1 and C2 represent the first trade-off parameter and the second trade-off parameter; e represents a unit vector;

[0129] At this time, the weight used in the quadratic programming problem formula (10) is The square of the L2 norm of the slack variable ξ, thereby converting the solution of formula (10) into a solution of a system of simultaneous linear equations; replacing the equality constraint into the objective function, we get:

[0130]

[0131] Solving for β1 gradient is zero, and the optimal value of β1 is calculated as:

[0132] β1=C1(H T H+C1H T H+ωI) -1 H T (Y-ε1e) (13)

[0133] Where ωI is the introduced regularization term, and ω is an infinitesimal positive number.

[0134] Based on similar linear problem solutions, the optimal value of β2 is calculated as:

[0135] β2=C2(H T H+C2H T H+ωI) -1 H T (Y+ε2e) (14)

[0136] Thus, the upper and lower bound functions f1(x) and f2(x) are obtained respectively, and finally the power generation output prediction model is obtained.

[0137] Step 4: Iteratively optimize the power generation output prediction model of each subsequence based on the improved grey wolf algorithm to obtain the optimal power generation output prediction model of each subsequence; Step 4 specifically includes the following method:

[0138] Initial configuration of parameters of the improved grey wolf algorithm; the parameters include the number of parameter schemes I, the maximum number of iterations L, the variable dimension, the upper limit and the lower limit of the variable;

[0139] Let counter l = 1 to overcome the problem that the initial population distribution of the original grey wolf algorithm is poor and it is difficult to find the global optimal solution. This scheme generates the initialization population based on the adaptive mapping strategy:

[0140]

[0141] In the formula, represents the i-th parameter solution generated in the l-th iteration, i=1,2,…,I; I represents the total number of parameter solutions; l=1,2,…,L; L represents the total number of iterations; coe represents the adjustment parameter; mod() represents the remainder; x max 、x min Respectively represent the upper and lower limits of the variable; C represents the intermediate variable; a represents the first auxiliary variable; b represents the second auxiliary variable; rand() represents the function of generating an N×D matrix, where each element of the matrix is ​​a random number uniformly distributed in the interval [0,1], N represents the number of rows of the matrix, and D represents the number of columns of the matrix;

[0142] Update all parameter solutions based on dynamic equations:

[0143] In the formula, represents the position vector of the i-th gray wolf at the l-th iteration; They represent the best, second best, and third best parameter solutions found from the first iteration to the lth iteration respectively; C1, C2, and C3 represent the coefficient vectors for adjusting the search direction in the process of finding the best, second best, and third best parameter solutions respectively; and Respectively The difference vector of

[0144] Update the first auxiliary variable al of the lth iteration based on the nonlinear decay strategy:

[0145]

[0146] in

[0147]

[0148] Where A1, A2, and A3 are coefficient vectors used to adjust the search step size; Respectively The direction of progress; represents the position vector of the i-th gray wolf at the l+1th iteration; r1 and r2 are random vectors uniformly distributed in the interval [0,1], which are used to increase the randomness of the search; It is the global optimal individual with the best fitness value;

[0149] To prevent over-convergence, the top 3γ% individuals in the global optimal population are selected; is the best parameter solution found in the lth iteration; trnd is the t distribution function; floor() represents the floor function; mean() represents the mean function; and exp represents the exponential function.

[0150] Step 5: Input the test sample set into the optimal power output prediction model to calculate the prediction results of each subsequence, and sum and denormalize the prediction results of each subsequence to obtain the prediction results of the hydropower system power generation capacity.

[0151] In order to demonstrate the effect achieved by the present invention, the present invention selects a hydropower station output sequence for comparative study. For a sample sequence of length n, if t i and are the measured output value and the model predicted value of the i-th sample, respectively. is the average value of the measured output sequence; the evaluation indicators are as follows: root mean square error (RMSE), mean absolute error (MAE), and determination coefficient (R2), and the calculation method is as follows:

[0152]

[0153] The experimental results are shown in Table 1.

[0154] Table 1 Results of prediction model evaluation indicators

[0155]

[0156] Figure 3 It is a schematic diagram comparing the prediction results of the method of the present invention with those of the extreme learning machine (ELM) and the twin extreme learning machine (TELM); wherein (a) is the prediction result of the method of the present invention, (a) is the prediction result of the extreme learning machine, and (c) is the prediction result of the twin extreme learning machine. According to the accompanying drawings and Table 1, it can be seen that compared with the extreme learning machine (ELM) method and the twin extreme learning machine (TELM) method, the determination coefficient (R2) of the method of the present invention is the highest, while the mean absolute error (MAE) and the root mean square error (RMSE) are reduced by more than 2 times. Therefore, the force prediction result obtained by the present invention has a smaller error, a higher degree of fit with the measured force value, and the model has good prediction performance.

[0157] Example 2

[0158] This embodiment provides a hydropower system power generation capacity prediction system, which is used to implement the above-mentioned hydropower system power generation capacity prediction method considering weather processes, and the system includes:

[0159] A preprocessing module, used for obtaining a measured power generation output sequence of the hydropower system and performing a first preprocessing on the measured power generation output sequence;

[0160] A decomposition module, used for decomposing the measured power generation output sequence after the first preprocessing into a plurality of subsequences based on an improved successive variational mode decomposition method;

[0161] A model building module, used to divide each subsequence into a training sample set and a test sample set after performing a second preprocessing on each subsequence, and to build a power generation output prediction model for each subsequence based on the training sample set;

[0162] A model optimization module is used to iteratively optimize the power generation output prediction model of each subsequence based on the improved grey wolf algorithm to obtain the optimal power generation output prediction model of each subsequence;

[0163] The prediction module is used to input the test sample set into the optimal power generation output prediction model to calculate the prediction results of each subsequence, and to obtain the prediction results of the hydropower system power generation capacity after summing and denormalizing the prediction results of each subsequence.

[0164] Example 3

[0165] This embodiment provides a computer-readable medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement a method for predicting the power generation capacity of a hydropower system as described in Embodiment 1; specifically, the following sub-steps are performed:

[0166] Step 1: obtaining a measured power generation output sequence of the hydropower system, and performing a first preprocessing on the measured power generation output sequence;

[0167] Step 2: Decomposing the measured power generation output sequence after the first preprocessing into multiple subsequences based on the improved successive variational mode decomposition method;

[0168] Step 3, performing a second preprocessing on each subsequence to divide it into a training sample set and a test sample set, and constructing a power generation output prediction model for each subsequence based on the training sample set;

[0169] Step 4: Iteratively optimize the power generation output prediction model of each subsequence based on the improved grey wolf algorithm to obtain the optimal power generation output prediction model of each subsequence;

[0170] Step 5: Input the test sample set into the optimal power output prediction model to calculate the prediction results of each subsequence, and sum and denormalize the prediction results of each subsequence to obtain the prediction results of the hydropower system power generation capacity.

[0171] Example 4

[0172] This embodiment provides an electronic device, such as Figure 4 As shown, including:

[0173] Memory for storing computer programs;

[0174] The processor is used to execute the computer program stored in the memory, and when the computer program is executed, the method for predicting the power generation capacity of the hydropower system of Example 1 is implemented. The specific implementation methods described above further explain the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for predicting the power generation capacity of a hydropower system, characterized in that: include: Acquire a measured power generation output sequence of the hydropower system, and perform a first preprocessing on the measured power generation output sequence; Based on the improved successive variational mode decomposition method, the measured power generation output sequence after the first preprocessing is decomposed into multiple subsequences; After performing the second preprocessing on each subsequence, a training sample set and a test sample set are divided, and a power generation output prediction model of each subsequence is constructed based on the training sample set; Based on the improved grey wolf algorithm, the power generation output prediction model of each subsequence is iteratively optimized to obtain the optimal power generation output prediction model of each subsequence; The test sample set is input into the optimal power generation output prediction model to calculate the prediction results of each subsequence. The prediction results of each subsequence are summed and denormalized to obtain the prediction results of the hydropower system power generation capacity.

2. A method for predicting the power generation capacity of a hydropower system according to claim 1, characterized in that: The step of obtaining a measured power generation output sequence of the hydropower system and performing a first preprocessing on the measured power generation output sequence; Included methods: Obtain the historical power generation output sequence of the hydropower system, and divide the historical power generation output sequence into benchmark output sequences of different levels according to the rainfall and its impact period s (s=1,2,…,7); Obtain the measured power generation output sequence of the hydropower system and follow the benchmark output sequence Υ s (s=1,2,…,7) Identify the missing values ​​in the measured power output sequence, and fill the missing values ​​based on the sequence similarity method and linear interpolation method, so as to obtain the complete measured power output sequence.

3. A method for predicting the power generation capacity of a hydropower system according to claim 2, characterized in that: According to the rainfall and its impact period, the historical power generation output sequence is divided into benchmark output sequences of different levels. s (s=1,2,…,7), including methods: S111, initial configuration period t=1; S112, determine the rainfall P of the hydropower system in period t t Impact period Z t ; S113, according to the period t to t+Z t The accumulated rainfall value for the period Determine the hydropower system from time t to t+Z t The power generation sequence N of the time period z (z=1,2,…,Z t ) belongs to the benchmark output sequence Υ s (s=1,2,…,7); where Indicates the period from t to t+Z t The rainfall in the period; P t represents the rainfall in period t; P z represents the rainfall in time period z; z represents the time period increment; Specific methods include: S114, set time period t=t+Z t +1, if time period t≤T, T is the maximum time period; then return to step S112; Otherwise, stop the calculation and output the benchmark output sequence Υ at different levels s (s=1,2,…,7), where N s,a A represents the ath output in the sth benchmark output sequence; s Represents the length of the sth benchmark output sequence, satisfying 4. A method for predicting the power generation capacity of a hydropower system according to claim 3, characterized in that: The S113 specifically includes the method: if It belongs to the benchmark output sequence Υ1; if It belongs to the benchmark output sequence Υ2; if It belongs to the benchmark output sequence Y3; if It belongs to the benchmark output sequence Y4; if It belongs to the benchmark output sequence Υ5; if It belongs to the benchmark output sequence Y6; if It belongs to the benchmark output sequence Y7; the hydropower system from time t to t+Z t The power generation sequence N of the time period z (z=1,2,…,Z t ) are added to the corresponding benchmark output sequence Y s .

5. A method for predicting the power generation capacity of a hydropower system according to claim 1, characterized in that: The method of decomposing the measured power generation output sequence after the first preprocessing into multiple subsequences based on the improved successive variational modal decomposition method includes: Based on the successive variational modal decomposition method, the measured power output sequence s(o) is decomposed into the Kth order modal component μ K (o) and the residual component s r (o); s μ (o) indicates the untreated part; μ i (o) represents the i-th order modal component; Where o represents the time series to be processed; Based on the decomposition principle, the successive variation modal decomposition is established to obtain multiple subsequences of the measured power output sequence; the decomposition principle includes: each modal component is closely around its central frequency principle, ensuring that the Kth order modal component μ K (o) Based on the effective component, the residual component s r (o) Minimization principle, the principle that both the Kth and K-1th order modes are identifiable, and the principle that the signal can be reconstructed during the decomposition process; The modal component of each subsequence is determined based on the correlation coefficient method, and the dimensionality reduction and denoising of each subsequence is performed based on the principal component analysis method.

6. A method for predicting the power generation capacity of a hydropower system according to claim 5, characterized in that: The method for implementing the principle that each modal component is closely around its central frequency includes: Minimize the K-th order modal component according to the constraint J1: Where δ(o) is the Dirac function; * is the convolution operator; ω K is the center frequency of the K-th mode; It means partial derivative of o; j is the unit of imaginary number; represents the square of L2 norm; The Kth order modal component μ is guaranteed to K (o) Based on the effective component, the residual component s r (o) Methods for implementing the minimization principle include: Minimize the spectral overlap of the residual component and the K-th mode according to the constraint J2: Among them, β K (o) represents the impulse response of the filter corresponding to the K-th order mode, and the response frequency is ω represents the frequency of the current signal; Ensure that the K-th order modal component μ K (o) Based on the effective component, the residual component s r (o) Methods for implementing the minimization principle include: According to the constraint J3, the frequency spectrum overlap from the Kth mode to the K-1th mode is minimized: Among them, β i (o) is the impulse response of the filter corresponding to the overlapping frequency, and the response frequency is ω i Represents the center frequency corresponding to the i-th component of the overlapping mode; Methods for realizing the principle that both the K-order mode and the K-1-order mode are identifiable and the principle that the signal can be reconstructed during the decomposition process include: make When the K-1th mode is known, the extraction problem of the Kth mode is expressed as a constrained minimization problem: Where α represents the equilibrium parameter.

7. A method for predicting the power generation capacity of a hydropower system according to claim 5, characterized in that: The method of determining the modal components of each subsequence based on the correlation coefficient method and performing dimensionality reduction and noise elimination processing on each subsequence based on the principal component analysis method includes the following methods: Create the original data set X of the low-frequency modal component P P , and remove the mean value to get the data set The low-frequency modal component P∈[1st-order modal component, K-1th-order modal component]; Computational Dataset The covariance matrix Φ P , and the covariance matrix Φ P The eigenvalue λ i (i=1,2,…,P) and eigenvector a i , according to the feature vector a i Arrange the principal component matrix A=a1,a2,…,a in descending order P ; Determine the number of elements P of the principal component matrix A * , extract the first P of the principal component matrix * The column vectors form the mapping matrix U; According to the projection matrix, we can find the reduced dimension data W = X P U, and finally get K P subsequences.

8. A method for predicting the power generation capacity of a hydropower system according to claim 1, characterized in that: The second preprocessing is performed on each subsequence to divide it into a training sample set and a test sample set, and a power generation output prediction model of each subsequence is constructed based on the training sample set; including method: Based on the partial autocorrelation analysis method, the number of previous influencing factors of each subsequence is determined and normalized; Divide the normalized subsequences into training sample sets according to the preset ratio and test sample set in Respectively represent the influencing factor set and output value of the i-th training sample; They represent the influencing factor set and output value of the jth test sample respectively; m and n represent the number of training samples and test samples respectively; Based on the training sample set, a power generation output prediction model is constructed based on the least squares twin extreme learning machine: Among them, f1(x) represents the upper bound function corresponding to the influencing factor x; f2(x) represents the lower bound function corresponding to the influencing factor x; β1 represents the upper limit of the prediction result; β2 represents the lower limit of the prediction result; h(x) represents the weighting function; f(x) represents the final prediction result corresponding to the influencing factor x.

9. A method for predicting the power generation capacity of a hydropower system according to claim 1, characterized in that: The improved grey wolf algorithm is used to iteratively optimize the power generation output prediction model of each subsequence to obtain the optimal power generation output prediction model of each subsequence; including the following method: Initial configuration improves the parameters of the Grey Wolf Algorithm; Generate an initialization population based on an adaptive mapping strategy: In the formula, represents the i-th parameter solution generated by the bottom l iterations, i=1,2,…,I; I represents the total number of parameter solutions; l=1,2,…,L; L represents the total number of iterations; coe represents the adjustment parameter; mod() represents the remainder; x max 、x min Respectively represent the upper and lower limits of the variable; C represents the intermediate variable; a represents the first auxiliary variable; b represents the second auxiliary variable; rand() represents the function of generating an N×D matrix, where each element of the matrix is ​​a random number uniformly distributed in the interval [0,1], N represents the number of rows of the matrix, and D represents the number of columns of the matrix; Update all parameter solutions based on dynamic equations: In the formula, represents the position vector of the i-th gray wolf at the l-th iteration; They represent the best, second best, and third best parameter solutions found from the first iteration to the lth iteration respectively; C1, C2, and C3 represent the coefficient vectors for adjusting the search direction in the process of finding the best, second best, and third best parameter solutions respectively; and Respectively The difference vector of Update the first auxiliary variable a of the lth iteration based on the nonlinear decay strategy l : Wherein, exp represents the exponential function.

10. A hydropower system power generation capacity prediction system, characterized in that: A method for predicting the power generation capacity of a hydropower system according to any one of claims 1 to 9, the system comprising: A preprocessing module, used for obtaining a measured power generation output sequence of the hydropower system and performing a first preprocessing on the measured power generation output sequence; A decomposition module, used for decomposing the measured power generation output sequence after the first preprocessing into a plurality of subsequences based on an improved successive variational mode decomposition method; A model building module, used to divide each subsequence into a training sample set and a test sample set after performing a second preprocessing on each subsequence, and to build a power generation output prediction model for each subsequence based on the training sample set; A model optimization module is used to iteratively optimize the power generation output prediction model of each subsequence based on the improved grey wolf algorithm to obtain the optimal power generation output prediction model of each subsequence; The prediction module is used to input the test sample set into the optimal power generation output prediction model to calculate the prediction results of each subsequence, and to obtain the prediction results of the hydropower system power generation capacity after summing and denormalizing the prediction results of each subsequence.

11. A computer readable medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor to implement the method for predicting the power generation capacity of a hydropower system taking into account weather processes as described in any one of claims 1 to 9.

12. An electronic device, characterized in that: include: Memory for storing computer programs; A processor is used to execute the computer program stored in the memory, and when the computer program is executed, a method for predicting the power generation capacity of a hydropower system as described in any one of claims 1 to 9 is implemented.

Images