Ultra-short-term wind power prediction method and device, electronic equipment and medium

Abstract

The invention discloses an ultra-short-term wind power prediction method and device, electronic equipment and a medium. The method comprises the following steps: acquiring a wind power time sequence, performing a mutual information method for many times, recording the optimal delay time of each time, and forming a time delay sequence by taking a plurality of optimal delay times as elements; constructing a multi-dimensional phase space for the wind power time sequence according to a time delay sequence, and searching an optimal embedding dimension; performing non-equal time delay phase space reconstruction on the wind power time sequence according to the time delay sequence and the embedded dimension to obtain a phase space; obtaining a space-time cooperation fragment set from the phase space, and constructing a model input Xin on the basis of the space-time cooperation fragment set; and taking the obtained Xin as the input of a model, training the model, and predicting a wind power value by adopting the model. According to the method, the predictability and the independence of the wind power phase space are better than those of equal-time-delay phase space reconstruction through non-equal-time-delay phase space reconstruction, and the wind power prediction precision is improved by combining time-space collaboration of model input.

Description

Technical Field

[0001] The present invention relates to the fields of wind power prediction and economic dispatch in power systems, and particularly to a very short-term wind power prediction method, device, electronic device and medium. Background Art

[0002] With the rapid development of the world economy, the energy demand has increased sharply. While traditional fossil energy faces the crisis of exhaustion, the large-scale consumption of traditional fossil energy has led to increasingly prominent problems of climate warming and environmental pollution. Wind energy, due to its large stock and wide distribution range, has become the "main force" in effectively adjusting the energy structure and accelerating the transformation process of the global energy system towards green and low-carbon. While wind energy penetrates into the power grid to alleviate energy shortages, its own volatility, intermittency and randomness also bring new challenges to the power grid. How to make good use of wind power has become an issue of concern. Improving the level of wind power prediction is one of the bases for solving this problem.

[0003] Wind power prediction models are mainly divided into physical models, traditional statistical models and machine learning models in terms of prediction principles. Physical models often rely on numerical weather prediction data, with a large amount of calculation, which may not be suitable for very short-term prediction and is more suitable for medium and long-term prediction. The traditional statistical model commonly used in wind power prediction at present is the time series analysis method, which has the characteristics of fast calculation speed and simple model. However, due to the often complex non-linear dynamic process of wind power changes, machine learning methods have been widely applied to wind power prediction, which can effectively make up for the deficiency of traditional statistical models relying solely on mathematical statistics to solve problems. However, there is currently a lack of a perfect technical solution. Summary of the Invention

[0004] To solve at least one of the technical problems existing in the prior art to a certain extent, an object of the present invention is to provide a very short-term wind power prediction method, device, electronic device and medium based on non-equal time-delay phase space reconstruction and spatio-temporal collaborative segments.

[0005] The first technical solution adopted by the present invention is:

[0006] A very short-term wind power prediction method, comprising the following steps:

[0007] Obtain the wind power time series, perform the mutual information method multiple times, and record the optimal delay time τ each time k , and use the multiple optimal delay time results as sequence elements to form a time-delay sequence T;

[0008] Construct phase spaces of various dimensions for the wind power time series according to the time-delay sequence T, and use the saturated correlation dimension method to find the optimal embedding dimension as the embedding dimension m for the final phase space reconstruction;

[0009] Perform non-equal time-delay phase space reconstruction on the wind power time series according to the time-delay sequence T and the embedding dimension m to obtain the phase space X m ;

[0010] Obtain the spatio-temporal collaborative fragment set X m from the obtained phase space X st , and construct the model input X st on the basis of the spatio-temporal collaborative fragment set X in ;

[0011] Use the obtained X in as the input of the model, train the model, and use the trained model to predict the future wind power value.

[0012] Furthermore, for the obtained wind power time series, perform the mutual information method multiple times, and record the optimal delay time τ k each time. Use the multiple optimal delay time results as sequence elements to form the time-delay sequence T, including:

[0013] Set the maximum embedding dimension n max , and use m max -1 as the number of times to perform the mutual information method;

[0014] Construct the operation sequence S k and the operation delay sequence Q k (τ) according to the wind power time series;

[0015] For the wind power time series, perform the mutual information method multiple times according to the operation sequence S k and the operation delay sequence Q k (τ), and record the optimal delay time τ k each time;

[0016] Use the obtained m max -1 optimal delay time results as sequence elements, and arrange them in the order of the number of times to form the time-delay sequence

[0017] Furthermore, the construction of the operation sequence S k and the operation delay sequence Q k (τ) according to the wind power time series includes:

[0018] For the wind power time series x(t) = {x1, x2,..., x N} with length N, the operation sequence S1 of the first mutual information method is the original wind power time series x(t), and the operation sequence S k of each subsequent time is based on the previous operation sequence S k-1 , and introduces the time delay τ k-1 according to the result of the previous mutual information method to obtain the sequence;

[0019] The operation sequence S k The expression of

[0020]

[0021] The operation delay sequence Q k (τ) is the delay sequence of the operation sequence S k at different delay times τ, and the operation delay sequence Q k (τ) has the following expression:

[0022]

[0023] Furthermore, for the wind power time series, according to the operation sequence S k and the operation delay sequence Q k (τ), the mutual information method is performed multiple times, and the optimal delay time τ is recorded each time k , including:

[0024] For the wind power time series x(t) = {x1, x2,..., x N} of length N, according to the operation sequence S k and the operation delay sequence Q k (τ), the mutual information method (Mutual Information, MI) is performed multiple times, and the optimal delay time τ is recorded each time k , with a sequence of length N s and the sequence at different time delays τ The mutual information calculation formula between them is:

[0025]

[0026] where P s (s i ) is the probability that s takes the value of s i , P q(τ) (q(τ) j ) is the probability that q(τ) takes the value of q(τ) j , and P sq(τ) (s i , q(τ) j ) is the probability that s and q(τ) simultaneously take s i and q(τ) j ;

[0027] The optimal delay time τ k is the delay time corresponding to the first local minimum point of I(Q k (τ), S k ).

[0028] Further, the phase space X m has the following expression:

[0029]

[0030] wherein, the number of points (number of rows) n of the phase space X m is:

[0031]

[0032] In the formula, N is the length of the wind power time series, and τ i is the i-th optimal delay time in the delay sequence T.

[0033] Further, obtaining the spatio-temporal collaborative fragment set X m from the obtained phase space X st , and constructing the model input X st on the basis of the spatio-temporal collaborative fragment set X in , including:

[0034] Segmenting the phase space X m by rows according to a preset fragment size k mini and a sliding step c to obtain a phase space fragment set X frag ;

[0035] Setting the neighboring target point X(i * ) as the last known point X(n) in the phase space X m , setting the number of points k contained in the spatio-temporal collaborative fragment set, and calculating the distance d frag between each fragment in the phase space fragment set X * and the neighboring target point X(i frag ); According to the fragment size k mini , the k / k mini nearest fragments form the spatio-temporal collaborative fragment set X st ;

[0036] Concatenating the fragments in the spatio-temporal collaborative fragment set X st by rows, and concatenating them by rows according to the fragment size k mini and the number of points k contained in the spatio-temporal collaborative fragment set, that is, converting the vector dimension from the original dimension (k / k st ,k mini ,m) of X mini to ((k / k mini )×k mini ,m) to obtain the model input X in with the dimension of (k,m).

[0037] Further, the expression of the phase space fragment set X frag is:

[0038]

[0039] Among them, each segment X frag The expression of (i) is:

[0040]

[0041] Distance d frag The calculation formula of is:

[0042]

[0043] In the formula, d is the Euclidean distance between two points, and X frag (i) (1) refers to the first phase point of the phase space segment.

[0044] The second technical solution adopted by the present invention is:

[0045] A very short-term wind power prediction device, comprising:

[0046] A time-delay sequence acquisition module, configured to acquire the wind power time series, perform the mutual information method multiple times, record the optimal delay time τ each time k , and use the multiple optimal delay time results as sequence elements to form a time-delay sequence T;

[0047] An embedding dimension acquisition module, configured to construct phase spaces of multiple dimensions for the wind power time series according to the time-delay sequence T, and find the optimal embedding dimension as the embedding dimension m for the final phase space reconstruction;

[0048] A phase space reconstruction module, configured to perform non-uniform time-delay phase space reconstruction on the wind power time series according to the time-delay sequence T and the embedding dimension m to obtain a phase space X m ;

[0049] A spatio-temporal collaboration acquisition module, configured to obtain a spatio-temporal collaboration segment set X from the obtained phase space X m , and construct a model input X on the basis of the spatio-temporal collaboration segment set X st , and on the basis of the spatio-temporal collaboration segment set X st ; in A model training module, configured to use the obtained X

[0050] as the input of the model to train the model, and use the trained model to predict future wind power values. in

[0051] The third technical solution adopted by the present invention is:

[0052] ​An electronic device, the electronic device includes a processor and a memory, and at least one instruction, at least one program, a code set or an instruction set is stored in the memory. The at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement a short-term wind power prediction method as described above.

[0053] The fourth technical solution adopted by the present invention is:

[0054] A computer-readable storage medium, and at least one instruction, at least one program, a code set or an instruction set is stored in the storage medium. The at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by a processor to implement a short-term wind power prediction method as described above.

[0055] The fifth technical solution adopted by the present invention is:

[0056] A computer program product or a computer program, the computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes a short-term wind power prediction method as described above.

[0057] The present invention has the following beneficial effects compared with the prior art:

[0058] (1) Experiments show that through the non-equal time delay of phase space reconstruction, the predictability and independence of the wind power phase space of the present invention are better than those of the equal time delay phase space reconstruction.

[0059] (2) By simulating and verifying the short-term wind power prediction model based on non-equal time delay phase space reconstruction and spatio-temporal collaborative segments established by the present invention, it can be seen that the prediction results reach the ideal range. The average absolute error of the prediction results for the next 180 minutes (after normalization) is within 5%, and the root mean square error is within 8%. The accuracy is improved compared with the comparative method. Therefore, this method can ultimately improve the accuracy of short-term wind power prediction.

[0060] (3) The prediction accuracy improvement effect of the data preprocessing method (based on non-equal time delay phase space reconstruction and spatio-temporal collaborative segments) proposed by the present invention can be reflected in different prediction models, and has a certain model universality. Description of the Drawings

[0061] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following introduces the accompanying drawings of the relevant technical solutions in the embodiments of the present invention or the prior art. It should be understood that the accompanying drawings in the following introduction are only for clearly expressing some embodiments of the technical solutions in the present invention. For those skilled in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0062] Figure 1 Schematic flowchart of the ultra-short-term wind power prediction method based on non-equal time-delay phase space reconstruction and spatio-temporal collaborative segments in the embodiment of the present invention;

[0063] Figure 2 Schematic flowchart of the process of obtaining the time-delay sequence by the cyclic mutual information method in the embodiment of the present invention;

[0064] Figure 3 Schematic flowchart of constructing phase spaces of multiple dimensions in the embodiment of the present invention;

[0065] Figure 4 Schematic flowchart of non-equal time-delay phase space reconstruction in the embodiment of the present invention;

[0066] Figure 5 Schematic flowchart of phase space segment segmentation in the embodiment of the present invention;

[0067] Figure 6 Schematic flowchart of model input construction in the embodiment of the present invention;

[0068] Figure 7 Schematic diagram of the wind power time series of the entire system of the Belgian Transmission System Operator in the embodiment of the present invention;

[0069] Figure 8 Schematic flowchart of the steps of an ultra-short-term wind power prediction method in the embodiment of the present invention. Detailed implementation manners

[0070] The following details the embodiments of the present invention. The examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals represent the same or similar elements or elements with the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as a limitation of the present invention. For the step numbers in the following embodiments, they are only set for the convenience of explanation and illustration, and no limitation is imposed on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0071] In the description of the present invention, it should be understood that with regard to the orientation description, such as the orientation or positional relationship indicated by up, down, front, back, left, right, etc., it is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation to the present invention.

[0072] In the description of the present invention, the meaning of "several" is one or more, the meaning of "multiple" is two or more. Understandings such as "greater than", "less than", "exceeding", etc. do not include the present number, and understandings such as "above", "below", "within", etc. include the present number. If there is a description of "first" and "second", it is only for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features or implicitly indicating the sequence relationship of the indicated technical features.

[0073] In the description of the present invention, unless otherwise clearly defined, words such as "set", "installed", "connected", etc. should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above words in the present invention in combination with the specific content of the technical solution.

[0074] The inventor studied the existing achievements in improving the accuracy of wind power prediction and found the following:

[0075] (1) The accuracy of wind power prediction is not only related to the prediction model itself, but also related to the input of the model. The research on data preprocessing methods is also crucial. Research shows that the external irregular volatility of wind power is actually a reflection of internal chaotic dynamics, and its chaotic characteristics need to be reflected in a high-dimensional space. Currently, the phase space reconstruction method for wind power is basically based on the time-delay coordinate method proposed by Takens.

[0076] (2) When using the time-delay coordinate method for phase space reconstruction, two parameters, namely the embedding dimension and the delay time, need to be determined. Commonly used methods for estimating the embedding dimension include the false nearest neighbor method, Cao's method, and the saturated correlation dimension method, among which the saturated correlation dimension method is relatively more robust. Commonly used methods for estimating the delay time include the average displacement method, the autocorrelation function method, and the mutual information method (Mutual Information, MI). The mutual information method has advantages over the autocorrelation function method. Therefore, the mutual information method has also become a commonly used method for phase space reconstruction of wind power sequences. The principle for the mutual information method to select the delay time is that the information contained in adjacent dimensions in the phase space is as independent as possible and can restore the evolution characteristics of the phase points. Therefore, the delay time when the mutual information function first reaches the local minimum is selected as the optimal delay time for phase space reconstruction. However, this method operates on the entire sequence and can only obtain a unified delay time, without considering that the information contained in different segments of a sequence may be different, and the delay times corresponding to the first local minimum points of mutual information are also different.

[0077] (3) The process of reconstructing the phase space using the time-delay coordinate method is actually to obtain and rearrange segments of the time series. The obtained phase space contains low-dimensional time series features. At the same time, each phase point reflects strong correlation in space, and the phase space contains high-dimensional space features. However, current methods for constructing the model input based on the phase space often cannot take into account the spatio-temporal information of the phase space. For example, directly taking adjacent historical phase points in the evolution trajectory only considers the low-dimensional time series features, and selecting the neighboring point set of the prediction target in the known phase space point set as the input only considers the high-dimensional space features.

[0078] Therefore, in accordance with the two improvement ideas of non-equal time-delay of phase space reconstruction and spatio-temporal coordination of model input, the present invention proposes a method for ultra-short-term wind power prediction with non-equal time-delay of phase space reconstruction, spatio-temporal coordination of model input, higher prediction accuracy, and stronger application significance, so as to overcome the defects in existing ultra-short-term wind power prediction technologies.

[0079] Example 1

[0080] As Figure 1 and Figure 8 shown, this embodiment provides a method for ultra-short-term wind power prediction based on non-equal time-delay phase space reconstruction and spatio-temporal coordinated segments, including the following steps:

[0081] S1. Obtain the wind power time series, perform the mutual information method multiple times, record the optimal delay time τ k for each time, and use the multiple optimal delay time results as sequence elements to form the time-delay sequence T.

[0082] For the wind power time series x(t) = {x1, x2, …, x N}(where \(t\) represents the time), set the maximum embedding dimension \(m\). max , perform the cyclic mutual information method, that is, perform the mutual information method (Mutual Information, MI) multiple times, and record the optimal delay time \(\tau\) each time. k , take multiple optimal delay time results as sequence elements, arrange them in the order of the number of times, and form a time delay sequence \(T\).

[0083] See Figure 2 , in some embodiments, step S1 specifically includes the following steps:

[0084] S11. Set the number of cycles.

[0085] Specifically, set the maximum embedding dimension \(m\). max , with \(m\). max - 1 as the number of times to perform the mutual information method.

[0086] S12. Construct an operation sequence \(S\) and an operation delay sequence \(Q\) k and operation delay sequence \(Q\) k (\(\tau\)).

[0087] Specifically, for the wind power time series \(x(t)=\{x_1,x_2,\cdots,x\}\) of length \(N\). N}, the operation sequence \(S1\) of the first mutual information method is the original wind power time series \(x(t)\), and the operation sequence \(S\) k for each subsequent time is based on the previous operation sequence \(S\) k-1 , introduce a time delay \(\tau\) according to the result of the previous mutual information method. k-1 The obtained sequence, that is, the expression of \(S\) k is:

[0088]

[0089] The operation delay sequence \(Q\) k (\(\tau\)) is the delay sequence of the operation sequence \(S\) k at different delay times \(\tau\), and the expression of the operation delay sequence \(Q\) k (\(\tau\)) is:

[0090]

[0091] S13. Perform the mutual information method multiple times.

[0092] Specifically, for the wind power time series \(x(t)=\{x_1,x_2,\cdots,x\}\) of length \(N\). N}, according to the number of times obtained in step S11 and the operation sequence \(S\) k and the operation delay sequence \(Q\) k(τ) Perform the Mutual Information (MI) method multiple times and record the optimal delay time τ each time. k , with a length of N s sequence and the sequence at different time delays τ The mutual information calculation formula between them is:

[0093]

[0094] In the formula, P s (s i ) is the probability that s takes the value of s i , P q(τ) (q(τ) j ) is the probability that q(τ) takes the value of q(τ) j , P sq(τ) (s i , q(τ) j ) is the probability that s and q(τ) simultaneously take s i and q(τ) j ;

[0095] The optimal delay time τ k is the delay time corresponding to the first local minimum point of I(Q k (τ), S k ), satisfying:

[0096]

[0097] I(Q k (τ k ), S k ) ≤ I(Q k (τ), S k )

[0098] S14. Construct the time delay sequence.

[0099] Take the m max - 1 optimal delay time results obtained in step S13 as sequence elements and arrange them in the order of occurrence to form the time delay sequence

[0100] S2. Construct phase spaces of multiple dimensions for the wind power time series according to the time delay sequence T, and use the saturated correlation dimension method to find the optimal embedding dimension as the embedding dimension m for the final phase space reconstruction.

[0101] See Figure 3 , for example, for the wind power time series x(t) = {x1, x2,..., x N}, construct phase spaces of multiple dimensions based on the time-delay sequence T obtained in step S1, and use the saturated correlation dimension method to find the optimal embedding dimension, which is used as the embedding dimension m for the final phase space reconstruction.

[0102] S3. Perform non-equal time-delay phase space reconstruction on the wind power time series according to the time-delay sequence T and the embedding dimension m to obtain the phase space X m .

[0103] Specifically, for the wind power time series x(t) = {x1, x2, …, x N} with a length of N, perform non-equal time-delay phase space reconstruction according to the time-delay sequence result T of step S1 and the embedding dimension result m of step S2. As Figure 4 shown, obtain the phase space X m , and its expression is:

[0104]

[0105] Among them, the number of points (rows) n of the phase space X m is:

[0106]

[0107] In the formula, N is the length of the wind power time series, and τ i is the i-th optimal delay time in the time-delay sequence T.

[0108] S4. Obtain the spatio-temporal collaborative fragment set X m from the obtained phase space X st , and construct the model input X st on the basis of the spatio-temporal collaborative fragment set X in .

[0109] As an implementation manner, step S4 specifically includes the following steps:

[0110] S41. Fragment segmentation of the phase space.

[0111] For the phase space X m with n points obtained in step S3, according to the set fragment size k mini and the sliding step c, perform fragment segmentation row by row. As Figure 5 shown, obtain the phase space fragment set X frag , and its expression is:

[0112]

[0113] Among them, the expression of each fragment X frag (i) is:

[0114]

[0115] S42. Acquisition of spatio-temporal collaborative segment set

[0116] Set the neighboring target point X(i*) as the phase space X obtained in step S3 m the last known point X(n) in, set the number of points k contained in the spatio-temporal collaborative segment set, and calculate the distance d between each segment in the phase space segment set X frag obtained in step S41 and the neighboring target point X(i*) frag , and the distance calculation formula is:

[0117]

[0118] where d is the Euclidean distance between two points, and X frag (i) (1) refers to the first phase point of the phase space segment X frag (i), and so on; according to the segment size k set in step S41 mini , the k / k mini nearest segments form the spatio-temporal collaborative segment set X st .

[0119] S43. Construction of model input

[0120] Refer to Figure 6 , and splice the segments in the spatio-temporal collaborative segment set X st obtained in step S42 row by row. According to the segment size k set in step S41 mini and the number of points k contained in the spatio-temporal collaborative segment set set in step S42, splicing row by row means converting the vector dimension from X st 's original dimension (k / k mini ,k mini ,m) to ((k / k mini )×k mini ,m), so as to obtain the model input X in with dimension (k,m).

[0121] The perspective of constructing the model input in step S4 is based on segments rather than single points, retaining certain low-dimensional temporal characteristics, and involving the calculation of distances between phase points in high-dimensional space, considering high-dimensional space characteristics, realizing the collaboration of spatio-temporal information. The spatial information here refers to the spatial information in the phase space, rather than the spatial information of the actual physical space.

[0122] S5. Use the obtained X in as the input of the model, train the model, and use the trained model to predict the future wind power value.

[0123] For the wind power time series x(t) = {x1, x2, …, x N} of length N, using X obtained in step S4 in as the model input, setting the prediction step size s, feeding it into the model for learning, training to obtain the optimal network parameters, and predicting the future wind power values X out = {x N+1 , x N+2 , …, x N+s}.

[0124] The ultra - short - term wind power prediction and its performance indicators of the above - mentioned method are described in detail below in combination with specific embodiments.

[0125] The wind power time series in this embodiment is the data of the entire system of the Belgian Transmission System Operator from January 1, 2018 to January 31, 2018. The time granularity of the data is 15 minutes, and the time series diagram is as Figure 7 shown.

[0126] Set the maximum embedding dimension m max = 8 in step S1. After steps S1 and S2, the optimal embedding dimension m = 5 is obtained, and the time - delay sequence T corresponding to the non - equal - time - delay phase - space reconstruction is T = {19, 19, 18, 19}. Set the segment size k mini = 4 and the sliding step c = 1 in step S41, the number of points k contained in the spatio - temporal collaborative segment set in step S42 is 16, and the prediction time scale is for predicting 60 / 120 / 180 minutes in advance.

[0127] The effect index of predictability uses the largest Lyapunov exponent, which can reflect the strength of chaos of a system. A system with stronger chaos is more sensitive to changes in initial conditions. A small change in initial conditions will lead to significantly different states of system evolution, and the predictability is poor. The effect of independence is judged by calculating the mutual information between adjacent dimensions in the phase space.

[0128] The predicted values need to be evaluated by different performance indicators. The following are several different prediction indicators, and the calculation formulas are as follows:

[0129] (1) Mean Absolute Error (MAE):

[0130]

[0131] where WS(i) is the true wind speed at time i, and WS(i) is the predicted wind speed at time i. The smaller this index, the higher the accuracy of the prediction algorithm, and N represents the total number of wind power data samples used for prediction.

[0132] (2) Root Mean Square Error (RMSE):

[0133]

[0134] The smaller this index is, the closer the predicted data of the prediction algorithm is to the characteristics of the real data.

[0135] 1) Perform equal-delay phase space reconstruction and non-equal-delay phase space reconstruction on the wind power time series of the same data set (Belgian Transmission System Operator), and compare the predictability and independence of the phase spaces obtained by the two.

[0136] Table 1 shows the comparison results of predictability. It can be seen from Table 1 that the largest Lyapunov exponent of the phase space obtained after non-equal-delay phase space reconstruction is smaller, the chaos is weaker, and the predictability is stronger:

[0137] Table 1 Comparison of the predictability of the phase space

[0138]

[0139] Table 2 shows the comparison results of independence. It can be seen from Table 2 that the mutual information between adjacent dimensions of the non-equal-delay phase space is smaller, the independence is stronger, and it more conforms to the selection principle of the mutual information method that "the information contained in adjacent dimensions in the phase space is as independent as possible".

[0140] Table 2 Comparison of the independence of the phase space

[0141]

[0142] 2) Perform non-equal-delay phase space reconstruction on the wind power time series of the same data set (Belgian Transmission System Operator). When constructing the model input, use the method based on time series features, the method based on space features, and the method of the present invention respectively, and compare the performance indicators of the prediction results of different methods with the method of the present invention. In order to verify the model universality of the method of the present invention, the prediction models are sequentially selected as Multilayer Perceptron (MLP), Convolutional Neural Network (CNN), and Attention-Based Model (ABM).

[0143] Table 3 shows the results of the comparison of the prediction performance of the method in this embodiment on MLP, CNN, and ABM, and the normalized evaluation indicators are used. It can be seen from Table 3 that the overall prediction effect of the method of the present invention is the best, and it has a certain model universality.

[0144] Table 3 Comparison of prediction performance

[0145]

[0146]

[0147] In summary, compared with the prior art, the method of this embodiment has the following beneficial effects:

[0148] 1) Through the non-equal time delay transformation of phase space reconstruction, the predictability and independence of the wind power phase space in the present invention are superior to those of the equal time delay phase space reconstruction.

[0149] 2) By simulating and verifying the ultra-short-term wind power prediction model established based on non-equal time delay phase space reconstruction and spatio-temporal cooperation segments in the present invention, it can be seen that the prediction results reach the ideal range. The average absolute error of the prediction results for the next 180 minutes (after normalization) is within 5%, and the root mean square error is within 8%, which is more accurate than the comparative method. Therefore, this method can ultimately improve the accuracy of ultra-short-term wind power prediction.

[0150] 3) The prediction accuracy improvement effect of the data preprocessing method (based on non-equal time delay phase space reconstruction and spatio-temporal cooperation segments) proposed in the present invention can be reflected in different prediction models, showing certain model universality.

[0151] Embodiment 2

[0152] This embodiment provides an ultra-short-term wind power prediction device, including:

[0153] A time delay sequence acquisition module, configured to acquire the wind power time sequence, perform the mutual information method multiple times, record the optimal delay time τ each time, k and use the multiple optimal delay time results as sequence elements to form a time delay sequence T;

[0154] An embedding dimension acquisition module, configured to construct phase spaces of multiple dimensions for the wind power time sequence according to the time delay sequence T, find the optimal embedding dimension, and use it as the embedding dimension m for the final phase space reconstruction;

[0155] A phase space reconstruction module, configured to perform non-equal time delay phase space reconstruction on the wind power time sequence according to the time delay sequence T and the embedding dimension m to obtain a phase space X m ;

[0156] A spatio-temporal cooperation acquisition module, configured to obtain a spatio-temporal cooperation segment set X from the obtained phase space X m , and construct a model input X on the basis of the spatio-temporal cooperation segment set X st , st ; in ;

[0157] A model training module, configured to use the obtained X in as the input of the model to train the model, and use the trained model to predict future wind power values.

[0158] Since this device is a very short-term wind power prediction device according to an embodiment of the present invention, and the principle of the device for solving problems is similar to that of the method, the implementation of this device can refer to the implementation process of the above method embodiment, and the repeated parts will not be described again.

[0159] Embodiment 3

[0160] An embodiment of the present invention further provides an electronic device, which includes a processor and a memory. At least one instruction, at least one segment of program, code set or instruction set is stored in the memory, and the at least one instruction, the at least one segment of program, the code set or instruction set is loaded and executed by the processor to implement a very short-term wind power prediction method as Figure 8 shown.

[0161] It can be understood that the memory may include a random access memory (RAM), or may also include a read-only memory (ROM). Optionally, the memory includes a non-transitory computer-readable storage medium. The memory can be used to store instructions, programs, codes, code sets or instruction sets. The memory may include a program storage area and a data storage area. Among them, the program storage area may store instructions for implementing an operating system, instructions for at least one function, instructions for implementing the above various method embodiments, etc.; the data storage area may store data created according to the use of the server, etc.

[0162] The processor may include one or more processing cores. The processor uses various interfaces and lines to connect various parts within the entire server, and by running or executing instructions, programs, code sets or instruction sets stored in the memory, and by calling data stored in the memory, it executes various functions of the server and processes data. Optionally, the processor may be implemented in at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), or programmable logic array (PLA). The processor may integrate one or several combinations of a central processing unit (CPU) and a modem, etc. Among them, the CPU mainly processes the operating system and application programs, etc.; the modem is used to process wireless communication. It can be understood that the above modem may not be integrated into the processor and may be implemented separately by a single chip.

[0163] Since the electronic device is the electronic device corresponding to a method for ultra-short-term wind power prediction according to an embodiment of the present invention, and the principle of the electronic device for solving problems is similar to that of the method, the implementation of the electronic device can refer to the implementation process of the above method embodiment, and the repeated parts will not be described again.

[0164] Embodiment 4

[0165] An embodiment of the present invention further provides a computer-readable storage medium, in which at least one instruction, at least one program, a code set or an instruction set is stored, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by a processor to implement Figure 8 a method for ultra-short-term wind power prediction as shown.

[0166] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a program. The program can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electrically-erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc memories, a magnetic disk memory, a tape memory, or any other computer-readable medium capable of carrying or storing data.

[0167] Since the storage medium is the storage medium corresponding to a method for ultra-short-term wind power prediction according to an embodiment of the present invention, and the principle of the storage medium for solving problems is similar to that of the method, the implementation of the storage medium can refer to the implementation process of the above method embodiment, and the repeated parts will not be described again.

[0168] Embodiment 5

[0169] In some possible embodiments, aspects of the method of the embodiments of the present invention can also be implemented in the form of a program product, which includes program code. When the program product runs on a computer device, the program code is used to cause the computer device to execute the steps of a very short-term wind power prediction method according to various exemplary embodiments described above in this specification. Among them, the executable computer program code or "code" for executing each embodiment can be written in high-level programming languages such as C, C++, C#, Python, Smalltalk, Java, JavaScript, Visual Basic, structured query language (e.g., Transact-SQL), Perl, or in various other programming languages.

[0170] It should be understood that each part of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one of the following techniques known in the art or a combination thereof can be used: discrete logic circuits with logic gate circuits for implementing logical functions on data signals, application-specific integrated circuits with appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0171] In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples.

[0172] The above embodiments are only for explaining the technical concept and characteristics of the present invention, and their purpose is to enable those of ordinary skill in the art to understand the content of the present invention and implement it accordingly, and cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made according to the essence of the content of the present invention should be covered by the protection scope of the present invention.

Claims

1. A very short-term wind power prediction method, characterized in that Including the following steps: Obtain the wind power time series, perform the mutual information method multiple times, and record the optimal delay time τ each time. k Use the multiple optimal delay time results as sequence elements to form the time delay sequence T. Construct phase spaces of multiple dimensions for the wind power time series according to the time delay sequence T, and find the optimal embedding dimension as the embedding dimension m for the final phase space reconstruction; Perform non-equidistant time-delay phase space reconstruction on the wind power time series according to the time-delay sequence \(t\) and the embedding dimension \(m\) to obtain the phase space \(X\). m ; Obtain the spatio-temporal collaborative segment set X from the obtained phase space X m and construct the model input X based on the spatio-temporal collaborative segment set X st ; st in ;​ To obtain the resulting X in As the input of the model, the model is trained, and the trained model is used to predict future wind power values.

2. The ultra-short-term wind power prediction method according to claim 1, wherein The wind power time series is obtained, and the mutual information method is performed multiple times, and the optimal delay time τ each time is recorded k , and multiple optimal delay time results are used as sequence elements to form a time delay sequence T, including: Set the maximum embedding dimension m max , and use m max - 1 as the number of times for the mutual information method; Construct the operation sequence S according to the wind power time series k and the operation delay sequence Q k (τ); For the wind power time series, according to the operation sequence S k and the operation delay sequence Q k (τ), perform the mutual information method multiple times and record the optimal delay time τ each time k ; Take the obtained m max -1 optimal delay time results as sequence elements, arranged in chronological order to form a time delay sequence 3. A very short-term wind power prediction method according to claim 2, characterized in that Construct the operation sequence S based on the wind power time series k and the operation delay sequence Q k (τ), including: For the wind power time series \(x(t)=\{x_1,x_2,\ldots,x_N\}\), the operation sequence \(S_1\) of the first mutual information method is the original wind power time series \(x(t)\), and the subsequent operation sequence \(S\) each time N is based on the previous operation sequence \(S\) k and introduces a time delay \(\tau\) according to the result of the previous mutual information method k-1 to obtain the sequence; k-1 ​ Operation sequence S k The expression is: Operation delay sequence Q k (τ) is the operation sequence S k The delay sequence at different delay times τ, the operation delay sequence Q k (τ) is expressed as:

4. The ultra-short-term wind power prediction method according to claim 2, characterized in that For the wind power time series, according to the operation sequence S k and the operation delay sequence Q k (τ), perform the mutual information method multiple times and record the optimal delay time τ each time k , including: For the wind power time series \(x(t)=\{x_1,x_2,\cdots,x\) N \}\) of length \(N\), perform the mutual information method multiple times according to the operation sequence \(S\) k and the operation delay sequence \(Q\) k \((\tau)\), and record the optimal delay time \(\tau\) each time k , with a sequence of length \(N\) s The mutual information calculation formula between the sequence and the sequence at different time delays \(\tau\) is as follows: Where, P s (s i ) is s the probability of taking the value of s i , P q(τ) (q(τ) j ) is the probability of q(τ) taking the value of q(τ) j , P sq(τ) (s i , q(τ) j ) is the probability that s and q(τ) simultaneously take the values of s i and q(τ) j ; Optimal delay time τ k is the delay time corresponding to the first local minimum point of I(Q k (τ), S k ).

5. A very short-term wind power prediction method according to claim 1, characterized in that The phase space X m has the following expression: Among them, the number of points n in the phase space X m is as follows: where N is the length of the wind power time series, and τ i is the i-th optimal delay time in the delay time series T.

6. The ultra-short-term wind power prediction method according to claim 1, wherein The obtained phase space X m is used to obtain the spatio-temporal collaborative fragment set X st . Based on the spatio-temporal collaborative fragment set X st , the model input X is constructed in , including: Phase space X m According to the preset segment size k mini and the sliding step c, perform segment division by row to obtain the phase space segment set X frag ; Set the neighboring target point X(i * ) as the last known point X(n) in the phase space X m . Set the number of points k contained in the spatio-temporal collaborative fragment set, and calculate the distances d frag between each fragment in the phase space fragment set X * and the neighboring target point X(i frag ); According to the fragment size k mini , the nearest k / k mini fragments form the spatio-temporal collaborative fragment set X st ; Concatenate the segments in the spatio-temporal collaborative segment set X st row by row. According to the segment size k mini and the number of points k contained in the spatio-temporal collaborative segment set, concatenate row by row, that is, convert the vector dimension from the original dimension of X st (k / k mini ,k mini ,m) to ((k / k mini )×k mini ,m), and obtain the model input X in with dimension (k,m).

7. A very short-term wind power prediction method according to claim 6, characterized in that, The phase space fragment set X frag has the following expression: Among them, each segment X frag (i) is expressed as: Distance d frag The calculation formula is as follows: where d is the Euclidean distance between two points, and X frag (i) (1) refers to the first phase point of the phase space segment.

8. An ultra-short-term wind power prediction device, characterized in that Including: The time delay sequence acquisition module is used to obtain the wind power time series, perform the mutual information method multiple times, and record the optimal delay time τ each time. k Using multiple optimal delay time results as sequence elements, a time delay sequence T is formed. An embedding dimension acquisition module, configured to construct phase spaces of multiple dimensions for the wind power time series according to the time delay sequence T, and find the optimal embedding dimension as the embedding dimension m for the final phase space reconstruction; The phase space reconstruction module is used to perform non-equidistant time-delay phase space reconstruction on the wind power time series according to the time-delay sequence T and the embedding dimension m, and obtain the phase space X m ; A spatio-temporal collaborative acquisition module for obtaining spatio-temporal collaborative fragment sets X from the obtained phase space X m and constructing model input X based on the spatio-temporal collaborative fragment sets X st from the spatio-temporal collaborative fragment sets X st ; in ; A model training module, which is used to obtain X in as the input of the model, train the model, and use the trained model to predict future wind power values.

9. An electronic device, characterized in that, The electronic device includes a processor and a memory. At least one instruction, at least one program, a code set or an instruction set is stored in the memory, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, At least one instruction, at least one program, a code set or an instruction set is stored in the storage medium, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by a processor to implement the method according to any one of claims 1 to 7.

Images