CN115630316A - Ultrashort-term wind speed prediction method based on improved long-term and short-term memory network - Google Patents
Ultrashort-term wind speed prediction method based on improved long-term and short-term memory network Download PDFInfo
- Publication number
- CN115630316A CN115630316A CN202211305960.XA CN202211305960A CN115630316A CN 115630316 A CN115630316 A CN 115630316A CN 202211305960 A CN202211305960 A CN 202211305960A CN 115630316 A CN115630316 A CN 115630316A
- Authority
- CN
- China
- Prior art keywords
- wind speed
- sequence
- short term
- term
- short
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 98
- 230000007787 long-term memory Effects 0.000 title claims abstract description 21
- 230000006403 short-term memory Effects 0.000 title claims abstract description 20
- 238000012549 training Methods 0.000 claims abstract description 34
- 238000004458 analytical method Methods 0.000 claims abstract description 13
- 238000012545 processing Methods 0.000 claims abstract description 4
- 230000015654 memory Effects 0.000 claims description 35
- 230000006870 function Effects 0.000 claims description 25
- 230000000739 chaotic effect Effects 0.000 claims description 16
- 238000013528 artificial neural network Methods 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 12
- 238000012795 verification Methods 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 9
- 239000013598 vector Substances 0.000 claims description 5
- 241000282461 Canis lupus Species 0.000 claims description 4
- 238000005096 rolling process Methods 0.000 claims description 3
- 230000009897 systematic effect Effects 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 238000000605 extraction Methods 0.000 abstract description 3
- 230000005611 electricity Effects 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 12
- 238000004422 calculation algorithm Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 3
- 238000012804 iterative process Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 230000004913 activation Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000011217 control strategy Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013277 forecasting method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000010355 oscillation Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
Landscapes
- Environmental & Geological Engineering (AREA)
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Atmospheric Sciences (AREA)
- Biodiversity & Conservation Biology (AREA)
- Ecology (AREA)
- Environmental Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the technical field of electricity, and discloses an ultra-short-term wind speed prediction method based on an improved long-term and short-term memory network, which comprises the following steps: calculating the wind speed sequence delay time and the embedding dimension by a coordinate delay method; analyzing the chaos characteristic of the wind speed sequence; constructing a wind speed sequence reconstruction phase space based on a mutual information method; wind speed sequence standardization processing; training a long-term and short-term memory network; and based on the steps, carrying out ultra-short-term wind speed prediction. In the aspect of ultra-short-term wind speed prediction, the chaos characteristics of a wind speed sequence are judged through Lyapunov exponent analysis, a mutual information method is introduced to construct a wind speed sequence reconstruction phase space, wind speed sequence characteristic analysis is carried out, extraction of high-dimensional characteristics of the wind speed sequence is achieved, a long-term and short-term memory network is introduced into regression prediction, and accurate prediction of the wind speed is achieved.
Description
Technical Field
The invention relates to the technical field of electricity, in particular to an ultra-short-term wind speed prediction method based on an improved long-term and short-term memory network.
Background
The vigorous development and utilization of clean new energy such as wind, light and the like is a solid force for supporting the transformation development of energy in China, is an important measure for gradually improving the energy structure in China to realize high-quality green development, and the grid connection of renewable energy also brings a serious challenge to the safe and stable operation of a power system. High-precision wind speed prediction is used as a time sequence data forecasting method, and references can be provided for day-ahead and day-in scheduling of a power distribution network and generation of a micro-grid control instruction by predicting wind speed points or wind speed sequence data in the next regulation and control period, so that the running economy and stability of a power distribution system are improved.
The existing research mainly develops a great deal of discussion around the application of neural network algorithms and regression prediction algorithms and scenes such as interval prediction and spatial correlation analysis. However, existing research rarely analyzes inherent characteristics of a time sequence, relies on a large amount of wind speed or power data as a basis in the actual process of training a prediction model, fails to fully dig out internal characteristics contained in sequence changes, and is low in data utilization rate. Therefore, how to introduce a multidimensional data analysis method on the basis of the above research and perform model training by using the analysis result as input to fit the wind speed high-dimensional characteristics so as to construct a wind speed prediction method with wider applicable scene and higher prediction accuracy needs more exploration and research.
The document entitled "ultrashort-term wind speed prediction based on sequence-to-sequence network and attention mechanism" in the 9 th period of volume 42 of the domestic journal "solar academic newspaper discloses an ultrashort-term wind speed prediction method based on a sequence-to-sequence network and attention mechanism, which adopts a 1-dimensional convolutional neural network and a gating cycle unit to encode wind speed sequence data, and then uses the attention mechanism and the gating cycle unit to dynamically decode semantic vectors so as to obtain a predicted value. The model considers the adaptability of the wind speed sequence and a specific neural network, and improves the ultra-short-term wind speed prediction precision. However, the control method does not analyze and study inherent characteristics of the wind speed sequence, the original multi-mechanical characteristics of the wind speed cannot be considered in the prediction process, and the ultra-short-term wind speed prediction accuracy is low.
Disclosure of Invention
In order to solve the defects in the prior art, the invention provides an ultra-short-term wind speed prediction method and device based on an improved long-term and short-term memory network, a computer readable storage medium and computer equipment.
An ultra-short-term wind speed prediction method based on an improved long-term and short-term memory network comprises the following steps:
(1) Calculating the delay time and the embedding dimension of the wind speed sequence by a coordinate delay method;
(2) Analyzing the chaos characteristics of the wind speed sequence;
(3) Constructing a wind speed sequence reconstruction phase space based on a mutual information method;
(4) Wind speed sequence standardization treatment;
(5) Training a long-term and short-term memory network;
(6) And (5) carrying out ultra-short-term wind speed prediction based on the steps (1) to (5).
Preferably, the coordinate delay method in step (1) is applied to a one-dimensional chaotic sequence { x } i I =1,2,. So, N }, which reconstructs the general of the phase spaceThe expression form is:
wherein x is i Is the chaotic sequence element, m is the embedding dimension, τ is the delay time, N is the length of the phase space, where N = N- (m-1) τ, and N is the chaotic sequence length.
Preferably, the calculation of the wind speed sequence delay time by the coordinate delay method in step (1) includes:
time series x (t) is recorded, then it is x (t + τ) at delay time τ, and in case of known series x (t), x (t + τ) and its mutual information I (x (t + τ), x (t)) can be expressed as:
I(x(t+τ),x(t))＝H(x(t+τ))-H(x(t+τ)|x(t))
where i is the number of the elements in the time series x (t), j is the number of the elements in the time series x (t + τ), P x(t) Probability of x (t), P x(t),x(t+τ) Is the joint probability density of x (t) and x (t + τ), and H is an intermediate variable function.
Preferably, the calculation of the wind speed sequence embedding dimension by the coordinate delay method in the step (1) comprises:
recording the phase point vector in the m-dimensional phase space as X i The nearest neighbor point is X ji The distance between them is:
r m,i ＝||X i -X j,i ||
when the phase space dimension increases by 1, the distance between them is updated as:
if r is m+1,i Is significantly greater than r m,i Then it can be determined as a false nearest point;
defining an embedding dimension evolution function E 1 (m)：
If the time series is a series of definite systems, E 1 (m) after m reaches a certain value, the function tends to be stable, and the corresponding m is selected as a proper embedding dimension when the function is stable;
defining an embedding dimension auxiliary function E 2 (m):
The greater the randomness of the sequence, E 2 The smaller the fluctuation of the value of (m) around 1; when E is 2 (m) when the fluctuation is large, the sequence can be regarded as a chaotic time sequence.
Preferably, the wind speed sequence chaotic characteristic analysis in the step (2) includes:
the calculation by adopting the wolf method of the Lyapunov index comprises the following steps:
1) Based on the reconstruction method of the coordinate delay method in the step (1), if the embedding dimension of the phase space is determined to be m and the delay time is τ, the phase space constructed by the chaos time sequence with the length of N can be expressed as:
X＝{X i |i∈1,2,...,n}
X i ＝{x i ,x i+τ ,...,x i+(m-1)τ }
wherein N = N- (m-1) τ;
2) Selecting X 0 As an initial point, its markDistance between nearest neighbor points is L 0 Setting the distance threshold to epsilon, then advancing the time to a time t such that L 0 When is greater than epsilon, record X t Most adjacent point X of t ' and the distance value L at that time 0 ', and searching for X according to the principle of minimum included angle t ' when the distance between two points is less than epsilon, the distance value at this time is recorded as L i ；
3) Repeating the step 2) until all points in the phase space are traversed, recording the total iteration number as kappa, and then the systematic Lyapunov exponent is as follows:
preferably, the constructing a wind speed sequence reconstruction phase space based on the mutual information method in step (3) includes:
suppose that the wind speed sequence in a certain observation point in a certain time period is V = { V = i I =1, 2.·, N }, where N is the number of data points, the known wind speed sequence is reconstructed as the following phase space:
wherein v is i For wind speed sequence elements, m is the embedding dimension, τ is the delay time, N = N + (m-1) τ.
Preferably, the wind speed sequence normalization process in step (4) includes:
the wind speed sequence is normalized using the following formula:
wherein,
is the mean value of the wind speed sequence, and sigma (V) is the variance of the wind speed sequence.
Preferably, the training of the long-short term memory network in step (5) comprises:
selecting the first 90% of the known wind speed sequence as a neural network training set and the last 10% as a verification set, respectively expanding the training set and the verification set sequence by m and tau as phase spaces, wherein the phase space information of the first point in the phase space of the training set is taken, namely
As input to the neural network, the actual normalized wind speed at the second point in time, i.e.
And as output, training the network, and rolling forward according to time for iteration until the last point of the training set is used as output to participate in network training.
Preferably, the step (6) of developing ultra-short-term wind speed prediction based on the steps (1) to (5) comprises:
for the trained neural network, the phase space information of the first N points in the verification set is still used as input, the actual standardized wind speed of the (N + 1) th point is used as output, the output value is subjected to standardization and then is compared with the actual wind speed to obtain the prediction accuracy, the prediction accuracy is measured by root mean square error RMSE, and the calculation formula is as follows:
wherein v is pre k To predict wind speed, v real k Is the actual wind speed.
An ultra-short term wind speed prediction device based on an improved long-short term memory network comprises:
the computing module is used for computing the wind speed sequence delay time and the embedding dimension of the coordinate delay method;
the analysis module is used for analyzing the chaos characteristic of the wind speed sequence;
the construction module is used for constructing a wind speed sequence reconstruction phase space based on a mutual information method;
the wind speed sequence is subjected to standardization processing;
the training module is used for training the long-term and short-term memory network;
and the prediction module predicts the ultra-short term wind speed.
A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the method steps of the above.
A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor realizes the above-mentioned method steps when executing the computer program.
The invention has the following positive and beneficial effects:
the invention provides an ultrashort-term wind speed prediction method based on an improved long-short term memory network LSTM, which analyzes the chaos characteristic of a wind speed sequence and a phase space reconstruction method, and aims at the current situation that the traditional LSTM network prediction method cannot completely extract and utilize the characteristic of the wind speed sequence, a wind speed prediction model based on the LSTM and the phase space reconstruction is constructed based on the randomness and the chaos characteristic of the wind speed sequence, the chaos characteristic of the wind speed sequence and the effectiveness of the prediction method are verified by adopting the actually measured wind speed, the chaos characteristic of the wind speed sequence is judged by Lyapunov index analysis, the phase space is constructed by introducing a mutual information method, the wind speed sequence characteristic analysis is carried out, the extraction of the high-dimensional characteristic of the wind speed sequence is realized, the long-short term memory network is introduced in the regression prediction, the accurate prediction of the wind speed is realized, and the details are as follows:
1) The wind speed sequence phase space reconstruction method can effectively extract wind speed sequence characteristics and restore the original kinetic phase space of the wind speed sequence characteristics, and provides effective data support for a subsequent prediction process;
2) The wind speed prediction method based on the LSTM and the phase space reconstruction can memorize the internal characteristics of the wind speed sequence through the training of a large data volume, so that the high-dimensional data characteristics of the wind speed sequence are more efficiently utilized in the prediction process, and the error of the prediction result of the method is greatly reduced compared with that of the traditional LSTM method;
3) The prediction precision cannot be effectively improved by simply increasing the input dimension of the neural network, and in order to improve the learning efficiency and the prediction accuracy of the neural network, the method needs to perform operations such as decomposition, reconstruction and the like on the internal features of the wind speed sequence to extract the feature parameters of the wind speed sequence, so that the prediction precision is high.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and together with the description, serve to explain the principles of the invention.
FIG. 1 is a schematic view of a sequence of wind speeds;
FIG. 2 is a block diagram of a prediction method according to the present invention;
FIG. 3 is a diagram illustrating an iterative process for delay time calculation according to an embodiment of the present invention;
FIG. 4 is a diagram illustrating an iterative evolution of an embedding dimension according to an embodiment of the present invention;
FIG. 5 is a waveform diagram of the ultra-short term wind speed prediction time sequence under the prediction method of the present invention;
FIG. 6 is an analysis diagram of the ultra-short term wind speed prediction error under the prediction method of the present invention;
FIG. 7 is a waveform diagram of a super short-term wind speed prediction time sequence by using a Case1 method;
FIG. 8 is an analysis diagram of the ultra-short term wind speed prediction error by using the Case1 method;
FIG. 9 is a waveform diagram of the ultra-short term wind speed prediction time sequence by using the Case2 method;
FIG. 10 is an analysis diagram of the ultra-short term wind speed prediction error by using the Case2 method;
FIG. 11 is a diagram of a computer device according to an embodiment of the present invention.
Detailed Description
The following description and the drawings sufficiently illustrate specific embodiments herein to enable those skilled in the art to practice them. Portions and features of some embodiments may be included in or substituted for those of others. The scope of embodiments of the invention encompasses the full ambit of the claims, as well as all available equivalents of the claims. The terms "first," "second," and the like, herein are used solely to distinguish one element from another element without requiring or implying any actual such relationship or order between such elements. In fact, a first element could be termed a second element, and vice versa. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a structure, apparatus, or device that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such structure, apparatus, or device. Without further limitation, an element defined by the phrases "comprising a," "8230," "8230," or "comprising" does not exclude the presence of other like elements in a structure, device, or apparatus that comprises the element. The embodiments of the invention are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments can be referred to each other.
The terms "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," and the like in this disclosure refer to orientations or positional relationships illustrated in the drawings, which are used for convenience in describing the disclosure and to simplify the description, and do not indicate or imply that the referenced device or element must have a particular orientation, be constructed in a particular orientation, and be constructed in a particular manner of operation, and thus are not to be construed as limiting the disclosure. In the description of the present invention, unless otherwise specified and limited, the terms "mounted," "connected," and "connected" are to be construed broadly, and may be, for example, mechanically or electrically connected, or two elements may be interconnected, directly or indirectly through an intermediate, and the specific meaning of the terms may be understood by those skilled in the art according to their specific situation.
In the present invention, the term "plurality" means two or more unless otherwise specified.
In the present invention, the character "/" indicates that the preceding and following objects are in an "or" relationship. For example, A/B represents: a or B.
In the present invention, the term "and/or" is an association relationship describing objects, and means that there may be three relationships. For example, a and/or B, represents: a or B, or A and B.
FIG. 1 is a schematic view of a wind speed sequence. In this embodiment, the present invention provides a method for predicting an ultra-short term wind speed based on an improved long-short term memory network, which is shown in fig. 2 and includes:
(1) Calculating the wind speed sequence delay time and the embedding dimension of a coordinate delay method;
(2) Analyzing the chaos characteristics of the wind speed sequence;
(3) Constructing a wind speed sequence reconstruction phase space based on a mutual information method;
(4) Wind speed sequence standardization treatment;
(5) Training a long-term and short-term memory network;
(6) And (5) carrying out ultra-short-term wind speed prediction based on the steps (1) to (5).
The method of the embodiment designs a phase space reconstruction method based on a coordinate delay method, and realizes the restoration of high-dimensional dynamic characteristics of a wind speed sequence; and designing an ultra-short-term wind speed prediction method based on a long-term and short-term memory network to realize accurate prediction of the ultra-short-term wind speed.
In one embodiment, the coordinate delay method described in step (1) is applied to a one-dimensional chaotic sequence { x } i I =1,2, ·, N }, whose general expression of the reconstruction phase space is:
wherein x is i Is a chaotic sequence element, m is an embedding dimension, tau is a delay time, and N is the length of a phase space, wherein N = N- (m-1) tau, and N is the chaotic sequence length.
Further, the calculation of the wind speed sequence delay time by the coordinate delay method in the step (1) includes:
time series x (t) is recorded, then it is x (t + τ) at delay time τ, and in case of known series x (t), x (t + τ) and its mutual information I (x (t + τ), x (t)) can be expressed as:
I(x(t+τ),x(t))＝H(x(t+τ))-H(x(t+τ)|x(t))
where i is the number of the elements in the time series x (t), j is the number of the elements in the time series x (t + T), P x(t) Probability of x (t), P x(t),x(t+τ) Is the joint probability density of x (t) and x (t + τ), and H is an intermediate variable function.
Further, in the phase space reconstruction, it is desirable to achieve an effect that the correlation between different state point vectors is as small as possible, and thus the first minimum value of the mutual information function may be selected as the delay time setting value.
Further, the calculation of the wind speed sequence embedding dimension of the coordinate delay method in the step (1) comprises the following steps:
recording the phase point vector in the m-dimensional phase space as X i The closest point being X ji The distance between them is:
r m,i ＝||X i -X j,i || (2)
when the phase space dimension increases by 1, the distance between them is updated as:
if r m+1,i Is significantly greater than r m,i Then it can be determined as a false nearest point;
further, an embedding dimension evolution function E is defined 1 (m)：
If the time series is a series of definite systems, E 1 (m) will tend to be stable after m reaches a certain value, and the corresponding m is selected as the appropriate embedding dimension when the function is stable.
Further, inIn practical applications, it is difficult to distinguish E for a finite length sequence due to data volume limitations 1 (m) the steady state and the slowly growing state of the function, i.e. it is difficult to distinguish between chaotic sequences reaching steady state and random sequences tending to saturate, thus defining the embedded dimension auxiliary function E 2 (m):
The greater the randomness of the sequence, E 2 The smaller the fluctuation of the value of (m) around 1; when E is 2 (m) when the fluctuation is large, the sequence can be considered as a chaotic time sequence.
In one embodiment, the wind speed sequence chaotic characteristic analysis in the step (2) comprises:
the convergence property of the system and the sensitivity to initial conditions are reflected by the Lyapunov index of the system, and the wolf method for calculating the Lyapunov index comprises the following specific steps of:
1) Based on the reconstruction method, if the embedding dimension of the phase space is determined to be m and the delay time is τ, the phase space constructed by the chaotic time sequence with the length of N can be expressed as:
wherein N = N- (m-1) τ;
2) Selecting X 0 As an initial point, noting that the distance between the initial point and the nearest point is L 0 Setting the distance threshold to epsilon, then advancing the time to a time t such that L 0 When is greater than epsilon, record X t Most adjacent point X of t ' and the distance value L at this time 0 ', and searching for X according to the principle of minimum included angle t ' when the distance between two points is less than epsilon, the distance value at this time is recorded as L i ；
3) Repeating the step 2) until all points in the phase space are traversed, recording the total iteration number as kappa, and then the systematic Lyapunov exponent is as follows:
in one embodiment, the constructing the wind speed sequence reconstruction phase space in step (3) based on a mutual information method includes:
suppose that the wind speed sequence in a certain observation point in a certain time period is V = { V = i I =1,2,. Cndot, N }, where N is the number of data points. The known wind speed sequence can be reconstructed to the following phase space according to the foregoing theory:
wherein v is i For wind speed sequence elements, m is the embedding dimension, τ is the delay time, N = N + (m-1) τ.
In one embodiment, the wind speed sequence is normalized in step (4), and in order to prevent the neural network training result from diverging, the wind speed sequence needs to be normalized by the following formula:
wherein,
is the mean of the wind speed sequence, and σ (V) is the variance of the wind speed sequence.
In one embodiment, the training of the long-short term memory network in step (5) comprises:
the forgetting gate in the long and short term memory network outputs S to the last unit by controlling the unit t-1 The receiving probability of the contained information is realized, so that the selective forgetting of the information transmitted by the previous unit is realized, the activation function selects a Sigmoid function, and the forgetting gate can be expressed as follows:
f t ＝Sigmoid(w fS S t-1 +w fx x t +δ f ) (7)
wherein, w fS ，w fx Respectively passing through a forgetting gateThe last unit outputs the weight actually input by the unit, and f is an iteration parameter calculated for the forgetting gate.
Further, the input gate inputs x to the unit by controlling t The receiving degree of the contained information realizes the selective receiving of the input of the unit to the unit memory state, and simultaneously, the function of changing the unit memory state can be realized. The entry gate contains two ports, which use Sigmoid and tanh as activation functions, respectively:
wherein w iS ，w ix I is the weight of the last cell output and the actual input of the cell through the input gate, w is the offset parameter calculated for the input gate iS ，w ix The weights of the last unit output and the actual unit input for updating the memory state of the unit are respectively, and C is the state updating calculation bias parameter.
Further, the input gate has the function of updating the memory state of the cell, wherein the inheritance of the cell to the long-term memory state is realized by multiplying the forgetting gate by the long-term memory state, and the inheritance to the short-term memory state actually input by the cell is realized by multiplying the information of two ports of the input gate, wherein, the expression is multiplied by elements:
further, the output gate sets a channel for the output considering the long term memory and the output considering only the short term memory, respectively, wherein the output corresponding to the short term memory is:
o t ＝Sigmoid(w oS S t-1 +w ox x t +δ o ) (10)
the corresponding outputs of long-term memory are:
S t ＝o t *tanh(C t ) (11)
further, for the neural network composed of the above units, four sets of parameters are required to be trained, and as with the general RNN, the model is still solved by adopting the BPTT algorithm, which mainly comprises the following steps:
1) Initializing the parameters to be trained, calculating the output of each gate unit and the expected output of the current sequence index
Wherein
The calculation formula is as follows:
wherein, V and c are regression layer weight and bias, which are regarded as calculation auxiliary parameters in the unit and do not participate in the forward propagation process.
2) Defining a loss function J and a unit error term delta t ：
The forward transfer error term can be expressed as:
wherein:
the error gradient of the parameters to be optimized is then:
3) In iteration, the unit connection weights are updated by using the gradient descent principle:
w ij +λδ i x ij →w′ ij (17)
wherein λ is learning efficiency, and is generally reduced after the number of iterations reaches a threshold value, so as to avoid network non-convergence due to oscillation.
In the step (5), the first 90% of the known wind speed sequence is selected as a neural network training set, the last 10% is selected as a verification set, the training set and the verification set sequence are respectively expanded into phase spaces by m and tau, and phase space information of a first point in the phase space of the training set, namely the phase space information
As input to the neural network, the actual normalized wind speed at the second point in time, i.e.
Training the network as output, rolling forward according to time and iterating until the last point of the training set is used as output, participating in network training, and finding that the input quantity is a sequence with the length of m and the number of the sequence points is 0.9n-1.
Further, the step (6) of developing ultra-short-term wind speed prediction based on the above steps (1) to (5) includes:
for the trained neural network, the phase space information of the first N points in the verification set is still used as input, the actual standardized wind speed of the (N + 1) th point is used as output, the output value is subjected to standardization and then is compared with the actual wind speed to obtain the prediction accuracy, the prediction accuracy is measured by a Root Mean Square Error (RMSE), and the calculation formula is as follows:
wherein v is pre k To predict wind speed, v real k Is the actual wind speed.
In order to verify the effectiveness of the control strategy provided by the invention, a phase space reconstruction and LSTM prediction model is constructed and example verification is carried out on the basis of a Matlab2020b platform.
The iterative process of the mutual information function with the change of the delay time is shown in fig. 3, and it can be seen that when the iterative process has a first minimum value point at τ =30, the phase space reconstruction delay time is determined to be 30.
After the delay time is determined, iterative computation of the embedding dimension is performed. Setting the maximum embedding dimension to 50, the embedding dimension evolution function E increases with the embedding dimension 1 (m) and E 2 The variation of (m) is shown in FIG. 4. It can be seen that E is obtained when the embedding dimension is at the point m =9 1 (m) transition to Steady State, while E 2 The value of (m) still fluctuates around 1, and thus it can be determined that the present wind speed sequence is not a random sequence, and the embedding dimension of the phase space reconstruction is 10.
Based on the delay time and the embedding dimension, if the Lyapunov exponent is calculated by a wolf method and is 0.000265 & gt 0, the original wind speed sequence has chaotic characteristics, and the wind speed sequence can be analyzed by adopting a chaotic sequence method.
The wind speed prediction method based on the improved LSTM is used for constructing a prediction model, the delay time of a wind speed phase space is 30, the embedding dimension is 10, the LSTM network comprises four layers, namely a sequence input layer, an LSTM layer, a full connection layer and a regression layer, and the LSTM layer comprises 200 hidden units. And (3) the number of network training iterations is designated as 150, the gradient threshold value is 1, the initial learning rate is 0.005, the learning rate starts to decrease after 75 generations, and the decreasing rate is 0.2. Because of the superiority of Adam's algorithm in dealing with LSTM regression problems, adam's algorithm is used here for solving. The predicted wind speed curve and the actual wind speed curve are shown in FIG. 5, and the error distribution and root mean square error are shown in FIG. 6.
It can be seen that the fitting degree of the predicted wind speed curve and the actual wind speed curve is good, the root mean square error is 0.39484, and the wind speed prediction method based on the LSTM and the phase space reconstruction is high in precision.
In order to verify the superiority of the proposed method, the present invention sets two comparative examples Case1 and Case2 based on the LSTM prediction method, where the inputs both use actual normalized wind speed, where Case1 uses a single input prediction mode of predicting a wind speed at a later point from an actual wind speed at a previous point, case2 uses a multiple input prediction mode of predicting a wind speed at a later point from an actual wind speed at a previous 10 points, and the results are shown in fig. 7 and 8 and fig. 9 and 10, respectively. From the above results, it can be seen that the root mean square error of Case1 is 0.42746, and the root mean square error of Case2 is 0.63626, then the root mean square error of the prediction method provided by the present invention is smaller than that of Case1 and Case2, and from the error distribution diagram, the error distribution of the method provided by the present invention is more uniform, which indicates that the method has better extraction effect on sequence features and higher prediction precision. Meanwhile, it can be seen that the prediction effect of Case2 is worse than that of Case1, and it can be seen that the prediction accuracy cannot be improved by simply increasing the input dimension of the prediction model, and the prediction effect can be effectively improved by extracting the wind speed sequence characteristics and reflecting the characteristics into the input quantity in real time.
In one embodiment, a computer device is provided, which may be a server, and its internal structure is shown in fig. 11. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing static information and dynamic information data. The network interface of the computer device is used for communicating with an external terminal through a network connection. Which computer program is executed by a processor to carry out the steps in the above-described method embodiments.
Those skilled in the art will appreciate that the architecture shown in fig. 11 is merely a block diagram of some of the structures associated with the inventive arrangements and is not intended to limit the computing devices to which the inventive arrangements may be applied, as a particular computing device may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In an embodiment, a computer-readable storage medium is provided, on which a computer program is stored which, when being executed by a processor, carries out the steps of the above-mentioned method embodiments.
In an embodiment, there is also provided a computer device comprising a memory and a processor, the memory having stored therein a computer program, the processor implementing the steps of the above method embodiments when executing the computer program.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database or other medium used in embodiments provided herein may include at least one of non-volatile and volatile memory. Non-volatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical storage, or the like. Volatile Memory can include Random Access Memory (RAM) or external cache Memory. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others.
Finally, the above embodiments are only used for illustrating the technical solutions of the present invention and not for limiting, and other modifications or equivalent substitutions made by the technical solutions of the present invention by those of ordinary skill in the art should be covered within the scope of the claims of the present invention as long as they do not depart from the spirit and scope of the technical solutions of the present invention.
Claims (10)
1. An ultra-short term wind speed prediction method based on an improved long-short term memory network is characterized by comprising the following steps:
(1) Calculating the wind speed sequence delay time and the embedding dimension by a coordinate delay method;
(2) Analyzing the chaos characteristic of the wind speed sequence;
(3) Constructing a wind speed sequence reconstruction phase space based on a mutual information method;
(4) Wind speed sequence standardization processing;
(5) Training a long-term and short-term memory network;
(6) And (5) performing ultra-short-term wind speed prediction based on the steps (1) - (5).
2. The ultra-short term wind speed prediction method based on the improved long and short term memory network as claimed in claim 1, wherein the calculation of the wind speed sequence delay time by the coordinate delay method in step (1) comprises:
time series x (t) is recorded, then it is x (t + τ) at delay time τ, and in case of known series x (t), x (t + τ) and its mutual information I (x (t + τ), x (t)) can be expressed as:
I(x(t+τ),x(t))＝H(x(t+τ))-H(x(t+τ)|x(t))
where i is the number of the elements in the time series x (t), j is the number of the elements in the time series x (t + τ), P x(t) Probability of x (t), P x(t),x(t+τ) Is the joint probability density of x (t) and x (t + τ), and H is an intermediate variable function.
3. The ultra-short term wind speed prediction method based on the improved long-short term memory network as claimed in claim 1, wherein the calculation of the wind speed sequence embedding dimension of the coordinate delay method in step (1) comprises:
recording the phase point vector in the m-dimensional phase space as X i The nearest neighbor point is X ji The distance between them is:
r m,i ＝||X i -X j,i ||
when the phase space dimension increases by 1, the distance between them is updated as:
if r m+1,i Is significantly greater than r m,i Then it can be determined as a false nearest point;
defining an embedding dimension evolution function E 1 (m)：
If the time series is a series of definite systems, E 1 (m) the function tends to be stable after m reaches a certain value, and the m corresponding to the function is selected as a proper embedding dimension when the function is stable;
defining an embedding dimension auxiliary function E 2 (m):
The greater the randomness of the sequence, E 2 The smaller the fluctuation of the value of (m) around 1; when E is 2 (m) when the fluctuation is large, the sequence can be regarded as a chaotic time sequence.
4. The ultra-short term wind speed prediction method based on the improved long-short term memory network as claimed in claim 1, wherein the wind speed sequence chaotic characteristic analysis in the step (2) comprises:
the calculation by adopting the wolf method of the Lyapunov index comprises the following steps:
1) Based on the reconstruction method of the coordinate delay method in step (1), if it is determined that the embedding dimension of the phase space is m and the delay time is τ, the phase space constructed by the chaos time sequence with the length of N can be expressed as:
X＝{X i |i∈1,2,...,n}
X i ＝{x i ,x i+τ ,...,x i+(m-1)τ }
wherein N = N- (m-1) τ;
2) Selecting X 0 As an initial point, recording the distance between the most adjacent point and the initial point as L 0 Setting the distance threshold to epsilon, then advancing the time to a time t such that L 0 When is greater than epsilon, record X t Closest point X of (2) t ' and the distance value L at this time 0 ', and searching for X according to the principle of minimum included angle t ' when the distance between two points is less than epsilon, the distance value at this time is recorded as L i ；
3) Repeating the step 2) until all points in the phase space are traversed, and recording the total iteration times as kappa, wherein the systematic Lyapunov exponent is as follows:
5. the ultra-short term wind speed prediction method based on the improved long-short term memory network as claimed in claim 1, wherein the step (3) of constructing the wind speed sequence reconstruction phase space based on the mutual information method comprises:
suppose that the wind speed sequence in a certain observation point in a certain time period is V = { V = i I =1, 2.·, N }, where N is the number of data points, the known wind speed sequence is reconstructed as the following phase space:
wherein v is i For wind speed sequence elements, m is the embedding dimension, τ is the delay time, N = N + (m-1) τ.
6. The ultra-short term wind speed prediction method based on the improved long-short term memory network as claimed in claim 1, wherein the wind speed sequence normalization process in step (4) comprises:
the wind speed sequence is normalized using the following formula:
wherein,
is the mean value of the wind speed sequence, and sigma (V) is the variance of the wind speed sequence.
7. The ultra-short term wind speed prediction method based on the improved long-short term memory network as claimed in claim 1, wherein the training of the long-short term memory network in the step (5) comprises:
selecting the first 90% of the known wind speed sequence as a neural network training set and the last 10% as a verification set, respectively expanding the training set and the verification set sequence by m and tau as phase spaces, wherein the phase space information of the first point in the phase space of the training set is taken, namely
As input to the neural network, the actual normalized wind speed at the second point in time, i.e.
And as output, training the network, and rolling and iterating forwards according to time until the last point of the training set is used as output to participate in network training.
8. The ultra-short term wind speed prediction method based on the improved long and short term memory network as claimed in claim 1, wherein the step (6) of developing ultra-short term wind speed prediction based on the above steps (1) - (5) comprises:
for the trained neural network, before the verification centralization is still carried outThe phase space information of N points is used as input, the actual standardized wind speed of the (N + 1) th point is used as output, the output value is subjected to standardization and then is compared with the actual wind speed to obtain the prediction precision, the prediction precision is measured by a root mean square error RMSE, and the calculation formula is as follows:
wherein v is pre k To predict wind speed, v real k Is the actual wind speed.
9. An ultra-short term wind speed prediction device based on an improved long-short term memory network is characterized by comprising:
the computing module is used for computing the wind speed sequence delay time and the embedding dimension of the coordinate delay method;
the analysis module is used for analyzing the chaos characteristic of the wind speed sequence;
the construction module is used for constructing a wind speed sequence reconstruction phase space based on a mutual information method;
the wind speed sequence is subjected to standardization processing;
the training module is used for training the long-term and short-term memory network;
and the prediction module predicts the ultra-short term wind speed.
10. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor realizes the method steps of any of claims 1-8 when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211305960.XA CN115630316A (en) | 2022-10-24 | 2022-10-24 | Ultrashort-term wind speed prediction method based on improved long-term and short-term memory network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211305960.XA CN115630316A (en) | 2022-10-24 | 2022-10-24 | Ultrashort-term wind speed prediction method based on improved long-term and short-term memory network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115630316A true CN115630316A (en) | 2023-01-20 |
Family
ID=84906016
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211305960.XA Pending CN115630316A (en) | 2022-10-24 | 2022-10-24 | Ultrashort-term wind speed prediction method based on improved long-term and short-term memory network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115630316A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116992274A (en) * | 2023-09-28 | 2023-11-03 | 国网山东省电力公司滨州市滨城区供电公司 | Short-term wind speed prediction method and system based on improved principal component regression model |
-
2022
- 2022-10-24 CN CN202211305960.XA patent/CN115630316A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116992274A (en) * | 2023-09-28 | 2023-11-03 | 国网山东省电力公司滨州市滨城区供电公司 | Short-term wind speed prediction method and system based on improved principal component regression model |
| CN116992274B (en) * | 2023-09-28 | 2024-02-02 | 国网山东省电力公司滨州市滨城区供电公司 | Short-term wind speed prediction method and system based on improved principal component regression model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113544711B (en) | Hybrid algorithm system and method for using cluster contraction | |
| Wang et al. | The study and application of a novel hybrid forecasting model–A case study of wind speed forecasting in China | |
| Kumar et al. | A hybrid multi-agent based particle swarm optimization algorithm for economic power dispatch | |
| Karayiannis et al. | On the construction and training of reformulated radial basis function neural networks | |
| CN111401755B (en) | Multi-new-energy output scene generation method, device and system based on Markov chain | |
| CN110969290A (en) | Runoff probability prediction method and system based on deep learning | |
| CN116644970A (en) | Photovoltaic power prediction method based on VMD decomposition and lamination deep learning | |
| CN112865170B (en) | Scene probability-based load recovery optimization method considering wind power output correlation | |
| CN116345555A (en) | CNN-ISCA-LSTM model-based short-term photovoltaic power generation power prediction method | |
| Hu et al. | ℓ-DARTS: Light-weight differentiable architecture search with robustness enhancement strategy | |
| CN115630316A (en) | Ultrashort-term wind speed prediction method based on improved long-term and short-term memory network | |
| Zhang et al. | Reinforcement learning with actor-critic for knowledge graph reasoning | |
| Huang et al. | A grid and density based fast spatial clustering algorithm | |
| CN118040678A (en) | Short-term offshore wind power combination prediction method | |
| Osama et al. | An optimized support vector regression using whale optimization for long term wind speed forecasting | |
| CN115630286A (en) | Electric power prediction method, system and equipment based on hybrid deep learning model | |
| CN116721327A (en) | Neural network architecture searching method based on generalization boundary | |
| Nzale et al. | A tool for automatic determination of model parameters using particle swarm optimization | |
| CN113112092A (en) | Short-term probability density load prediction method, device, equipment and storage medium | |
| Li et al. | Complementary Fusion of Deep Spatio-Temporal Network and Tree Model for Wind Power Forecasting (Team: HIK) | |
| CN114722490A (en) | Agent model global optimization method based on mixed increase and interval reduction | |
| CN114757533A (en) | Power grid reactive power reserve demand rapid calculation method based on graph convolution deep network | |
| Liu et al. | Wind Power Prediction Model Based on WOA-BiLSTM-Attention | |
| CN112232482A (en) | Method and equipment for constructing neural network architecture search and intelligent equipment | |
| Jain et al. | Simulation model calibration with dynamic stratification and adaptive sampling |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |