CN117689082A - Short-term wind power probability prediction method, system and storage medium - Google Patents
Short-term wind power probability prediction method, system and storage medium Download PDFInfo
- Publication number
- CN117689082A CN117689082A CN202410074006.7A CN202410074006A CN117689082A CN 117689082 A CN117689082 A CN 117689082A CN 202410074006 A CN202410074006 A CN 202410074006A CN 117689082 A CN117689082 A CN 117689082A
- Authority
- CN
- China
- Prior art keywords
- data
- wind power
- expression
- sample
- prediction
- 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 51
- 238000003860 storage Methods 0.000 title claims abstract description 9
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 25
- 238000012545 processing Methods 0.000 claims description 35
- 230000006870 function Effects 0.000 claims description 31
- 238000012549 training Methods 0.000 claims description 31
- 230000008569 process Effects 0.000 claims description 22
- 239000011159 matrix material Substances 0.000 claims description 18
- 238000007781 pre-processing Methods 0.000 claims description 18
- 230000007704 transition Effects 0.000 claims description 18
- 238000004364 calculation method Methods 0.000 claims description 17
- 230000002159 abnormal effect Effects 0.000 claims description 16
- 238000004140 cleaning Methods 0.000 claims description 16
- 238000013480 data collection Methods 0.000 claims description 15
- 238000013501 data transformation Methods 0.000 claims description 15
- 230000010354 integration Effects 0.000 claims description 15
- 238000011176 pooling Methods 0.000 claims description 15
- 238000004422 calculation algorithm Methods 0.000 claims description 14
- 238000013528 artificial neural network Methods 0.000 claims description 13
- 238000004458 analytical method Methods 0.000 claims description 10
- 230000001131 transforming effect Effects 0.000 claims description 10
- 238000010224 classification analysis Methods 0.000 claims description 7
- 238000010586 diagram Methods 0.000 claims description 7
- 238000013527 convolutional neural network Methods 0.000 claims description 6
- 238000005520 cutting process Methods 0.000 claims description 6
- 238000001514 detection method Methods 0.000 claims description 6
- 238000009826 distribution Methods 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 6
- 238000001228 spectrum Methods 0.000 claims description 6
- 230000009466 transformation Effects 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 5
- 238000010248 power generation Methods 0.000 claims description 4
- 241001334134 Rugopharynx epsilon Species 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 3
- 238000013459 approach Methods 0.000 claims description 3
- 230000001174 ascending effect Effects 0.000 claims description 3
- 238000004590 computer program Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 3
- 238000012544 monitoring process Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000011084 recovery Methods 0.000 claims description 3
- 238000005096 rolling process Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 3
- 238000012546 transfer Methods 0.000 claims description 3
- 230000005856 abnormality Effects 0.000 claims description 2
- 230000009191 jumping Effects 0.000 claims description 2
- 238000005457 optimization Methods 0.000 description 3
- 238000007418 data mining Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000005265 energy consumption Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000004888 barrier function Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Abstract
The invention discloses a short-term wind power probability prediction method, a system and a storage medium. The wind power data can be decomposed by utilizing signal decomposition to reduce the influence of fluctuation of the wind power data, so that the wind power prediction precision is improved.
Description
Technical Field
The invention belongs to the technical field of wind power prediction, and particularly relates to a short-term wind power probability prediction method, a short-term wind power probability prediction system and a storage medium.
Background
With the increase of energy consumption, the traditional fossil energy is excessively consumed, so that the problems of environmental pollution and the like are more and more serious, the worldwide energy consumption is rapidly increased, the contradiction between supply and demand is continuously worsened, and the development of new energy is taken as an effective way for coping with the energy problem. Among renewable energy sources, wind energy which is abundant in reserve, low in environmental requirement and very widely distributed is being developed and utilized by countries around the world gradually. And thermal power. The conventional power supplies such as hydropower and the like are different, wind power generation mainly captures wind energy through a wind turbine generator, then converts the wind energy into kinetic energy of a generator blade, and finally converts the kinetic energy into electric energy to be combined with a power grid for use, so that the magnitude of wind power is mainly determined by wind speed, but the wind speed has obvious randomness, fluctuation, intermittence and uncontrollability due to the restriction and influence of geographic factors and meteorological conditions, and the wind power also has strong random fluctuation. Large-scale wind power is integrated into a power grid, and a large amount of rotation reserve capacity is reserved for a conventional power supply to balance unknown wind power to meet load requirements, so that the running economy of a power system is reduced, and the power quality of the power grid is reduced.
Therefore, it is necessary to provide an accurate wind power prediction technology, and data support is provided for economic dispatching of a power system by predicting future wind power, so that a countermeasure strategy is conveniently made in advance for future risks.
Disclosure of Invention
In view of the above, the present invention provides a short-term wind power probability prediction method, system and storage medium for accurately predicting wind power and improving wind power utilization rate, which solves the above technical problems, and is implemented by adopting the following technical scheme.
In a first aspect, the invention provides a short-term wind power probability prediction method, which comprises the following steps:
acquiring wind power related data required by wind power plant prediction, wherein the wind power related data comprises historical data and geographic information;
obtaining original data through data collection, data cleaning, data integration, data protocol and data transformation of noise signals, missing data and abnormal data in historical data, wherein the data collection comprises determining the purpose of prediction and the content of required data, the data cleaning comprises deleting missing and faulty data from a database, the data integration comprises redundant processing and conflict data detection, the influence of invalid and wrong data on building a model is reduced through the data protocol, and the data transformation comprises converting the data into a form used for model training;
Carrying out data preprocessing on the original data to obtain sample data, wherein the data preprocessing comprises wind power data clustering processing and wind power data decomposition processing;
and extracting characteristic information of the sample data, inputting the characteristic information into a trained deep neural network for training to obtain a wind power prediction result, and performing classification analysis on the wind power prediction result to complete short-term wind power probability prediction.
As a preferable mode of the above technical solution, the wind power data clustering process includes:
presetting a sample data set X to be clustered, wherein the X comprises n samples, namely X= { X 1 ,x 2 ...x n Each sample has m-dimensional data, i.e., the ith sample vector is x i ＝{x i1 ,x i2 ...x im Index set I x ＝{1,2...n}；
Calculating the distance d between the samples ij The expression of (2) isSetting a cut-off distance d c And calculating the local density ρ of each sample i According to distance matrix d ij Comprising N=n (N-1)/2 distance values, sorting the distance values in ascending order, taking the distance value at the position of 2% of the number of rows of the ordered distance matrix as the cut-off distance, if the data set N<1000, local density ρ i The expression of (2) is +.>
If the data set n is more than or equal to 1000, the local density ρ i Can be calculated by adopting a truncated kernel function, and the corresponding expression isWhen d ij -d c <At 0, x (d) ij -d c ) =1; when d ij -d c When not less than 0, x (d) ij -d c )＝0；
The local densities are ordered in a descending order, and the ordered sequence is ρ= { ρ 1 ,ρ 2 ...ρ n Meeting the expression asThe expression for calculating the cluster center distance δ for each sample is +.>
Calculating an evaluation index gamma i And descending order is carried out on the clustering clusters to draw a decision diagram, and the clustering number c and the clustering center position Q= { Q are known according to the decision diagram 1 ,q 2 ...q c }，q∈I x ，γ i The expression of (C) is gamma i ＝ρ i *δ i I=1, 2..n, and an initial cluster center value z= { Z was obtained 1 ,Z 2 ...Z c }，Z k For the q-th data set X k The method comprises the steps of (1) substituting c sample values, k=1, 2..c, c clustering central values Z and the clustering number c in an original data set X into a clustering algorithm, and calculating the expression of a membership matrix U to beWherein r represents a fuzzy weight index, r.epsilon.1, +.infinity), d ik Represents the distance between the ith sample and the center of the kth cluster, U ik Representing the membership degree of the ith sample belonging to the kth class;
calculating and updating the expression of the clustering center Z asCalculating a clustering target value, judging whether the condition is met or whether the maximum iteration number is met, if so, entering the next step, otherwise, jumping to a calculation membership matrix, and clustering the target function with the expression of +.>
Classifying the samples according to a softening classification rule, and obtaining the maximum membership U of the sample i ik If U is satisfied ik >0.5+0.5c -1 Sample x i Uniquely divided into categories k; if it meets U ik >(c+δ) -1 Sample x i Can belong to other categories while being divided into categories k, wherein c represents the number of clusters and delta represents the degree of overlap; sample set X ′ After the division is completed, correspondingly classifying the deleted clustering center samples into c categories until the clustering is completed.
As a preferable mode of the above technical solution, the wind power data decomposition processing includes:
decomposing VMD to decompose wind power original data into a low frequency sequence and a high frequency sequence by adopting a variation mode, wherein the low frequency sequence reflects the overall fluctuation trend of the original power data, the high frequency component reflects a detail signal and a noise signal at corresponding moments, and the specific decomposition process of the VMD comprises the following steps:
calculation of Modal function u Using Hilbert transform k The expression of resolving the signal of (t) to obtain a single-sided spectrum isWherein (1)>
Decomposing signals and center frequencies thereof by modeTerm mixing, the expression of modulating each mode spectrum to the base band is +.>
According to the square L of the gradient direction of the modulated signal 2 The expression of the bandwidth of each mode is estimated by the norm
Wherein { u } k (t)}＝{u 1 (t),u 2 (t)...u k (t) } represents K modal components, { ω k }＝{ω 1 ,ω 2 ...ω k -represents the center frequency of each mode;
introducing a quadratic penalty functionAnd the Lagrangian function obtained by the Lagrangian multiplier λ is:
the variation problem is solved by adopting an alternate direction multiplier algorithm and the alternate updating is carried out Solving Lagrangian expression, C n Of the optimal solution of (a) of the modal component u k And a center frequency omega k Respectively expressed as->
As the optimization of the technical scheme, the method for obtaining the original data by collecting the noise signals, the missing data and the abnormal data in the historical data, cleaning the data, integrating the data, and transforming the data comprises the following steps:
constructing a noisy wind power data signal model, wherein the expression of the noisy wind power data signal model is P (x) =p (x) +omega (x), wherein P (x) represents an actual wind power data signal, omega (x) represents a white noise data signal, and P (x) represents a noisy wind power data signal;
wavelet transforming noisy wind power data signals to satisfy C in space ψ Function or signal ψ (x), corresponding to the expressionWherein R is + Representing all positive real numbers, < >>Representing the fourier transform of ψ (x), ψ (a, b) (x) representing the continuous wavelet operation function automatically generated by the wavelet mother function ψ (x) depending on the parameters a and b, a representing the elastic telescoping influence factor, b representing the parallel movement influence factor;
the data signal f (x) is continuously wavelet transformed by the expression The expression of inversely transforming the data signal with the automatic recovery adjustment and the data signal repeated a plurality of times is that The computational expression for discrete wavelet transformation of signal f (x) is The expression of inverse transforming the automatically restored adjusted data signal and the reconstructed data signal is +.>Where B represents a constant number independent of the data signal.
As a preferable mode of the technical scheme, the operation function f (x) ∈L 2 (R) is composed of low frequency components with identification rate of 2-n, and is composed of high frequency components with identification rate of 2-j (i is less than or equal to j is less than or equal to n), and the expression of reconstruction of data signal expansion is f (x) =A n +D n +D n-1 +...D 2 +D 1 Wherein f (x) represents a data signal, a represents a low-frequency component, D represents a high-frequency component, and n represents an analysis layer number;
comprehensively processing the expansion threshold value of the wind power signal after wavelet transformation, when meeting the requirement of |c i When the I is less than or equal to tau, the corresponding expression isWhen meeting |c i |>At τ, its calculation expression is +.> τ represents a threshold value, corresponding toThe expression of (2) is +.>Wherein c i Representing the original wavelet transform coefficients, < >>Represents the wavelet transform coefficients after the limit value overall processing, M represents the number of wavelet transform coefficients of the corresponding scale, σ represents the standard deviation of the noise data signal.
As a preferable mode of the above-described aspect, the processing of the abnormal data and the missing data includes:
the calculation expression of the output power of the preset wind power generation is that Wherein p represents the impeller power, ρ represents the air density, and C ρ The method for monitoring the abnormal data by using the isolated forest algorithm comprises the following steps of:
the first step: from a given n samples of data y= { Y to be measured 1 ,y 2 ...y n Random selection in }Taking the data points as sub-samples, and putting the sub-samples into a root node of an isolated binary tree;
and a second step of: randomly selecting sample attribute q, randomly generating a cutting point p in the current node data range, and generating p epsilon q min ,q max ]；
And a third step of: dividing a hyperplane by taking the cut point as a basis, dividing the current node data space into two subspaces, placing a point smaller than p under the current selected attribute on the left branch of the node, and otherwise placing the point on the right branch of the node;
fourth step: recursively executing the second and third steps on the left branch and the right branch of the node, and continuously constructing new leaf nodes until any one of the following is satisfiedThe following conditions are: only one data is arranged on the leaf node, and the segmentation can not be continued; the height of the tree reaches the set height
Repeating the first step to the fourth step, wherein each cutting process is random, and t isolated tree forms are obtained to form a forest;
For each sample y i Comprehensively calculating the result of each tree, wherein the result is expressed asWhere h (x) represents the path length required for sample y to be isolated for each tree t, i.e., the number of branches that sample y passes from the root node to the leaf node of a single isolated tree; e (h (y)) represents the average path length required for an isolated sample y in an isolated forest; c (n) represents the average path length of an isolated tree constructed from n sample data, and is used for normalizing the path length h (y) of the sample y, and the calculation expression of c (n) is +.>Wherein H (i) represents a sum and a corresponding expression of H (i) ≡ln (i) = 0.5772, if the anomaly score s (y, n) is closer to 1, the depth of the data point in the isolated tree branch is smaller, and the data point is judged to be an anomaly point; when the abnormality score approaches 0, indicating that the data point is large in the depth of the isolated tree branch, the data point is determined as a normal point.
As the optimization of the technical scheme, extracting the characteristic information of the sample data and inputting the characteristic information into the trained deep neural network for training to obtain the wind power prediction result, comprising the following steps:
the VMD processed modal component is used as an input sample of a prediction model, feature extraction is carried out through a convolution layer and a pooling layer in CNN, and the specific process of training by adopting the VMD-CNN-LSTM model comprises the following steps:
Inputting original power data, decomposing the original power data through VMD, determining a decomposition mode number K by calculating an energy entropy value before the original power data is decomposed into K components;
each subsequence after VMD decomposition is used as input data to train in a CNN-LSTM model, wherein effective expression characteristics are obtained by adding a CNN convolution layer and a pooling layer into the size of an input data sample, then feature fusion is carried out by a full-connection layer, the size and the number of convolution kernels are initialized, the optimal size of the convolution kernels is selected by continuously changing the size of the convolution kernels in the pooling layer, a ReLU function is selected by a convolution layer activation function, the maximum pooling mode is selected by the pooling layer, and network training is carried out by adding the LSTM model so as to optimize training times and learning rate;
and respectively predicting the modal components to obtain respective prediction results, carrying out equal-weight superposition reconstruction, and finally carrying out error evaluation and comparison analysis on the output results to realize final wind power prediction.
As a preferable mode of the above technical solution, the predictive analysis of the predictive model using a markov chain specifically includes:
if arbitrary random process { x (t) n-1 ) t.epsilon.T } for any finite time sequence { x } (t1) ，x (t2) …x (tn) A corresponding state of b 1 ，b 2 …b m E, B, the corresponding expression is:
P{x(t n )≤b m |x(t n-1 )，x(t n-2 )…x(t 1 )}＝P{x(t n )≤b m |x(t n-1 ) And { x (t) n-1 ) T e T is a markov process;
if the random process { x (T), T ε T }, for any integer and state b 1 ，b 2 …b m E B, if its state transition conditional probability satisfies P (x n ＝b m |x n-1 ＝b m-1 ，x n-2 ＝b m-2 ，...，x 1 ＝b 1 )＝P(x n ＝b m |x n-1 ＝b m-1 ) Then { x (T), t.epsilon.T } is a Markov chain and P is recorded ij(k) ＝P{x n+k ＝b j |x n ＝b i }, wherein P ij(k) Indicated as b in the initial state i Under the condition that the initial time is n,through transition of k steps, from output state b i Transition to state b j Which corresponds to a one-step state transition probability matrix ofLet->Wherein m represents the total number of states, and the molecule is state b i One step transfer to b j The denominator of the number of times of (a) is state b i The number of occurrences;
k-step transition probability based on Markov chain characteristicsPresetting an initial state as B (0), and setting a corresponding initial state vector as B 0 The calculation expression of the state probability distribution after the k steps is Q (k) =b 0 P k The specific steps of the construction of the wind power prediction model based on the Markov chain are as follows:
carrying out normalization processing and state division on wind power data to form a state sequence;
according to the expression asStatistically calculating a state transition probability matrix for a state sequence>
Determining initial state B (0), i.e. the corresponding state of the power value at the moment before the predicted moment, to obtain corresponding initial state vector B 0 ；
According to the expression Q (k) =b 0 P k Obtaining the power corresponding state probability distribution at the moment k, and taking the state corresponding to the maximum probability value as the predicted moment corresponding state;
and determining the predicted value as a state interval median corresponding to the predicted time, adding the obtained predicted value into the original sequence, and continuing rolling prediction.
In a second aspect, the present invention also provides a short-term wind power probability prediction system, including:
the data acquisition module is used for acquiring wind power related data required by wind power plant prediction, wherein the wind power related data comprises historical data and geographic information;
the data screening module is used for obtaining original data through data collection, data cleaning, data integration, data protocol and data transformation of noise signals, missing data and abnormal data in the historical data, wherein the data collection comprises determining the purpose of prediction and the content of required data, the data cleaning comprises deleting missing and faulty data from a database, the data integration comprises redundant processing and conflict data detection, the influence of invalid and wrong data on a constructed model is reduced through the data protocol, and the data transformation comprises converting the data into a form used for model training;
The data preprocessing module is used for carrying out data preprocessing on the original data to obtain sample data, wherein the data preprocessing comprises wind power data clustering processing and wind power data decomposition processing;
and the model training module is used for extracting characteristic information of the sample data, inputting the characteristic information into the trained deep neural network for training to obtain a wind power prediction result, and carrying out classification analysis on the wind power prediction result to complete short-term wind power probability prediction.
In a third aspect, the present invention also provides a computer readable storage medium storing a computer program for implementing the short-term wind power probability prediction method described above.
The invention provides a short-term wind power probability prediction method, a system and a storage medium. The wind power data can be decomposed by utilizing signal decomposition to reduce the influence of fluctuation of the wind power data, so that the wind power prediction precision is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a short-term wind power probability prediction method of the present invention;
FIG. 2 is a block diagram of a short-term wind power probability prediction system of the present invention.
Detailed Description
Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to like or similar elements or elements having like or similar functions throughout. The embodiments described below by referring to the drawings are illustrative only and are not to be construed as limiting the invention.
Referring to fig. 1, the invention provides a short-term wind power probability prediction method, which comprises the following steps:
s1: acquiring wind power related data required by wind power plant prediction, wherein the wind power related data comprises historical data and geographic information;
S2: obtaining original data through data collection, data cleaning, data integration, data protocol and data transformation of noise signals, missing data and abnormal data in historical data, wherein the data collection comprises determining the purpose of prediction and the content of required data, the data cleaning comprises deleting missing and faulty data from a database, the data integration comprises redundant processing and conflict data detection, the influence of invalid and wrong data on building a model is reduced through the data protocol, and the data transformation comprises converting the data into a form used for model training;
s3: carrying out data preprocessing on the original data to obtain sample data, wherein the data preprocessing comprises wind power data clustering processing and wind power data decomposition processing;
s4: and extracting characteristic information of the sample data, inputting the characteristic information into a trained deep neural network for training to obtain a wind power prediction result, and performing classification analysis on the wind power prediction result to complete short-term wind power probability prediction.
In this embodiment, the wind power data clustering process includes:
presetting a sample data set X to be clustered, wherein the X comprises n samples, namely X= { X 1 ，x 2 …x n Each sample has m-dimensional data, i.e., the ith sample vector is x i ＝{x i1 ，x i2 …x im Index set I x ＝{1，2…n}；
Calculating the distance d between the samples ij The expression of (2) isSetting a cut-off distance d c And calculating the local density ρ of each sample i According to distance matrix d ij Comprising N=n (N-1)/2 distance values, sorting the distance values in ascending order, taking the distance value at the position of 2% of the number of rows of the ordered distance matrix as the cut-off distance, if the data set N<1000, local density ρ i The expression of (C) is ρ i ＝/>
If the data set n is more than or equal to 1000, the local density ρ i Can be calculated by adopting a truncated kernel function, and the corresponding expression isWhen d ij -d c <At 0, x (d) ij -d c ) =1; when d ij -d c When not less than 0, x (d) ij -d c )＝0；
The local densities are ordered in a descending order, and the ordered sequence is ρ= { ρ 1 ,ρ 2 ...ρ n Meeting the expression asThe expression for calculating the cluster center distance δ for each sample is +.>
Calculating an evaluation index gamma i And descending order is carried out on the clustering clusters to draw a decision diagram, and the clustering number c and the clustering center position Q= { Q are known according to the decision diagram 1 ,q 2 ...q c }，q∈I x ，γ i The expression of (C) is gamma i ＝ρ i *δ i I=1, 2..n, and an initial cluster center value z= { Z was obtained 1 ,Z 2 ...Z c }，Z k For the q-th data set X k The method comprises the steps of (1) substituting c sample values, k=1, 2..c, c clustering central values Z and the clustering number c in an original data set X into a clustering algorithm, and calculating the expression of a membership matrix U to beWherein r represents a fuzzy weight index, r.epsilon.1, +.infinity), d ik Represents the distance between the ith sample and the center of the kth cluster, U ik Representing the membership degree of the ith sample belonging to the kth class;
calculating and updating the expression of the clustering center Z asComputing clustered targetsThe value, judge whether to meet the condition or whether to reach the maximum iteration number, if yes, go to the next step, otherwise jump to calculate the membership matrix, the expression of the clustering objective function is +.>
Classifying the samples according to a softening classification rule, and obtaining the maximum membership U of the sample i ik If U is satisfied ik >0.5+0.5c -1 Sample x i Uniquely divided into categories k; if it meets U ik >(c+δ) -1 Sample x i Can belong to other categories while being divided into categories k, wherein c represents the number of clusters and delta represents the degree of overlap; after the sample set X' is divided, correspondingly classifying the deleted clustering center samples into c categories until the clustering is completed.
The wind power data decomposition processing includes:
decomposing VMD to decompose wind power original data into a low frequency sequence and a high frequency sequence by adopting a variation mode, wherein the low frequency sequence reflects the overall fluctuation trend of the original power data, the high frequency component reflects a detail signal and a noise signal at corresponding moments, and the specific decomposition process of the VMD comprises the following steps:
calculation of Modal function u Using Hilbert transform k The expression of resolving the signal of (t) to obtain a single-sided spectrum isWherein (1)>
Decomposing signals and center frequencies thereof by modeTerm mixing, the expression of modulating each mode spectrum to the base band is +.>
According to the square L of the gradient direction of the modulated signal 2 The expression of the bandwidth of each mode is estimated by the norm
Wherein { u } k (t)}＝{u 1 (t),u 2 (t)...u k (t) } represents K modal components, { ω k }＝{ω 1 ,ω 2 ...ω k -represents the center frequency of each mode;
introducing a quadratic penalty functionAnd the Lagrangian function obtained by the Lagrangian multiplier λ is:
the variation problem is solved by adopting an alternate direction multiplier algorithm and the alternate updating is carried out Solving Lagrangian expression, C n Of the optimal solution of (a) of the modal component u k And a center frequency omega k Respectively expressed as->
The wind power prediction mainly comprises three relatively independent integers of data mining, power prediction and data conversion, wherein the data mining comprises data selection and data preprocessing, and the power prediction comprises a prediction model construction and an optimization prediction model construction. The clustering algorithm generally adopts a clustering effectiveness index to determine the final clustering number, namely the clustering number is preset as a plurality of integers respectively, so that multiple clustering processing is carried out, and the effectiveness index after each clustering processing is calculated, so that the clustering number corresponding to the clustering result with better effectiveness is selected as the optimal clustering number. The variable-division modal decomposition VMD is a complex signal decomposition algorithm, can decompose a complex sequence into a plurality of signals with adjustable negative frequency, calculates the center frequency and bandwidth of each mode through an iterative algorithm, is a non-recursive decomposition algorithm, reduces the whole data nonlinearity after decomposition, and is beneficial to improving the model prediction accuracy because the fluctuation of local power variation tends to be gentle. Noise signals, missing data and abnormal data in historical data are subjected to data collection, data cleaning, data integration, data specification and data transformation to obtain original data, the original data is subjected to data preprocessing to obtain sample data, characteristic information of the sample data is extracted and input into a trained deep neural network to be trained to obtain a wind power prediction result, the wind power prediction result is subjected to classification analysis to complete short-term wind power probability prediction, data sample dimension is increased to conduct data diversity prediction, a prediction model is optimized according to different wind power data, and the applicability of the model is increased. The wind power data can be decomposed by utilizing signal decomposition to reduce the influence of fluctuation of the wind power data, so that the wind power prediction precision is improved.
Optionally, the noise signal, missing data and abnormal data in the historical data are subjected to data collection, data cleaning, data integration, data protocol and data transformation to obtain the original data, which comprises the following steps:
constructing a noisy wind power data signal model, wherein the expression of the noisy wind power data signal model is P (x) =p (x) +omega (x), wherein P (x) represents an actual wind power data signal, omega (x) represents a white noise data signal, and P (x) represents a noisy wind power data signal;
wavelet transforming noisy wind power data signals to satisfy C in space ψ Function or signal ψ (x), corresponding to the expressionWherein R is + Representing all positive real numbers, < >>Representing the fourier transform of ψ (x), ψ (a, b) (x) representing the continuous wavelet operation function automatically generated by the wavelet mother function ψ (x) depending on the parameters a and b, a representing the elastic telescoping influence factor, b representing the parallel movement influence factor;
the data signal f (x) is continuously wavelet transformed by the expression The expression of inversely transforming the data signal with the automatic recovery adjustment and the data signal repeated a plurality of times is thatThe computational expression for discrete wavelet transformation of signal f (x) is The expression of inverse transforming the automatically restored adjusted data signal and the reconstructed data signal is +. >Where B represents a constant number independent of the data signal.
In the present embodiment, the operation function f (x) ∈L 2 (R) is composed of low frequency components with identification rate of 2-n, and is composed of high frequency components with identification rate of 2-j (i is less than or equal to j is less than or equal to n), and the expression of reconstruction of data signal expansion is f (x) =A n +D n +D n-1 +...D 2 +D 1 Wherein f (x) represents a data signal, a represents a low-frequency component, D represents a high-frequency component, and n represents an analysis layer number;
comprehensively processing the expansion threshold value of the wind power signal after wavelet transformation, when meeting the requirement of |c i When the I is less than or equal to tau, the corresponding expression isWhen meeting |c i |>At τ, its calculation expression is +.> τ represents a threshold value, the corresponding expression is +.>Wherein c i Representing the original wavelet transform coefficients, < >>Represents the wavelet transform coefficients after the limit value overall processing, M represents the number of wavelet transform coefficients of the corresponding scale, σ represents the standard deviation of the noise data signal.
The processing of the abnormal data and the missing data includes:
the calculation expression of the output power of the preset wind power generation is thatWherein p represents the impeller power, ρ represents the air density, and C ρ The method for monitoring the abnormal data by using the isolated forest algorithm comprises the following steps of:
The first step: from a given n samples of data y= { Y to be measured 1 ,y 2 ...y n Random selection in }Taking the data points as sub-samples, and putting the sub-samples into a root node of an isolated binary tree;
and a second step of: randomly selecting sample attribute q, randomly generating a cutting point p in the current node data range, and generating p epsilon q min ,q max ]；
And a third step of: dividing a hyperplane by taking the cut point as a basis, dividing the current node data space into two subspaces, placing a point smaller than p under the current selected attribute on the left branch of the node, and otherwise placing the point on the right branch of the node;
fourth step: recursively executing the second step and the third step on the left branch node and the right branch node of the node, and continuously constructing new leaf nodes until any one of the following conditions is met: only one data is arranged on the leaf node, and the segmentation can not be continued; the height of the tree reaches the set height
Repeating the first step to the fourth step, wherein each cutting process is random, and t isolated tree forms are obtained to form a forest;
for each sample y i Comprehensively calculating the result of each tree, wherein the result is expressed asWhere h (x) represents the path length required for sample y to be isolated for each tree t, i.e., the number of branches that sample y passes from the root node to the leaf node of a single isolated tree; e (h (y)) represents the average path length required for an isolated sample y in an isolated forest; c (n) represents the average path length of an isolated tree constructed from n sample data, and is used for normalizing the path length h (y) of the sample y, and the calculation expression of c (n) is +. >Wherein H (i) represents a sum and a corresponding expression of H (i) ≡ln (i) = 0.5772, if the anomaly score s (y, n) is closer to 1, the depth of the data point in the isolated tree branch is smaller, and the data point is judged to be an anomaly point; when the anomaly score approaches 0, indicating that the data point is large in the depth of the orphan tree branch, the data point is determinedIs a normal point, thereby improving the accuracy of data preprocessing.
Optionally, extracting feature information of the sample data and inputting the feature information into a trained deep neural network for training to obtain a wind power prediction result, including:
the VMD processed modal component is used as an input sample of a prediction model, feature extraction is carried out through a convolution layer and a pooling layer in CNN, and the specific process of training by adopting the VMD-CNN-LSTM model comprises the following steps:
inputting original power data, decomposing the original power data through VMD, determining a decomposition mode number K by calculating an energy entropy value before the original power data is decomposed into K components;
each subsequence after VMD decomposition is used as input data to train in a CNN-LSTM model, wherein effective expression characteristics are obtained by adding a CNN convolution layer and a pooling layer into the size of an input data sample, then feature fusion is carried out by a full-connection layer, the size and the number of convolution kernels are initialized, the optimal size of the convolution kernels is selected by continuously changing the size of the convolution kernels in the pooling layer, a ReLU function is selected by a convolution layer activation function, the maximum pooling mode is selected by the pooling layer, and network training is carried out by adding the LSTM model so as to optimize training times and learning rate;
And respectively predicting the modal components to obtain respective prediction results, carrying out equal-weight superposition reconstruction, and finally carrying out error evaluation and comparison analysis on the output results to realize final wind power prediction.
In this embodiment, a markov chain is used to perform prediction analysis on a prediction model, which specifically includes:
if arbitrary random process { x (t) n-1 ) t.epsilon.T } for any finite time sequence { x } (t1) ,x (t2) …x (tn) A corresponding state of b 1 ，b 2 …b m E, B, the corresponding expression is:
P{x(t n )≤b m |x(t n-1 )，x(t n-2 )…x(t 1 )}＝P{x(t n )≤b m |x(t n-1 ) And { x (t) n-1 ) T e T is a markov process;
if the random process { x (T), T ε T }, for any integer and state b 1 ，b 2 …b m E B, if its state transition conditional probability satisfies P (x n ＝b m |x n-1 ＝b m-1 ,x n-2 ＝b m-2 ,...,x 1 ＝b 1 )＝P(x n ＝b m |x n-1 ＝b m-1 ) Then { x (T), t.epsilon.T } is a Markov chain and P is recorded ij(k) ＝P{x n+k ＝b j |x n ＝b i }, wherein P ij(k) Indicated as b in the initial state i Under the condition of n initial time, through k steps of transition, the output state b i Transition to state b j Which corresponds to a one-step state transition probability matrix ofLet->Wherein m represents the total number of states, and the molecule is state b i One step transfer to b j The denominator of the number of times of (a) is state b i The number of occurrences;
k-step transition probability based on Markov chain characteristicsPresetting an initial state as B (0), and setting a corresponding initial state vector as B 0 The calculation expression of the state probability distribution after the k steps is Q (k) =b 0 P k The specific steps of the construction of the wind power prediction model based on the Markov chain are as follows:
carrying out normalization processing and state division on wind power data to form a state sequence;
according to the expression asStatistically calculating a state transition probability matrix for a state sequence>
Determining initial state B (0), i.e. the corresponding state of the power value at the moment before the predicted moment, to obtain corresponding initial state vector B 0 ；
According to the expression Q (k) =b 0 P k Obtaining the power corresponding state probability distribution at the moment k, and taking the state corresponding to the maximum probability value as the predicted moment corresponding state;
and determining the predicted value as a state interval median corresponding to the predicted time, adding the obtained predicted value into the original sequence, and continuing rolling prediction.
It should be noted that the markov chain is a dynamic analysis algorithm suitable for time series, and has better application to time series prediction with larger randomness and volatility, so that the application of the markov chain in wind power prediction is also suitable. If the state cannot be obtained at a certain moment, the random process with the characteristic is called a Markov process, wherein the Markov chain is a special Markov process with discrete time and state. Firstly, training sample data is input in an input layer, then the processed training sample data is transmitted to an output layer through an implicit layer, when the error between actual output and ideal output exceeds the expected value, the back propagation process of the error is started, error signals of the output data are generated from the output layer and sequentially propagated to the implicit layer and the input layer according to the original path, the weight of each layer is corrected according to the gradient descending mode through neurons of each layer by the error signals, and the weight of each layer is continuously adjusted through continuous forward information propagation and back error propagation, so that the robustness of neural network training is improved.
Referring to fig. 2, the invention further provides a short-term wind power probability prediction system, which comprises:
the data acquisition module is used for acquiring wind power related data required by wind power plant prediction, wherein the wind power related data comprises historical data and geographic information;
the data screening module is used for obtaining original data through data collection, data cleaning, data integration, data protocol and data transformation of noise signals, missing data and abnormal data in the historical data, wherein the data collection comprises determining the purpose of prediction and the content of required data, the data cleaning comprises deleting missing and faulty data from a database, the data integration comprises redundant processing and conflict data detection, the influence of invalid and wrong data on a constructed model is reduced through the data protocol, and the data transformation comprises converting the data into a form used for model training;
the data preprocessing module is used for carrying out data preprocessing on the original data to obtain sample data, wherein the data preprocessing comprises wind power data clustering processing and wind power data decomposition processing;
and the model training module is used for extracting characteristic information of the sample data, inputting the characteristic information into the trained deep neural network for training to obtain a wind power prediction result, and carrying out classification analysis on the wind power prediction result to complete short-term wind power probability prediction.
In this embodiment, there are a plurality of meteorological factors that can affect the wind power prediction accuracy, in which the meteorological factors that affect the wind power prediction accuracy are also many, but because of the limitation of theoretical development and application technology, if the meteorological information with excessive dimensions is used as the training data of the neural network, the neural network structure is complicated, the convergence speed can not even converge, so that a large amount of meteorological factor information becomes a barrier for improving the wind power prediction accuracy. The main component PCA is adopted to carry out data dimension reduction, a small amount of variables are used for reflecting information expressed by the whole variables, and the dimension of input information is reduced as much as possible under the condition of ensuring that the original information loss is smaller, so that the wind power prediction efficiency and precision are improved.
In another possible embodiment, the invention also provides a computer readable storage medium storing a computer program for implementing the short-term wind power probability prediction method.
Any particular values in all examples shown and described herein are to be construed as merely illustrative and not a limitation, and thus other examples of exemplary embodiments may have different values.
It should be noted that: like reference numerals and letters denote like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures.
The above examples merely represent a few embodiments of the present invention, which are described in more detail and are not to be construed as limiting the scope of the present invention. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the invention, which are all within the scope of the invention.
Claims (10)
1. The short-term wind power probability prediction method is characterized by comprising the following steps of:
acquiring wind power related data required by wind power plant prediction, wherein the wind power related data comprises historical data and geographic information;
obtaining original data through data collection, data cleaning, data integration, data protocol and data transformation of noise signals, missing data and abnormal data in historical data, wherein the data collection comprises determining the purpose of prediction and the content of required data, the data cleaning comprises deleting missing and faulty data from a database, the data integration comprises redundant processing and conflict data detection, the influence of invalid and wrong data on building a model is reduced through the data protocol, and the data transformation comprises converting the data into a form used for model training;
Carrying out data preprocessing on the original data to obtain sample data, wherein the data preprocessing comprises wind power data clustering processing and wind power data decomposition processing;
and extracting characteristic information of the sample data, inputting the characteristic information into a trained deep neural network for training to obtain a wind power prediction result, and performing classification analysis on the wind power prediction result to complete short-term wind power probability prediction.
2. The short-term wind power probability prediction method according to claim 1, wherein the wind power data clustering process comprises:
presetting a sample data set X to be clustered, wherein the X comprises n samples, namely X= { X 1 ,x 2 ...x n Each sample has m-dimensional data, i.e., the ith sample vector is x i ＝{x i1 ,x i2 ...x im Index set I x ＝{1,2...n}；
Calculating the distance d between the samples ij The expression of (2) isSetting a cut-off distance d c And calculating the local density ρ of each sample i According to distance matrix d ij Comprising N=n (N-1)/2 distance values, sorting the distance values in ascending order, taking the distance value at the position of 2% of the number of rows of the ordered distance matrix as the cut-off distance, if the data set N<1000, local density ρ i The expression of (2) is +.>
If the data set n is more than or equal to 1000, the local density ρ i Can be calculated by adopting a truncated kernel function, and the corresponding expression is When d ij -d c <At 0, x (d) ij -d c ) =1; when d ij -d c When not less than 0, x (d) ij -d c )＝0；
The local densities are ordered in a descending order, and the ordered sequence is ρ= { ρ 1 ,ρ 2 ...ρ n Meeting the expression asThe expression for calculating the cluster center distance δ for each sample is +.>
Calculating an evaluation index gamma i And descending order is carried out on the clustering clusters to draw a decision diagram, and the clustering number c and the clustering center position Q= { Q are known according to the decision diagram 1 ,q 2 ...q c }，q∈I x ，γ i The expression of (C) is gamma i ＝ρ i *δ i I=1, 2..n, and an initial cluster center value z= { Z was obtained 1 ,Z 2 ...Z c }，Z k For the q-th data set X k The method comprises the steps of (1) substituting c sample values, k=1, 2..c, c clustering central values Z and the clustering number c in an original data set X into a clustering algorithm, and calculating the expression of a membership matrix U to beWherein r represents a fuzzy weight index, r.epsilon.1, +.infinity), d ik Represents the distance between the ith sample and the center of the kth cluster, U ik Representing the membership degree of the ith sample belonging to the kth class;
calculating and updating the expression of the clustering center Z asCalculating a clustering target value, judging whether the condition is met or whether the maximum iteration number is met, if so, entering the next step, otherwise, jumping to a calculation membership matrix, and clustering the target function with the expression of +.>
Classifying the samples according to a softening classification rule, and obtaining the maximum membership U of the sample i ik If U is satisfied ik >0.5+0.5c -1 Sample x i Uniquely divided into categories k; if it meets U ik >(c+δ) -1 Sample x i Can belong to other categories while being divided into categories k, wherein c represents the number of clusters and delta represents the degree of overlap; after the sample set X' is divided, correspondingly classifying the deleted clustering center samples into c categories until the clustering is completed.
3. The short-term wind power probability prediction method according to claim 1, wherein the wind power data decomposition processing includes:
decomposing VMD to decompose wind power original data into a low frequency sequence and a high frequency sequence by adopting a variation mode, wherein the low frequency sequence reflects the overall fluctuation trend of the original power data, the high frequency component reflects a detail signal and a noise signal at corresponding moments, and the specific decomposition process of the VMD comprises the following steps:
calculation of Modal function u Using Hilbert transform k The expression of resolving the signal of (t) to obtain a single-sided spectrum isWherein (1)>
Decomposing signals and center frequencies thereof by modeTerm mixing, the expression of modulating each mode spectrum to the base band is +.>
According to the square L of the gradient direction of the modulated signal 2 Norm estimation of modesThe bandwidth of (a) is expressed as
Wherein { u } k (t)}＝{u 1 (t),u 2 (t)...u k (t) } represents K modal components, { ω k }＝{ω 1 ,ω 2 ...ω k -represents the center frequency of each mode;
Introducing a quadratic penalty functionAnd the Lagrangian function obtained by the Lagrangian multiplier λ is:
the variation problem is solved by adopting an alternate direction multiplier algorithm and the alternate updating is carried out Solving Lagrangian expression, C n Of the optimal solution of (a) of the modal component u k And a center frequency omega k Respectively expressed as->
4. The short-term wind power probability prediction method according to claim 1, wherein obtaining the original data from the noise signals, the missing data and the abnormal data in the historical data through data collection, data cleaning, data integration, data specification and data transformation comprises the following steps:
constructing a noisy wind power data signal model, wherein the expression of the noisy wind power data signal model is P (x) =p (x) +omega (x), wherein P (x) represents an actual wind power data signal, omega (x) represents a white noise data signal, and P (x) represents a noisy wind power data signal;
wavelet transforming noisy wind power data signals to satisfy C in space ψ Function or signal ψ (x), corresponding to the expressionWherein R is + Representing all positive real numbers, < >>Representing the fourier transform of ψ (x), ψ (a, b) (x) representing the continuous wavelet operation function automatically generated by the wavelet mother function ψ (x) depending on the parameters a and b, a representing the elastic telescoping influence factor, b representing the parallel movement influence factor;
The data signal f (x) is continuously wavelet transformed by the expression The expression of inversely transforming the data signal with the automatic recovery adjustment and the data signal repeated a plurality of times is thatThe computational expression for discrete wavelet transformation of signal f (x) is The expression of inverse transforming the automatically restored adjusted data signal and the reconstructed data signal is +.>Wherein B represents one and dataA constant number of signal independence.
5. The short-term wind power probability prediction method of claim 4, further comprising:
for the operation function f (x) E L 2 (R) is composed of low frequency components with identification rate of 2-n, and is composed of high frequency components with identification rate of 2-j (i is less than or equal to j is less than or equal to n), and the expression of reconstruction of data signal expansion is f (x) =A n +D n +D n-1 +...D 2 +D 1 Wherein f (x) represents a data signal, a represents a low-frequency component, D represents a high-frequency component, and n represents an analysis layer number;
comprehensively processing the expansion threshold value of the wind power signal after wavelet transformation, when meeting the requirement of |c i When the I is less than or equal to tau, the corresponding expression isWhen meeting |c i |>At τ, its calculation expression is +.>=0, τ represents a threshold value, corresponding to the expressionWherein c i Representing the original wavelet transform coefficients, < >>Represents the wavelet transform coefficients after the limit value overall processing, M represents the number of wavelet transform coefficients of the corresponding scale, σ represents the standard deviation of the noise data signal.
6. The short-term wind power probability prediction method according to claim 4, wherein the processing of the abnormal data and the missing data comprises:
the calculation expression of the output power of the preset wind power generation is thatWherein p represents the impeller power, ρ represents the air density, and C ρ The method for monitoring the abnormal data by using the isolated forest algorithm comprises the following steps of:
the first step: from a given n samples of data y= { Y to be measured 1 ,y 2 ...y n Random selection in }Taking the data points as sub-samples, and putting the sub-samples into a root node of an isolated binary tree;
and a second step of: randomly selecting sample attribute q, randomly generating a cutting point p in the current node data range, and generating p epsilon q min ,q max ]；
And a third step of: dividing a hyperplane by taking the cut point as a basis, dividing the current node data space into two subspaces, placing a point smaller than p under the current selected attribute on the left branch of the node, and otherwise placing the point on the right branch of the node;
fourth step: recursively executing the second step and the third step on the left branch node and the right branch node of the node, and continuously constructing new leaf nodes until any one of the following conditions is met: only one data is arranged on the leaf node, and the segmentation can not be continued; the height of the tree reaches the set height
Repeating the first step to the fourth step, wherein each cutting process is random, and t isolated tree forms are obtained to form a forest;
for each sample y i Comprehensively calculating the result of each tree, wherein the result is expressed asWhere h (x) represents the path length required for sample y to be isolated for each tree t, i.e., sample y is from the root of a single isolated treeThe number of branches traversed from node to leaf node; e (h (y)) represents the average path length required for an isolated sample y in an isolated forest; c (n) represents the average path length of an isolated tree constructed from n sample data, and is used for normalizing the path length h (y) of the sample y, and the calculation expression of c (n) is +.>Wherein H (i) represents a sum and a corresponding expression of H (i) ≡ln (i) = 0.5772, if the anomaly score s (y, n) is closer to 1, the depth of the data point in the isolated tree branch is smaller, and the data point is judged to be an anomaly point; when the abnormality score approaches 0, indicating that the data point is large in the depth of the isolated tree branch, the data point is determined as a normal point.
7. The short-term wind power probability prediction method according to claim 1, wherein the feature information of the sample data is extracted and input into a trained deep neural network for training to obtain a wind power prediction result, and the method comprises the following steps:
The VMD processed modal component is used as an input sample of a prediction model, feature extraction is carried out through a convolution layer and a pooling layer in CNN, and the specific process of training by adopting the VMD-CNN-LSTM model comprises the following steps:
inputting original power data, decomposing the original power data through VMD, determining a decomposition mode number K by calculating an energy entropy value before the original power data is decomposed into K components;
each subsequence after VMD decomposition is used as input data to train in a CNN-LSTM model, wherein effective expression characteristics are obtained by adding a CNN convolution layer and a pooling layer into the size of an input data sample, then feature fusion is carried out by a full-connection layer, the size and the number of convolution kernels are initialized, the optimal size of the convolution kernels is selected by continuously changing the size of the convolution kernels in the pooling layer, a ReLU function is selected by a convolution layer activation function, the maximum pooling mode is selected by the pooling layer, and network training is carried out by adding the LSTM model so as to optimize training times and learning rate;
and respectively predicting the modal components to obtain respective prediction results, carrying out equal-weight superposition reconstruction, and finally carrying out error evaluation and comparison analysis on the output results to realize final wind power prediction.
8. The short-term wind power probability prediction method according to claim 7, wherein the prediction model is subjected to prediction analysis by using a markov chain, and the method specifically comprises:
If arbitrary random process { x (t) n-1 ) t.epsilon.T } for any finite time sequence { x } (t1) ,x (t2) ...x (tn) A corresponding state of b 1 ,b 2 ...b m E, B, the corresponding expression is:
P{x(t n )≤b m |x(t n-1 ),x(t n-2 )...x(t 1 )}＝P{x(t n )≤b m |x(t n-1 ) And { x (t) n-1 ) T e T is a markov process;
if the random process { x (T), T ε T }, for any integer and state b 1 ,b 2 ...b m E B, if its state transition conditional probability satisfies P (x n ＝b m |x n-1 ＝b m-1 ,x n-2 ＝b m-2 ,...,x 1 ＝b 1 )＝P(x n ＝b m |x n-1 ＝b m-1 ) Then { x (T), t.epsilon.T } is a Markov chain and P is recorded ij(k) ＝P{x n+k ＝b j |x n ＝b i }, wherein P ij(k) Indicated as b in the initial state i Under the condition of n initial time, through k steps of transition, the output state b i Transition to state b j Which corresponds to a one-step state transition probability matrix ofLet->Wherein m represents the total number of states, and the molecule is state b i One step transfer to b j The denominator of the number of times of (a) is state b i The number of occurrences;
k-step transition probability based on Markov chain characteristicsPresetting an initial state as B (0), and setting a corresponding initial state vector as B 0 The calculation expression of the state probability distribution after the k steps is Q (k) =b 0 P k The specific steps of the construction of the wind power prediction model based on the Markov chain are as follows:
carrying out normalization processing and state division on wind power data to form a state sequence;
according to the expression asStatistically calculating a state transition probability matrix for a state sequence >
Determining initial state B (0), i.e. the corresponding state of the power value at the moment before the predicted moment, to obtain corresponding initial state vector B 0 ；
According to the expression Q (k) =b 0 P k Obtaining the power corresponding state probability distribution at the moment k, and taking the state corresponding to the maximum probability value as the predicted moment corresponding state;
and determining the predicted value as a state interval median corresponding to the predicted time, adding the obtained predicted value into the original sequence, and continuing rolling prediction.
9. A short-term wind power probability prediction system according to any one of claims 1 to 8, characterized by comprising:
the data acquisition module is used for acquiring wind power related data required by wind power plant prediction, wherein the wind power related data comprises historical data and geographic information;
the data screening module is used for obtaining original data through data collection, data cleaning, data integration, data protocol and data transformation of noise signals, missing data and abnormal data in the historical data, wherein the data collection comprises determining the purpose of prediction and the content of required data, the data cleaning comprises deleting missing and faulty data from a database, the data integration comprises redundant processing and conflict data detection, the influence of invalid and wrong data on a constructed model is reduced through the data protocol, and the data transformation comprises converting the data into a form used for model training;
The data preprocessing module is used for carrying out data preprocessing on the original data to obtain sample data, wherein the data preprocessing comprises wind power data clustering processing and wind power data decomposition processing;
and the model training module is used for extracting characteristic information of the sample data, inputting the characteristic information into the trained deep neural network for training to obtain a wind power prediction result, and carrying out classification analysis on the wind power prediction result to complete short-term wind power probability prediction.
10. A computer readable storage medium storing a computer program for implementing the short-term wind power probability prediction method according to any one of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410074006.7A CN117689082A (en) | 2024-01-17 | 2024-01-17 | Short-term wind power probability prediction method, system and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410074006.7A CN117689082A (en) | 2024-01-17 | 2024-01-17 | Short-term wind power probability prediction method, system and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117689082A true CN117689082A (en) | 2024-03-12 |
Family
ID=90139227
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410074006.7A Pending CN117689082A (en) | 2024-01-17 | 2024-01-17 | Short-term wind power probability prediction method, system and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117689082A (en) |
-
2024
- 2024-01-17 CN CN202410074006.7A patent/CN117689082A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101414366B (en) | Method for forecasting electric power system short-term load based on method for improving uttermost learning machine | |
| CN111414717A (en) | XGboost-L ightGBM-based unit power prediction method | |
| US11042802B2 (en) | System and method for hierarchically building predictive analytic models on a dataset | |
| CN110889603A (en) | Power system economic dispatching method considering wind power correlation based on PCA-Copula theory | |
| CN110717610B (en) | Wind power prediction method based on data mining | |
| CN108717579B (en) | Short-term wind power interval prediction method | |
| CN109471049B (en) | Satellite power supply system anomaly detection method based on improved stacked self-encoder | |
| CN114792156A (en) | Photovoltaic output power prediction method and system based on curve characteristic index clustering | |
| CN114792158A (en) | Multi-wind-farm short-term power prediction method based on space-time fusion graph neural network | |
| CN114021483A (en) | Ultra-short-term wind power prediction method based on time domain characteristics and XGboost | |
| CN116933175A (en) | Electric automobile charging load prediction method and device | |
| CN113627685B (en) | Wind driven generator power prediction method considering wind power internet load limit | |
| CN109886488B (en) | Distributed wind power plant layered hybrid short-term prediction method considering wind speed time lag | |
| CN113627674A (en) | Distributed photovoltaic power station output prediction method and device and storage medium | |
| CN112418504B (en) | Wind speed prediction method based on mixed variable selection optimization deep belief network | |
| Quan et al. | Short-term wind speed forecasting based on ensemble online sequential extreme learning machine and Bayesian optimization | |
| CN116937579A (en) | Wind power interval prediction considering space-time correlation and interpretable method thereof | |
| CN116842337A (en) | Transformer fault diagnosis method based on LightGBM (gallium nitride based) optimal characteristics and COA-CNN (chip on board) model | |
| CN117114160A (en) | Short-term photovoltaic power prediction method | |
| CN116070768A (en) | Short-term wind power prediction method based on data reconstruction and TCN-BiLSTM | |
| CN115526376A (en) | Multi-feature fusion generation countermeasure network ultra-short-term wind power prediction method | |
| CN117689082A (en) | Short-term wind power probability prediction method, system and storage medium | |
| CN115564155A (en) | Distributed wind turbine generator power prediction method and related equipment | |
| CN114298132A (en) | Wind power prediction method and device and electronic equipment | |
| CN114154684A (en) | Short-term photovoltaic power prediction method based on data mining and multi-core support vector machine |
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 |