CN117117968A - Wind farm climbing event prediction method based on data enhancement - Google Patents
Wind farm climbing event prediction method based on data enhancement Download PDFInfo
- Publication number
- CN117117968A CN117117968A CN202311097368.XA CN202311097368A CN117117968A CN 117117968 A CN117117968 A CN 117117968A CN 202311097368 A CN202311097368 A CN 202311097368A CN 117117968 A CN117117968 A CN 117117968A
- Authority
- CN
- China
- Prior art keywords
- climbing
- data
- wind power
- algorithm
- 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
- 230000009194 climbing Effects 0.000 title claims abstract description 122
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000012549 training Methods 0.000 claims abstract description 25
- 238000013507 mapping Methods 0.000 claims abstract description 17
- 238000001514 detection method Methods 0.000 claims abstract description 15
- 238000012360 testing method Methods 0.000 claims abstract description 14
- 238000012795 verification Methods 0.000 claims abstract description 14
- 230000000739 chaotic effect Effects 0.000 claims abstract description 11
- 238000004140 cleaning Methods 0.000 claims abstract description 8
- 230000006870 function Effects 0.000 claims description 24
- 230000001932 seasonal effect Effects 0.000 claims description 22
- 239000013598 vector Substances 0.000 claims description 12
- 230000035772 mutation Effects 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 10
- 230000019637 foraging behavior Effects 0.000 claims description 9
- 230000007246 mechanism Effects 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 9
- 238000009499 grossing Methods 0.000 claims description 7
- 238000005457 optimization Methods 0.000 claims description 7
- 230000002431 foraging effect Effects 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 6
- VDRGNAMREYBIHA-UHFFFAOYSA-N 2c-e Chemical group CCC1=CC(OC)=C(CCN)C=C1OC VDRGNAMREYBIHA-UHFFFAOYSA-N 0.000 claims description 3
- 230000008485 antagonism Effects 0.000 claims description 3
- 125000004122 cyclic group Chemical group 0.000 claims description 3
- 238000013016 damping Methods 0.000 claims description 3
- 238000013508 migration Methods 0.000 claims description 3
- 230000005012 migration Effects 0.000 claims description 3
- 238000011084 recovery Methods 0.000 claims description 3
- 230000000630 rising effect Effects 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000001174 ascending effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011423 initialization method Methods 0.000 description 1
- 230000009191 jumping Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0475—Generative networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/094—Adversarial learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/0985—Hyperparameter optimisation; Meta-learning; Learning-to-learn
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/08—Computing arrangements based on specific mathematical models using chaos models or non-linear system models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/004—Generation forecast, e.g. methods or systems for forecasting future energy generation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2123/00—Data types
- G06F2123/02—Data types in the time domain, e.g. time-series data
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/10—Power transmission or distribution systems management focussing at grid-level, e.g. load flow analysis, node profile computation, meshed network optimisation, active network management or spinning reserve management
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
- H02J2300/20—The dispersed energy generation being of renewable origin
- H02J2300/28—The renewable source being wind energy
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Business, Economics & Management (AREA)
- Power Engineering (AREA)
- Economics (AREA)
- Primary Health Care (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- Human Resources & Organizations (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Nonlinear Science (AREA)
- Algebra (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a wind farm climbing event prediction method based on data enhancement, which comprises the following steps: acquiring historical power data of a wind farm, and cleaning the data; extracting sectional trend of the wind power data obtained in the previous step by using an MK-sliding window detection method, and performing climbing detection; constructing a time sequence to generate an countermeasure network TimeGAN, carrying out data enhancement on detected wind power climbing data, and dividing the data into a training set, a verification set and a test set; establishing an ETSformer model, and inputting the obtained training set and verification set into the ETSformer model for training; improving an artificial humming algorithm AHA by adopting a Logistic chaotic mapping strategy and a Gaussian-cauchy mixed variation strategy to obtain an IAHA algorithm, optimizing the hyper-parameters of an ETSformer model by using the IAHA algorithm, and improving the accuracy of wind power climbing event prediction; and finally, predicting the wind power climbing event by using the test set. The invention can solve the problem of missing climbing data and improve the prediction accuracy.
Description
Technical Field
The invention relates to the technical field of wind farm climbing event prediction, in particular to a wind farm climbing event prediction method based on data enhancement.
Background
The renewable, pollution-free and wide-prospect characteristics of wind energy, etc., have become the fastest growing new energy in the world. However, wind power is affected by extreme weather in a short time, so that a wind power climbing event occurs when the output power is increased or reduced in a large scale, and great potential safety hazards or serious economic losses are caused for the operation of a power grid. In order to keep the grid system running stably, wind power hill climbing prediction is extremely necessary.
Currently, the prediction methods of the climbing event are classified into a direct prediction method and an indirect prediction method. The direct prediction method is to directly predict relevant parameters of climbing, such as climbing rate, climbing duration and the like, on the basis of historical climbing event records. The result of the direct prediction method is more visual and accurate, but a large amount of historical data is needed to train a model, and the wind power climbing prediction is different from the wind power prediction in that wind power climbing is a small probability event occurring in a wind power sequence, so that the prediction precision of the direct prediction method is greatly influenced by the lack of the historical data. The indirect prediction method is to extract data characteristics according to a prediction result to further judge whether a climbing event occurs or not on the basis of wind power prediction, and in the wind power prediction process, extreme data are often ignored in order to improve the overall prediction accuracy, so that climbing information is lost. Therefore, it is important for safe operation and economic dispatch of the power system to be able to obtain accurate prediction results of climbing events.
Disclosure of Invention
The invention aims to solve the technical problem of providing a wind farm climbing event prediction method based on data enhancement, which can solve the problem of climbing data missing and improve the prediction precision.
In order to solve the technical problems, the invention provides a wind farm climbing event prediction method based on data enhancement, which comprises the following steps:
step 1, acquiring historical power data of a wind power plant, and cleaning the data;
step 2, extracting the sectional trend of the wind power data obtained in the step 1 by using an MK-sliding window detection method, and performing climbing detection;
step 3, constructing a time sequence to generate an countermeasure network TimeGAN, carrying out data enhancement on the wind power climbing data detected in the step 2, and according to the step 6:2:2 is divided into a training set, a verification set and a test set;
step 4, an ETSformer model is established, and the training set and the verification set obtained in the step 3 are input into the ETSformer model for training;
step 5, improving the artificial humming algorithm AHA by adopting a Logistic chaotic mapping strategy and a Gaussian-cauchy mixed variation strategy to obtain an IAHA algorithm; optimizing the super parameters of the ETSformer model by using an IAHA algorithm, and improving the prediction accuracy of the ETSformer model in the wind power climbing event;
and 6, predicting the wind power climbing event by using the test set.
Preferably, in step 1, the data cleaning includes missing value processing, outlier processing, data deduplication, data normalization, and normalization.
Preferably, in step 2, the sectional trend extraction is performed on the wind power data obtained in step 1 by using an MK-sliding window detection method, and the climbing detection specifically includes the following steps:
step 21, defining a wind power time sequence x t (t=1, 2, …, phi) and statistics a:
wherein phi is the length of the sequence to be detected, X τ And X ν Year data corresponding to the time series; sgn (X) τ -X ν ) As a sign function, statistics a obey normal distribution;
step 22, calculating the expected E (a) and variance Var (a):
E(Α)=0 (2)
Var(Α)=φ(φ-1)(2φ+5)/18 (3)
step 23, constructing a standardized test statistic n:
wherein, pi obeys normal distribution, when pi >0, there is a rising trend; when pi <0, there is a downward trend;
step 24, detecting wind power trend data by adopting a sliding window with a fixed time width deltat, and defining the characteristics of wind power climbing: amplitude, direction, slope, start time and duration;
step 25, if the power sampling in the sliding window at the current time point t meets the following formula, the climbing event is considered to occur, and the time is the starting time of the climbing event;
wherein P (t) is the wind power at the moment t, and θ is the wind power threshold value for judging the occurrence of climbing;
step 26, if the power sampling in the sliding window at the current time point t meets the following formula, the climbing event is considered to be ended, and the moment is the ending time of the climbing event;
step 27, recording the amplitude, direction, slope, starting time and duration of the wind power climbing event, and returning to step 25 to continuously detect a new climbing event possibly occurring later.
Preferably, in step 3, a time sequence is constructed to generate an countermeasure network TimeGAN, the wind power climbing data detected in step 2 is enhanced, and according to 6:2:2 into training set, verification set and test set, comprising the following steps:
step 31, carrying out sample expansion on wind power climbing data u (t); the joint learning process from the encoder and the antagonism network is as follows:
reconstructing a loss function
Wherein O is u(t) [·]To calculate the euclidean distance of similarity between vectors, u (t) represents the actual time series data,representing time sequence data obtained through self-encoder mapping-inverse mapping learning, and optimizing parameters of the self-encoder;
supervision of loss function
Wherein h (t) is a hidden layer vector mapped in a potential space, g (·) is a cyclic network in a generator, and I (t) is a random vector for learning a time sequence dependency relationship of a sample;
unsupervised loss function
Wherein y (t) andrespectively representing the classification of the real sample and the synthesized sample by the discriminator;
step 32, parameter optimization process of the self-encoder:
optimization process against parameters in the generation network:
wherein θ e ,θ r ,θ g And theta d Respectively representing parameters in the embedding function, the recovery function, the generator and the discriminator, wherein lambda and eta are weight parameters which are all larger than 0;
and 33, generating a large amount of simulated climbing data similar to the characteristics of the wind power climbing data through the countermeasure training of the generator and the discriminator, and expanding a wind power climbing data set.
Preferably, in step 4, an ETSformer model is built, and the training set and the verification set obtained in step 3 are input into the ETSformer model to perform training, which specifically includes the following steps:
step 41, climbing the input wind power into a time sequence Q t-L:t The method comprises the steps of decomposing the seasonal component and the trend component, and further decomposing the trend component into a horizontal component and a growth component; mapping to the hidden space through an input Embedding module;
step 42, extracting an increasing part and a seasonal part of the wind power climbing time sequence through an encoder, and performing nonlinear transformation before moving to the next layer;
wherein t is the current time, L is the length of the historical sequence, and n represents the number of layers; LN is layer normalization, FF is position feed forward network;a residual error that is an nth encoder layer; />For the residual of the previous encoder layer, MH-ESA is the head index smoothing attention mechanism, FA is the frequency attention mechanism, +.>For history growth part, ++>Is a seasonal portion of history;
step 43, defining a horizontal portion at the current time
Where alpha is a smoothing parameter, and,for the growth part of the last moment, +.>Is a seasonal part of the current time;
step 44, obtaining, by the decoder, a prediction of the H-step growth part and a prediction of the seasonal part;
wherein,for the increment of the current time, +.>And->Respectively representing the prediction of the H step growth part and the prediction of the seasonal part; e (E) t:t+H A prediction representing the horizontal part of the H step; n represents the stack number, and GD is growth damping;
step 45, obtaining H-step prediction of wind power climbing event through combination of level, growth and seasonal prediction
Preferably, in step 5, the climbing event is divided into an ascending climbing, a non-climbing and a descending climbing.
Preferably, in step 5, the modification of the artificial humming algorithm AHA by using a Logistic chaotic mapping strategy and a gauss-cauchy hybrid variation strategy to obtain an IAHA algorithm, and optimizing the hyper-parameters of the ETSformer model by using the IAHA algorithm specifically comprises the following steps:
step 51, setting the population size and the iteration times of an IAHA algorithm, and searching the upper limit and the lower limit of a space; initializing an access table;
52, improving the population initialization of the original AHA algorithm, and initializing the population of the AHA algorithm by using a Logistic chaotic mapping strategy;
W σ+1 =εW σ (1-W σ ) (20)
wherein W is σ+1 Positions of sigma+1 populations; epsilon [0,4 ]],
Step 53, defining an objective function as the deviation between the predicted value and the actual value of the wind power climbing event, and calculating the fitness value of the population through the objective function;
step 54, defining three flight skills;
step 55, updating the position of the humming bird; searching an ETSformer model optimal super parameter through different position updating methods, so that the deviation value of the predicted value and the actual value of the wind power climbing event is reduced;
step 56, continuously updating the current solution through algorithm iteration, and synchronously updating the global optimal solution;
and 57, judging whether an algorithm termination condition is reached or not through the given maximum iteration times, and finally outputting the optimal super-parameters of the ETSformer model within the maximum iteration times.
Preferably, in step 54, the three flight skills include omni-directional, diagonal, and axial flight.
Preferably, in step 55, the position updating mode of the foraging behavior guided by the buzzers is adopted;
wherein Γ is i (k+1) is the position of the ith food source at the kth+1th iteration; Γ -shaped structure j,tar (k) Is the location of the target food source that the jth bird is planning to visit; a is a guide foraging factor conforming to standard normal distribution, and D is a direction vector.
Preferably, in step 55, the location of the foraging behavior of the buzzers is updated;
wherein b is a regional foraging factor which obeys standard normal distribution;
a Gauss-Cauchy mixed variation strategy is introduced to replace a migration foraging behavior position updating mode in the original AHA algorithm; the optimal individuals in the population are subjected to variation disturbance, so that the optimal individuals jump out of the local optimal;
wherein Γ is k Representing the position after the kth iteration gaussian-cauchy mixing disturbance,representing the optimal position of the kth iteration of the individual j; gauss (ψ) is a gaussian mutation operator, cauchy (ψ) is a cauchy mutation operator; ρ 1 、ρ 2 Is the weight coefficient of the mutation operator.
Preferably, in step 5, the super parameters include learning rate and attention header number.
The beneficial effects of the invention are as follows: (1) According to the method, the MK-sliding window detection method is used for identifying the wind power climbing event in the wind power, and the sectional trend extraction is carried out on the original wind power, so that the tiny fluctuation of the power is eliminated, the obvious change trend of the data is reserved, and the identification rate of the climbing event in the wind power data is improved; (2) According to the method, the data enhancement is carried out on the wind power climbing data by adopting the TimeGAN, the model training data is expanded, the problem of data missing in a wind power climbing event can be solved, and the reliability and the accuracy of model prediction under extreme weather conditions are improved; (3) The ETSformer model provided by the invention can learn the dependency relationship of level, growth and seasonal part complexity in the time sequence, and replaces the self-attention mechanism in the original converter model by the exponential smoothing attention mechanism and the frequency attention mechanism, thereby improving the accuracy and the efficiency of prediction. Compared with the traditional wind power climbing event prediction model, the prediction precision of the wind power climbing event prediction model can be remarkably improved; (4) The invention provides an IAHA algorithm combining various improved strategies, and a Logistic chaotic mapping strategy population initialization method is adopted, so that a better optimizing effect can be obtained; introducing a Gaussian-Cauchy mixed mutation strategy in the position updating of the original AHA algorithm, and jumping out of local optimum by carrying out mutation disturbance on the optimum individuals in the population; the learning rate and the attention head number of the ETSformer model are optimized by using IAHA, so that overfitting can be prevented, and the generalization capability of the model is improved.
Drawings
FIG. 1 is a schematic flow chart of the method of the present invention.
FIG. 2 is a schematic diagram of a wind power ramp event according to the present invention.
FIG. 3 is a schematic diagram of an ETSformer model according to the present invention.
FIG. 4 is a flowchart of the IAHA algorithm optimization ETSformer model of the present invention.
Detailed Description
As shown in fig. 1, a wind farm climbing event prediction method based on data enhancement includes the following steps:
step 1, acquiring historical power data of a wind power plant, and cleaning the data; the cleaning comprises missing value processing, outlier processing, data deduplication, data normalization and normalization.
Step 2, extracting the sectional trend of the wind power data obtained in the step 1 by using an MK-sliding window detection method, and performing climbing detection; the method specifically comprises the following steps:
step 21, defining a wind power time sequence x t (t=1, 2, …, phi) and statistics a:
wherein phi is the length of the sequence to be detected, X τ And X ν Year data corresponding to the time series; sgn (X) τ -X ν ) As a sign function, statistics a obey normal distribution;
step 22, calculating the expected E (a) and variance Var (a):
E(Α)=0 (2)
Var(Α)=φ(φ-1)(2φ+5)/18 (3)
step 23, constructing a standardized test statistic n:
wherein, pi obeys normal distribution, when pi >0, there is a rising trend; when pi <0, there is a downward trend;
step 24, detecting wind power trend data by adopting a sliding window with a fixed time width deltat, and defining the characteristics of wind power climbing: amplitude, direction, slope, start time and duration;
step 25, if the power sampling in the sliding window at the current time point t meets the following formula, the climbing event is considered to occur, and the time is the starting time of the climbing event;
wherein P (t) is the wind power at the moment t, and θ is the wind power threshold value for judging the occurrence of climbing;
step 26, if the power sampling in the sliding window at the current time point t meets the following formula, the climbing event is considered to be ended, and the moment is the ending time of the climbing event;
step 27, recording the amplitude, direction, slope, starting time and duration of the wind power climbing event, and returning to step 25 to continuously detect a new climbing event possibly occurring later.
Step 3, constructing a time sequence to generate an countermeasure network TimeGAN, carrying out data enhancement on the wind power climbing data detected in the step 2, and according to the step 6:2:2 is divided into a training set, a verification set and a test set; the method specifically comprises the following steps:
step 31, carrying out sample expansion on wind power climbing data u (t); the joint learning process from the encoder and the antagonism network is as follows:
reconstructing a loss function
Wherein O is u(t) [·]To calculate the euclidean distance of similarity between vectors, u (t) represents the actual time series data,representing time sequence data obtained through self-encoder mapping-inverse mapping learning, and optimizing parameters of the self-encoder;
supervision of loss function
Wherein h (t) is a hidden layer vector mapped in a potential space, g (·) is a cyclic network in a generator, and I (t) is a random vector for learning a time sequence dependency relationship of a sample;
unsupervised loss function
Wherein y (t) andrespectively represent the true sample and the synthetic sample of the discriminatorClassifying the books;
step 32, parameter optimization process of the self-encoder:
optimization process against parameters in the generation network:
wherein θ e ,θ r ,θ g And theta d Respectively representing parameters in the embedding function, the recovery function, the generator and the discriminator, wherein lambda and eta are weight parameters which are all larger than 0;
and 33, generating a large amount of simulated climbing data similar to the characteristics of the wind power climbing data through the countermeasure training of the generator and the discriminator, and expanding a wind power climbing data set.
Step 4, an ETSformer model is established, and as shown in fig. 3, the training set and the verification set obtained in the step 3 are input into the ETSformer model for training; the method specifically comprises the following steps:
step 41, climbing the input wind power into a time sequence Q t-L:t The method comprises the steps of decomposing the seasonal component and the trend component, and further decomposing the trend component into a horizontal component and a growth component; mapping to the hidden space through an input Embedding module;
step 42, extracting an increasing part and a seasonal part of the wind power climbing time sequence through an encoder, and performing nonlinear transformation before moving to the next layer;
wherein t is the current time, L is the length of the historical sequence, and n represents the number of layers; LN is layer normalization, FF is position feed forward network;a residual error that is an nth encoder layer; />For the residual of the previous encoder layer, MH-ESA is the head index smoothing attention mechanism, FA is the frequency attention mechanism, +.>For history growth part, ++>Is a seasonal portion of history;
step 43, defining a horizontal portion at the current time
Where alpha is a smoothing parameter, and,for the growth part of the last moment, +.>Is a seasonal part of the current time;
step 44, obtaining, by the decoder, a prediction of the H-step growth part and a prediction of the seasonal part;
wherein,for the increment of the current time, +.>And->Respectively representing the prediction of the H step growth part and the prediction of the seasonal part; e (E) t:t+H A prediction representing the horizontal part of the H step; n represents the stack number, and GD is growth damping;
step 45, obtaining H-step prediction of wind power climbing event through combination of level, growth and seasonal prediction
Step 5, improving the artificial humming algorithm AHA by adopting a Logistic chaotic mapping strategy and a Gaussian-cauchy mixed variation strategy to obtain an IAHA algorithm; optimizing the super parameters of the ETSformer model by using an IAHA algorithm, and improving the accuracy of wind power climbing event prediction, as shown in figure 4; climbing events are classified into uphill, non-uphill, and downhill, as shown in fig. 2; the method specifically comprises the following steps:
step 51, setting the population size and the iteration times of an IAHA algorithm, and searching the upper limit and the lower limit of a space; initializing an access table;
52, improving the population initialization of the original AHA algorithm, and initializing the population of the AHA algorithm by using a Logistic chaotic mapping strategy;
W σ+1 =εW σ (1-W σ ) (20)
wherein W is σ+1 Positions of sigma+1 populations; epsilon [0,4 ]],
Step 53, defining an objective function as the deviation between the predicted value and the actual value of the wind power climbing event, and calculating the fitness value of the population through the objective function;
step 54, defining three flight skills; three flight skills include omni-directional, diagonal and axial flight;
step 55, updating the position of the humming bird; searching an ETSformer model optimal super parameter through different position updating methods, so that the deviation value of the predicted value and the actual value of the wind power climbing event is reduced; a position updating mode of the foraging behavior guided by the buzzers;
wherein Γ is i (k+1) is the position of the ith food source at the kth+1th iteration; Γ -shaped structure j,tar (k) Is the location of the target food source that the jth bird is planning to visit; a is a guide foraging factor conforming to standard normal distribution, and D is a direction vector.
A regional foraging behavior position updating mode of the buzzers;
wherein b is a regional foraging factor which obeys standard normal distribution;
a Gauss-Cauchy mixed variation strategy is introduced to replace a migration foraging behavior position updating mode in the original AHA algorithm; the optimal individuals in the population are subjected to variation disturbance, so that the optimal individuals jump out of the local optimal;
wherein Γ is k Representing the position after the kth iteration gaussian-cauchy mixing disturbance,representing the optimal position of the kth iteration of the individual j; gauss (ψ) is a gaussian mutation operator, cauchy (ψ) is a cauchy mutation operator; ρ 1 、ρ 2 Is the weight coefficient of the mutation operator.
Step 56, continuously updating the current solution through algorithm iteration, and synchronously updating the global optimal solution;
and 57, judging whether an algorithm termination condition is reached or not through the given maximum iteration times, and finally outputting the optimal super-parameters of the ETSformer model within the maximum iteration times, wherein the super-parameters comprise the learning rate and the attention head number.
According to the method, historical power data of the wind power plant are obtained, and data cleaning is performed; extracting sectional trend of the wind power data obtained in the previous step by using an MK-sliding window detection method, and performing climbing detection; constructing a time sequence to generate an countermeasure network TimeGAN, carrying out data enhancement on detected wind power climbing data, and dividing the data into a training set and a verification set test set; establishing an ETSformer model, and inputting the obtained training set and verification set into the ETSformer model for training; improving an artificial humming algorithm AHA by adopting a Logistic chaotic mapping strategy and a Gaussian-cauchy mixed variation strategy to obtain an IAHA algorithm, optimizing the hyper-parameters of an ETSformer model by using the IAHA algorithm, and improving the accuracy of wind power climbing event prediction; and finally, predicting the wind power climbing event by using the test set. The invention can solve the problem of missing climbing data and improve the prediction accuracy.
Claims (10)
1. The wind farm climbing event prediction method based on data enhancement is characterized by comprising the following steps of:
step 1, acquiring historical power data of a wind power plant, and cleaning the data;
step 2, extracting the sectional trend of the wind power data obtained in the step 1 by using an MK-sliding window detection method, and performing climbing detection;
step 3, constructing a time sequence to generate an countermeasure network TimeGAN, carrying out data enhancement on the wind power climbing data detected in the step 2, and according to the step 6:2:2 is divided into a training set, a verification set and a test set;
step 4, an ETSformer model is established, and the training set and the verification set obtained in the step 3 are input into the ETSformer model for training;
step 5, improving the artificial humming algorithm AHA by adopting a Logistic chaotic mapping strategy and a Gaussian-cauchy mixed variation strategy to obtain an IAHA algorithm; optimizing the super parameters of the ETSformer model by using an IAHA algorithm, and improving the prediction accuracy of the ETSformer model in the wind power climbing event;
and 6, predicting the wind power climbing event by using the test set.
2. The method for predicting a wind farm climbing event based on data enhancement according to claim 1, wherein in step 1, the data cleaning includes missing value processing, outlier processing, data deduplication, data normalization and normalization.
3. The method for predicting a climbing event of a wind farm based on data enhancement according to claim 1, wherein in step 2, the wind power data obtained in step 1 is extracted in a sectional trend by using an MK-sliding window detection method, and the climbing detection specifically includes the following steps:
step 21, defining a wind power time sequence x t (t=1, 2, …, phi) and statistics a:
wherein phi is the length of the sequence to be detected, X τ And X ν Year data corresponding to the time series; sgn (X) τ -X ν ) As a sign function, statistics a obey normal distribution;
step 22, calculating the expected E (a) and variance Var (a):
E(Α)=0 (2)
Var(Α)=φ(φ-1)(2φ+5)/18 (3)
step 23, constructing a standardized test statistic n:
wherein, pi obeys normal distribution, when pi >0, there is a rising trend; when pi <0, there is a downward trend;
step 24, detecting wind power trend data by adopting a sliding window with a fixed time width deltat, and defining the characteristics of wind power climbing: amplitude, direction, slope, start time and duration;
step 25, if the power sampling in the sliding window at the current time point t meets the following formula, the climbing event is considered to occur, and the time is the starting time of the climbing event;
wherein P (t) is the wind power at the moment t,a wind power threshold value for judging the occurrence of climbing;
step 26, if the power sampling in the sliding window at the current time point t meets the following formula, the climbing event is considered to be ended, and the moment is the ending time of the climbing event;
step 27, recording the amplitude, direction, slope, starting time and duration of the wind power climbing event, and returning to step 25 to continuously detect a new climbing event possibly occurring later.
4. The method for predicting a wind farm climbing event based on data enhancement according to claim 1, wherein in step 3, a time series generation countermeasure network TimeGAN is constructed, the wind power climbing data detected in step 2 is data enhanced, and according to 6:2: the proportion of 2 is divided into a training set, a verification set and a test set, which specifically comprises the following steps:
step 31, carrying out sample expansion on wind power climbing data u (t); the joint learning process from the encoder and the antagonism network is as follows:
reconstructing a loss function
Wherein O is u(t) [·]To calculate the euclidean distance of similarity between vectors, u (t) represents the actual time series data,representing time sequence data obtained through self-encoder mapping-inverse mapping learning, and optimizing parameters of the self-encoder;
supervision of loss function
Wherein h (t) is a hidden layer vector mapped in a potential space, g (·) is a cyclic network in a generator, and I (t) is a random vector for learning a time sequence dependency relationship of a sample;
unsupervised loss function
Wherein y (t) andrespectively representing the classification of the real sample and the synthesized sample by the discriminator;
step 32, parameter optimization process of the self-encoder:
optimization process against parameters in the generation network:
wherein θ e ,θ r ,θ g And theta d Respectively representing parameters in the embedding function, the recovery function, the generator and the discriminator, wherein lambda and eta are weight parameters which are all larger than 0;
and 33, generating a large amount of simulated climbing data similar to the characteristics of the wind power climbing data through the countermeasure training of the generator and the discriminator, and expanding a wind power climbing data set.
5. The method for predicting the climbing event of the wind farm based on data enhancement according to claim 1, wherein in step 4, an ETSformer model is established, and the training set and the verification set obtained in step 3 are input into the ETSformer model for training, and specifically comprises the following steps:
step 41, climbing the input wind power into a time sequence Q t-L:t The method comprises the steps of decomposing the seasonal component and the trend component, and further decomposing the trend component into a horizontal component and a growth component; mapping to the hidden space through an input Embedding module;
step 42, extracting an increasing part and a seasonal part of the wind power climbing time sequence through an encoder, and performing nonlinear transformation before moving to the next layer;
wherein t is the current time, L is the length of the historical sequence, and n represents the number of layers; LN is layer normalization, FF is position feed forward network;a residual error that is an nth encoder layer; />For the residual of the previous encoder layer, MH-ESA is the head index smoothing attention mechanism, FA is the frequency attention mechanism, +.>For history growth part, ++>Is a seasonal portion of history;
step 43, defining a horizontal portion at the current time
Where alpha is a smoothing parameter, and,for the growth part of the last moment, +.>Is a seasonal part of the current time;
step 44, obtaining, by the decoder, a prediction of the H-step growth part and a prediction of the seasonal part;
wherein,for the increment of the current time, +.>And->Respectively representing the prediction of the H step growth part and the prediction of the seasonal part; e (E) t:t+H A prediction representing the horizontal part of the H step; n represents the stack number, and GD is growth damping;
step 45, obtaining H-step prediction of wind power climbing event through combination of level, growth and seasonal prediction
6. The method for predicting a hill climbing event in a wind farm based on data enhancement according to claim 1, wherein in step 5, the hill climbing event is divided into an upward hill climbing, a non-hill climbing and a downward hill climbing.
7. The method for predicting a climbing event of a wind farm based on data enhancement according to claim 1, wherein in step 5, an artificial humming algorithm AHA is improved by adopting a Logistic chaotic mapping strategy and a gauss-cauchy hybrid variation strategy to obtain an IAHA algorithm, and optimizing super parameters of an ETSformer model by using the IAHA algorithm specifically comprises the following steps:
step 51, setting the population size and the iteration times of an IAHA algorithm and the upper limit and the lower limit of a search space; initializing an access table;
52, improving the population initialization of the original AHA algorithm, and initializing the population of the AHA algorithm by using a Logistic chaotic mapping strategy;
W σ+1 =εW σ (1-W σ ) (20)
wherein W is σ+1 Positions of sigma+1 populations; epsilon [0,4 ]],
Step 53, defining an objective function as the deviation between the predicted value and the actual value of the wind power climbing event, and calculating the fitness value of the population through the objective function;
step 54, defining three flight skills;
step 55, updating the position of the humming bird; searching an ETSformer model optimal super parameter through different position updating methods, so that the deviation value of the predicted value and the actual value of the wind power climbing event is reduced;
step 56, continuously updating the current solution through algorithm iteration, and synchronously updating the global optimal solution;
and 57, judging whether an algorithm termination condition is reached or not through the given maximum iteration times, and finally outputting the optimal super-parameters of the ETSformer model within the maximum iteration times.
8. The data enhanced wind farm climbing event prediction method according to claim 7, wherein in step 54, the three flight skills include omni-directional, diagonal and axial flight.
9. The method for predicting a wind farm climbing event based on data enhancement according to claim 7, wherein in step 55, the foraging behavior is guided by a buzzer in a location update manner;
wherein Γ is i (k+1) is the position of the ith food source at the kth+1th iteration; Γ -shaped structure j,tar (k) Is the location of the target food source that the jth bird is planning to visit; a is a guide foraging factor conforming to standard normal distribution, and D is a direction vector.
10. The method for predicting a wind farm climbing event based on data enhancement according to claim 7, wherein in step 55, the foraging behavior location of the buzzers is updated in a regional manner;
wherein b is a regional foraging factor which obeys standard normal distribution;
a Gauss-Cauchy mixed variation strategy is introduced to replace a migration foraging behavior position updating mode in the original AHA algorithm; the optimal individuals in the population are subjected to variation disturbance, so that the optimal individuals jump out of the local optimal;
wherein Γ is k Representing the position after the kth iteration gaussian-cauchy mixing disturbance,representing the optimal position of the kth iteration of the individual j; gauss (ψ) is a gaussian mutation operator, cauchy (ψ) is a cauchy mutation operator; ρ 1 、ρ 2 Is the weight coefficient of the mutation operator.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311097368.XA CN117117968A (en) | 2023-08-28 | 2023-08-28 | Wind farm climbing event prediction method based on data enhancement |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311097368.XA CN117117968A (en) | 2023-08-28 | 2023-08-28 | Wind farm climbing event prediction method based on data enhancement |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117117968A true CN117117968A (en) | 2023-11-24 |
Family
ID=88808828
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311097368.XA Pending CN117117968A (en) | 2023-08-28 | 2023-08-28 | Wind farm climbing event prediction method based on data enhancement |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117117968A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117354058A (en) * | 2023-12-04 | 2024-01-05 | 武汉安域信息安全技术有限公司 | Industrial control network APT attack detection system and method based on time sequence prediction |
CN117688505A (en) * | 2024-02-04 | 2024-03-12 | 河海大学 | Prediction method and system for vegetation large-range regional negative abnormality |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060132993A1 (en) * | 2004-12-17 | 2006-06-22 | General Electric Company | Wind farm power ramp rate control system and method |
CN116029425A (en) * | 2022-12-16 | 2023-04-28 | 南方电网能源发展研究院有限责任公司 | Method, device and computer equipment for predicting fan output climbing event |
CN116595356A (en) * | 2023-07-17 | 2023-08-15 | 腾讯科技(深圳)有限公司 | Time sequence signal prediction method and device, electronic equipment and storage medium |
CN116646927A (en) * | 2023-05-30 | 2023-08-25 | 淮阴工学院 | Wind power prediction method based on segmented filtering and longitudinal and transverse clustering |
-
2023
- 2023-08-28 CN CN202311097368.XA patent/CN117117968A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060132993A1 (en) * | 2004-12-17 | 2006-06-22 | General Electric Company | Wind farm power ramp rate control system and method |
CN116029425A (en) * | 2022-12-16 | 2023-04-28 | 南方电网能源发展研究院有限责任公司 | Method, device and computer equipment for predicting fan output climbing event |
CN116646927A (en) * | 2023-05-30 | 2023-08-25 | 淮阴工学院 | Wind power prediction method based on segmented filtering and longitudinal and transverse clustering |
CN116595356A (en) * | 2023-07-17 | 2023-08-15 | 腾讯科技(深圳)有限公司 | Time sequence signal prediction method and device, electronic equipment and storage medium |
Non-Patent Citations (1)
Title |
---|
黄麒元 等: "基于ASTFA-ELM的风电功率爬坡预测", 电器与能效管理技术, vol. 2017, no. 3, 15 February 2017 (2017-02-15), pages 54 - 59 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117354058A (en) * | 2023-12-04 | 2024-01-05 | 武汉安域信息安全技术有限公司 | Industrial control network APT attack detection system and method based on time sequence prediction |
CN117688505A (en) * | 2024-02-04 | 2024-03-12 | 河海大学 | Prediction method and system for vegetation large-range regional negative abnormality |
CN117688505B (en) * | 2024-02-04 | 2024-04-19 | 河海大学 | Prediction method and system for vegetation large-range regional negative abnormality |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN117117968A (en) | Wind farm climbing event prediction method based on data enhancement | |
CN106971240A (en) | The short-term load forecasting method that a kind of variables choice is returned with Gaussian process | |
CN111260126B (en) | Short-term photovoltaic power generation prediction method considering correlation degree of weather and meteorological factors | |
CN115293415A (en) | Multi-wind-farm short-term power prediction method considering time evolution and space correlation | |
CN110942194A (en) | Wind power prediction error interval evaluation method based on TCN | |
CN109239669B (en) | Self-evolution radar target detection algorithm based on deep learning | |
CN111832838B (en) | Method for predicting short-term wind power generation output power | |
CN105574615A (en) | Spatial correlation and genetic algorithm (GA) based wind power forecast method for wavelet-BP neural network | |
CN115481791A (en) | Water-wind power generation and power generation combined prediction method, device and equipment | |
CN113988215B (en) | Power distribution network metering cabinet state detection method and system | |
CN116628510A (en) | Self-training iterative artificial intelligent model training method | |
CN114330100A (en) | Short-term photovoltaic power probability interval prediction method | |
CN116842337A (en) | Transformer fault diagnosis method based on LightGBM (gallium nitride based) optimal characteristics and COA-CNN (chip on board) model | |
Bhatt et al. | Statistical and ANN based prediction of wind power with uncertainty | |
Li et al. | GMM-HMM-based medium-and long-term multi-wind farm correlated power output time series generation method | |
CN110766215B (en) | Wind power climbing event prediction method based on feature adaptive selection and WDNN | |
CN115965135A (en) | New energy prediction error modeling method and system based on naive Bayes classification | |
CN115965177A (en) | Improved autoregressive error compensation wind power prediction method based on attention mechanism | |
CN104794546A (en) | Wind power climbing forecasting method based on deep confidence network classifying method | |
CN117477551A (en) | Photovoltaic power generation power prediction method and system based on domain knowledge embedded model | |
CN116070458A (en) | New wind power plant scene generation method based on RAC-GAN | |
CN116646927A (en) | Wind power prediction method based on segmented filtering and longitudinal and transverse clustering | |
CN115296298A (en) | Wind power plant power prediction method | |
CN116488151A (en) | Short-term wind power prediction method based on condition generation countermeasure network | |
CN115758137A (en) | Photovoltaic power prediction method and detection method based on differential evolution algorithm |
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 |