CN112906935A - Method for predicting ultra-short-term power of wind power plant - Google Patents
Method for predicting ultra-short-term power of wind power plant Download PDFInfo
- Publication number
- CN112906935A CN112906935A CN202011522117.8A CN202011522117A CN112906935A CN 112906935 A CN112906935 A CN 112906935A CN 202011522117 A CN202011522117 A CN 202011522117A CN 112906935 A CN112906935 A CN 112906935A
- Authority
- CN
- China
- Prior art keywords
- whale
- algorithm
- short
- long
- memory network
- 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.)
- Withdrawn
Links
- 238000000034 method Methods 0.000 title claims abstract description 29
- 241000283153 Cetacea Species 0.000 claims abstract description 51
- 230000007787 long-term memory Effects 0.000 claims abstract description 15
- 230000006403 short-term memory Effects 0.000 claims abstract description 15
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 11
- 230000006870 function Effects 0.000 claims abstract description 8
- 238000012216 screening Methods 0.000 claims abstract description 6
- 230000009466 transformation Effects 0.000 claims abstract description 4
- 238000005457 optimization Methods 0.000 claims description 7
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 230000003247 decreasing effect Effects 0.000 claims description 4
- 230000015654 memory Effects 0.000 claims description 4
- 210000002569 neuron Anatomy 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 4
- 238000005259 measurement Methods 0.000 claims description 3
- 230000008602 contraction Effects 0.000 claims description 2
- 230000002596 correlated effect Effects 0.000 claims description 2
- 230000000875 corresponding effect Effects 0.000 claims description 2
- 230000007246 mechanism Effects 0.000 claims description 2
- 238000012549 training Methods 0.000 claims description 2
- 238000001514 detection method Methods 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 3
- 238000010248 power generation Methods 0.000 description 3
- 239000013598 vector Substances 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Images
Classifications
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Economics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Strategic Management (AREA)
- Artificial Intelligence (AREA)
- Human Resources & Organizations (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for predicting the ultra-short term power of a wind power plant, and particularly relates to a method for predicting the ultra-short term power of the wind power plant based on a long-term and short-term memory network optimized by an ensemble empirical mode decomposition and improved whale algorithm. Adding white noise to an original detection signal; decomposing the signal added with the white noise by using an EMD (empirical mode decomposition) algorithm to obtain an IMF (intrinsic mode function) component; carrying out Hilbert transformation on the obtained IMF mean value; screening the IMF components after the change by using the sample entropy to reduce the data volume; improving the convergence factor of the original whale algorithm, and balancing the global convergence capability and the local convergence capability of the algorithm; unsupervised generation of appropriate model prediction parameters is performed through combination of an improved whale algorithm and a long-term and short-term memory network; predicting each component data in sequence and accumulating the result to obtain a final predicted value; the method shows better accuracy and generalization performance for wind power data prediction.
Description
Technical Field
The invention relates to a method for predicting ultra-short-term wind power of a wind power plant, in particular to a method for predicting the ultra-short-term power of the wind power plant based on long and short-term memory network prediction optimized by an improved whale algorithm and based on sample entropy measurement screening data after ensemble empirical mode decomposition.
Background
At present, the whole mankind faces the dilemma of energy depletion, so that the development of renewable energy sources is increased to become the best choice for solving the problems. Wind power generation is a main form of wind energy utilization, and is widely distributed in all parts of the world due to the characteristics of environmental protection, sustainability, low cost and remarkable scale benefit. But compared with the traditional water-fire power generation, the strong fluctuation and randomness of the wind speed bring huge challenges to the power system. Today, however, people can predict some laws in the natural world through neural networks.
At present, the modeling method for wind power generation prediction at home and abroad mainly comprises a time sequence method, a Kalman filtering method, an artificial intelligence method and the like. The time series prediction method such as the autoregressive moving average method and the autoregressive differential moving average method can establish a prediction model with higher precision by utilizing sequence self data, but have the defects of lower prediction precision of a low-order model, higher parameter estimation difficulty of a high-order model and the like. Artificial intelligence methods such as artificial neural networks, while dealing with complex problems, are difficult to determine the structure of the network. While the support vector machine, the extreme learning machine and the like have good self-learning and self-adaptive capabilities, stronger nonlinear mapping capabilities and parallel processing capabilities, but rely on parameter setting.
Disclosure of Invention
In order to solve the problem of ultra-short-term power prediction of the wind power plant, the invention provides the ultra-short-term power prediction method of the wind power plant, which has higher precision and stronger robustness.
The technical scheme for solving the technical problems is as follows:
adding white noise to the original signal;
obtaining an IMF component by using an EMD decomposition algorithm;
carrying out Hilbert transformation on the obtained IMF component;
carrying out measurement screening on the obtained IMF components by using sample entropy;
improving a whale algorithm;
unsupervised optimization is carried out on parameters of the long-term and short-term memory network by adopting a whale algorithm;
and predicting the screened data by using the optimized neural network.
Drawings
FIG. 1 is a general flow diagram of the present invention
FIG. 2 is a diagram of the process of adding white noise to the original signal in the present invention
FIG. 3 is a comparison of convergence factors of whale algorithm before and after improvement in the present invention
FIG. 4 is a flow chart of optimizing long-short term memory network by improving whale algorithm in the invention
FIG. 5 is a diagram of the structure of the long and short term memory network of the present invention
Detailed Description
The method comprises the following steps: white noise is added, on the basis of an original signal x (t), the white noise is added into a signal to be decomposed to obtain a signal s (t), and the defect that the original signal lacks a time scale is compensated through optimization of the original signal, so that the signal is smoother, and the phenomenon of modal aliasing is overcome;
step two: decomposing, calculating extreme value distribution of s (t) of signal, constructing envelope curve of maximum value and minimum value by cubic spline interpolation method, and setting the extreme value envelope function of signal as f (t) to obtain formula
In the above formula, u (t) and v (t) are the upper envelope and the lower envelope of the signal, respectively. Setting the mean value of the envelope function f (t) of the extremum as e1Calculating signals s (t) and e1And the difference is set as c1
c1＝s(t)-e1 (2)
By comparison with IMF components, if c1Conform toThe condition of the IMF is then labeled as the first intrinsic mode IMF, labeled c1If not, replacing original c with f (t)1Continuing with equation 2, the first IMF component c is added1R is separated from the signal x (t)1Is composed of
r1＝x(t)-e1 (3)
Step three: separating the obtained r1Taking the obtained IMF as a new decomposition signal x (t), continuing to execute the second step, repeating the steps in the same way, and separating to obtain each sub-component of the IMF until the nth component r is circulatednAnd when the function is a monotone function, the execution is finished. Thereby obtaining a reconstructed signal of
In the above formula 4, rnRepresenting a residual component, ciIndicating that the signal has different frequency components from low to high. The above method of separating IMF from the raw signal is called "screening". However, in practical process, since the envelope mean m1 is hard to be zero, a standard deviation coefficient is introduced as a criterion for judging whether the IMF condition is satisfied. Wherein the standard deviation coefficient SD is expressed as follows
In formula 5: the value of 1 is usually between 0.2 and 0.3, i is the number of decomposition layers, and when the standard deviation coefficient satisfies formula 5, the intrinsic mode component of the IMF obtained by decomposition is considered to meet the requirement;
step four: carrying out Hilbert transformation on the obtained IMF inherent modal components meeting the requirements
The sample entropy screening steps are as follows:
step one, selecting an embedded dimensionReconstructing the transformed signal x to a phase space by a number m to obtain a state vector xi＝(xi,xi+1,...,xi+m+1)
Step two: the distances of all the different state vectors described above are calculated,
step three: setting a similarity margin r, counting with x according to equation 9iThe number of similar state vectors in proportion, where H (-) is a unit step function;
step four: find all
Average value of (d):
step five: resetting the embedding dimension to m +1, and obtaining B according to the formula 7-9m+1。
Step six: obtaining the sample entropy SampEn of the analysis signal x
SampEn(x,m,r)＝ln Bm(r)-ln Bm+1(r) (10)
The whale algorithm was modified as follows:
the convergence factor of the original whale algorithm is as in formula 7
a＝2-2t/Tmax (11)
Wherein T ismaxThe maximum number of iterations is indicated. Through analysis, the convergence factor a at this time is linearly decreased from 2 to 0. However, in the group intelligent algorithm, two search types of global search and local search exist simultaneously, the former has strong capability to ensure the diversity of the population, and the latter and the algorithm are precise for local searchThe accuracy is positively correlated. The original linear decreasing strategy of the convergence factor a cannot fully embody the actual optimization searching process, so the patent provides a new nonlinear convergence mode.
Wherein e is the base of the natural logarithm; t is the number of current iterations; l ismaxIs the maximum number of iterations;
the method for optimizing the long-term and short-term memory network and predicting the ultra-short-term power of the wind power plant by improving the whale algorithm comprises the following steps:
the method comprises the following steps: combining an improved whale algorithm with a long-term and short-term memory network to obtain appropriate model parameters in an unsupervised mode;
step two: the iteration times epoch of the long and short term memory network and the number N of neurons in the hidden layer are used as decision variables of the improved whale algorithm;
step three: and respectively calculating the fitness of the individual whales in the population. Wherein, the fitness function of the whale optimization algorithm is as follows:
fitness (l)＝(Y′-Y)T(Y′-Y) (13)
y' is the output value of the LSTM model, Y is the label value corresponding to the training sample X, and l is the current iteration number.
Step five: comparing the fitness value of each whale, and taking the whale with the lowest fitness as the current optimal whale position X (epoch, N);
step six: and (4) judging the hunting mode of the whale with standing head, and updating the position by selecting a contraction and enclosure mechanism or a spiral bubble net attack mode.
Wherein l and p are random numbers between (-1, 1) and (0, 1), respectively, DpRepresenting the distance between the prey and whale, b being the logarithmic spiral coefficient, the shape of the spiral following bThe value is changed.
When the whale is caught, 1 of the 2 catching modes can be selected according to the probability of p;
step seven: calculating the updated individual fitness value of the whale, and replacing the individual position of the original whale population with the individual position of the new whale population if the individual fitness of the new whale population is superior to that of the previous generation whale population; otherwise, the positions of original whale population individuals are reserved.
Step eight, enabling t to be t +1, judging whether the algorithm reaches a termination condition, if so, outputting an optimal individual position X' and finishing the algorithm of the fitness value; otherwise, repeating the above steps;
step nine: obtaining an optimal individual position X' by utilizing an improved whale algorithm, namely the iteration times epoch of a long-term and short-term memory network and the number N of neurons in a hidden layer;
step ten: and predicting the IMF components after empirical mode decomposition by using a long-short term memory network obtained by unsupervised learning, and finally combining each predicted value to obtain a final result.
Claims (2)
1. A wind power plant ultra-short term power prediction method based on long-short term memory network prediction optimized by an improved whale algorithm is disclosed. The method comprises the following steps:
adding white noise to the original signal;
obtaining an IMF component by using an EMD decomposition algorithm;
carrying out Hilbert transformation on the obtained IMF component;
carrying out measurement screening on the obtained IMF components by using sample entropy;
improving a whale algorithm;
unsupervised optimization is carried out on parameters of the long-term and short-term memory network by adopting a whale algorithm;
and predicting the data by using the optimized neural network.
2. The wind power plant ultra-short term power prediction method based on the collective empirical mode decomposition and whale algorithm optimized long and short term memory network as claimed in claim 1, wherein the step of optimizing the long and short term memory network by using the improved whale algorithm to perform wind power ultra-short term prediction comprises the following steps:
the method comprises the following steps: the convergence factor of the original whale algorithm is 2-2T/TmaxWherein T ismaxThe maximum number of iterations is indicated. Through analysis, the convergence factor a at this time is linearly decreased from 2 to 0. However, in the group intelligent algorithm, two search types, namely global search and local search, exist simultaneously, the former has strong capability to ensure the diversity of the population, and the latter is positively correlated with the accuracy of the algorithm to the local search. The original linear decreasing strategy of the convergence factor a cannot fully reflect the actual optimization searching process, so the patent provides a new nonlinear convergence mode
Wherein e is the base of the natural logarithm; t is the number of current iterations; l ismaxIs the maximum number of iterations;
step two: combining an improved whale algorithm with a long-term and short-term memory network to obtain appropriate model parameters in an unsupervised mode;
step three: the iteration times epoch of the long and short term memory network and the number N of neurons in the hidden layer are used as decision variables of the improved whale algorithm;
step four: and respectively calculating the fitness of the individual whales in the population. Wherein, the fitness function of the whale optimization algorithm is as follows: fitness (l) ═ Y' -Y)T(Y '-Y), wherein Y' is an output value of the LSTM model, Y is a label value corresponding to the training sample X, and l is the current iteration number;
step five: comparing the fitness value of each whale, and taking the whale with the smallest fitness as the current optimal whale position X (epoch, N);
step six: judging the hunting mode of the whale at the head, and updating the position by selecting a contraction and enclosure mechanism or a spiral bubble net attack mode;
wherein l and p are random numbers between (-1, 1) and (0, 1), respectively, DpThe distance between the prey and the whale is represented, b is a logarithmic spiral coefficient, and the shape of the spiral can be changed along with the value of b. When the whale is caught, 1 of the 2 catching modes can be selected according to the probability of p;
step seven: calculating the updated individual fitness value of the whale, and replacing the individual position of the original whale population with the individual position of the new whale population if the individual fitness of the new whale population is superior to that of the previous generation whale population; otherwise, keeping the position of the original whale population;
step eight, enabling t to be t +1, judging whether the algorithm reaches a termination condition, if so, outputting an optimal individual position X' and finishing the algorithm of the fitness value; otherwise, repeating the above steps;
step nine: obtaining the optimal individual position X' by using an improved whale algorithm, namely the iteration times epoch of the long-term and short-term memory network and the number N of hidden layer neurons;
step seven: and predicting the IMF components after empirical mode decomposition by using a long-short term memory network obtained by unsupervised learning, and finally combining each predicted value to obtain a final result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011522117.8A CN112906935A (en) | 2020-12-21 | 2020-12-21 | Method for predicting ultra-short-term power of wind power plant |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011522117.8A CN112906935A (en) | 2020-12-21 | 2020-12-21 | Method for predicting ultra-short-term power of wind power plant |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112906935A true CN112906935A (en) | 2021-06-04 |
Family
ID=76111638
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011522117.8A Withdrawn CN112906935A (en) | 2020-12-21 | 2020-12-21 | Method for predicting ultra-short-term power of wind power plant |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112906935A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113487068A (en) * | 2021-06-21 | 2021-10-08 | 湖北工业大学 | Short-term wind power prediction method based on long-term and short-term memory module |
| CN113537582A (en) * | 2021-07-02 | 2021-10-22 | 东北电力大学 | Photovoltaic power ultra-short-term prediction method based on short-wave radiation correction |
| CN115438833A (en) * | 2022-07-29 | 2022-12-06 | 国网浙江省电力有限公司 | Short-term power load hybrid prediction method |
-
2020
- 2020-12-21 CN CN202011522117.8A patent/CN112906935A/en not_active Withdrawn
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113487068A (en) * | 2021-06-21 | 2021-10-08 | 湖北工业大学 | Short-term wind power prediction method based on long-term and short-term memory module |
| CN113537582A (en) * | 2021-07-02 | 2021-10-22 | 东北电力大学 | Photovoltaic power ultra-short-term prediction method based on short-wave radiation correction |
| CN113537582B (en) * | 2021-07-02 | 2022-05-24 | 东北电力大学 | Photovoltaic power ultra-short-term prediction method based on short-wave radiation correction |
| CN115438833A (en) * | 2022-07-29 | 2022-12-06 | 国网浙江省电力有限公司 | Short-term power load hybrid prediction method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112906935A (en) | Method for predicting ultra-short-term power of wind power plant | |
| CN116596044B (en) | Power generation load prediction model training method and device based on multi-source data | |
| CN117117859B (en) | Photovoltaic power generation power prediction method and system based on neural network | |
| CN113722980B (en) | Ocean wave height prediction method, ocean wave height prediction system, computer equipment, storage medium and terminal | |
| CN112488208B (en) | Method for acquiring remaining life of island pillar insulator | |
| CN114548591A (en) | Time sequence data prediction method and system based on hybrid deep learning model and Stacking | |
| CN113947182B (en) | Traffic flow prediction model construction method based on dual-stage stacked graph convolution network | |
| CN113688869A (en) | Photovoltaic data missing reconstruction method based on generation countermeasure network | |
| Wang et al. | Quantile deep learning model and multi-objective opposition elite marine predator optimization algorithm for wind speed prediction | |
| CN111355633A (en) | Mobile phone internet traffic prediction method in competition venue based on PSO-DELM algorithm | |
| CN115099461A (en) | Solar radiation prediction method and system based on double-branch feature extraction | |
| CN117458480A (en) | Photovoltaic power generation power short-term prediction method and system based on improved LOF | |
| CN116384576A (en) | Wind speed prediction method, device, system and storage medium | |
| CN109886488B (en) | Distributed wind power plant layered hybrid short-term prediction method considering wind speed time lag | |
| CN118296963A (en) | Method for optimizing oil well yield prediction model based on multi-strategy improvement marine predator algorithm | |
| CN113239503B (en) | New energy output scene analysis method and system based on improved k-means clustering algorithm | |
| Chen et al. | Short-term Wind Speed Forecasting Based on Singular Spectrum Analysis, Fuzzy C-Means Clustering, and Improved POABP | |
| CN118040678A (en) | Short-term offshore wind power combination prediction method | |
| CN118038032A (en) | Point cloud semantic segmentation model based on super point embedding and clustering and training method thereof | |
| CN117592593A (en) | Short-term power load prediction method based on improved quadratic modal decomposition and WOA optimization BILSTM-intent | |
| CN116865255A (en) | Short-term wind power prediction method based on improved entropy weight method and SECEEMD | |
| CN116681154A (en) | Photovoltaic power calculation method based on EMD-AO-DELM | |
| CN113807005B (en) | Bearing residual life prediction method based on improved FPA-DBN | |
| CN111882106B (en) | Short-term power load prediction method based on comprehensive factors and CEEMD-IGWO-GRNN | |
| CN112381226B (en) | Deep convolutional neural network architecture searching method based on particle swarm optimization |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20210604 |