CN114971027A - Power load interval prediction method based on high-performance CIG hybrid model - Google Patents
Power load interval prediction method based on high-performance CIG hybrid model Download PDFInfo
- Publication number
- CN114971027A CN114971027A CN202210597333.1A CN202210597333A CN114971027A CN 114971027 A CN114971027 A CN 114971027A CN 202210597333 A CN202210597333 A CN 202210597333A CN 114971027 A CN114971027 A CN 114971027A
- Authority
- CN
- China
- Prior art keywords
- interval
- signal
- population
- power load
- igwo
- 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 43
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 31
- 241000282461 Canis lupus Species 0.000 claims abstract description 20
- 238000005457 optimization Methods 0.000 claims abstract description 13
- 238000012545 processing Methods 0.000 claims abstract description 12
- 230000006870 function Effects 0.000 claims description 17
- 238000004590 computer program Methods 0.000 claims description 12
- 238000000354 decomposition reaction Methods 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 10
- 238000009826 distribution Methods 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 5
- 238000012549 training Methods 0.000 claims description 5
- 241000717544 Aconitum lycoctonum subsp. vulparia Species 0.000 claims description 3
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 claims description 3
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000005295 random walk Methods 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 238000012952 Resampling Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000007619 statistical method 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
- 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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- 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 provides a power load interval prediction method based on a high-performance CIG (common integrated grid) mixed model, which is used for predicting a short-term power load interval by combining a self-adaptive noise complete set empirical mode decomposition-permutation entropy (CEEMDAN-PE) signal reconstruction algorithm, a Gated Round Unit (GRU) with strong nonlinear fitting capability and the CIG mixed model of an improved Grey wolf algorithm (IGWO) with excellent parameter optimization capability. Firstly, the CEEMDAN-PE-based power load signal is utilized to decompose and reconstruct. And secondly, processing the reconstructed main trend signal by using an interval preset value based on an interval preset criterion. And then, respectively predicting the IGWO-GRU load points of the secondary signal part, and constructing two IGWO-GRU network models for the main trend signal to respectively predict the upper limit and the lower limit of the main trend signal interval. Finally, the point prediction result of the secondary signal is added to the section prediction result of the main tendency signal, thereby obtaining the power load section prediction result.
Description
Technical Field
The invention belongs to the technical field of short-term power load interval prediction, and particularly relates to a power load interval prediction method based on a high-performance CIG (common information group) hybrid model.
Background
Due to the fact that various external influence factors and burst factors exist in actual conditions, the prediction of the power load point has a certain degree of deviation, and the deviation may interfere with the judgment of the power decision, so that the risk decision is caused. In recent years, more and more researchers have studied and studied the field of the power load section prediction in which the sprouting is performed. The good load interval prediction result can provide more load information for power decision-making personnel, so that the power decision-making personnel can capture the approximate range of future power load fluctuation and make a proper power decision. The existing commonly used interval prediction strategies include a Bootstrap method, an upper and lower limit interval prediction method (LUBE), a quantile regression method (QR), a hybrid algorithm based on artificial intelligence, a probability prediction method and the like.
In recent years, various prediction methods have been widely used and developed for short-term power load interval prediction. 1. Bootstrap is a resampling algorithm with a return based on an original sample, a plurality of groups of sample sets are obtained through Bootstrap, the sample sets are respectively predicted, and finally statistical analysis is carried out on the prediction result to obtain a load interval prediction result; 2. the LUBE is a novel, rapid and reliable interval prediction algorithm, and the method does not need any upper and lower bound information of a Prediction Interval (PI) to train a neural network. The method estimates the predicted region boundary by constructing a dual-output Neural Network (NN). The training objective function of the neural network is a PI evaluation index, and the objective function is minimized by optimizing neural network parameters through a parameter optimization algorithm; 3. quantile regression is used for estimating the linear relationship between independent variable quantiles and dependent variable quantiles, and is currently used in the field of interval prediction. However, the linear fitting of the QR model is not accurate enough; 4. for example, a load interval prediction strategy of an improved scale factor method is to perform load point prediction by an LSTM algorithm, and finally apply the improved scale factor algorithm to a load point prediction result, thereby obtaining a load prediction interval. A CEEMDAN-SA-DBN interval prediction model carries out point prediction on each component decomposed by CEEMDAN by using a DBN algorithm as a prediction model, and simultaneously introduces the SA algorithm to carry out optimization on DBN network parameters. And obtaining a load point prediction result after the component prediction results are superposed, and obtaining a load prediction interval after considering a density curve of a historical load prediction error.
Disclosure of Invention
There is further room for improvement in consideration of the prior art, in order to avoid that load point prediction cannot accurately fit future load trends, thereby affecting power decisions. The invention provides a power load interval prediction method based on a high-performance CIG (CEEMDAN-IGWO-GRU) hybrid model, which is used for realizing the power load interval prediction of high interval coverage rate and low interval width, and can enable power grid workers to give more information to the change trend of the load, thereby providing reference for decision making and improving the accuracy of the decision making.
The method is combined with an adaptive noise complete set empirical mode decomposition-permutation entropy (CEEMDAN-PE) signal reconstruction algorithm, a Gated Round Unit (GRU) with strong nonlinear fitting capability and a CIG mixed model of an improved Grey wolf algorithm (IGWO) with excellent parameter optimization capability to predict the short-term power load interval. Firstly, the CEEMDAN-PE-based power load signal is utilized to decompose and reconstruct. And secondly, processing the reconstructed main trend signal by using an interval preset value based on an interval preset criterion. And then, respectively predicting the IGWO-GRU load points of the secondary signal part, and constructing two IGWO-GRU network models for the main trend signal to respectively predict the upper limit and the lower limit of the main trend signal interval. Finally, the point prediction result of the secondary signal and the section prediction result of the main tendency signal are added, thereby obtaining a power load section prediction result.
Firstly, the strategy adopts a CEEMDAN-PE algorithm to carry out characteristic value screening by decomposing a power load signal into a main trend signal and a secondary trend signal and utilizing Pearson correlation coefficient-mutual information entropy (PCCs-MI). Secondly, interval presetting is carried out on the original signal based on the interval presetting criterion provided by the invention. Then, the IGWO-GRU network is used for respectively carrying out interval prediction and point prediction on the main trend signal and the secondary trend signal. And finally, summing the interval prediction and the point prediction results of the primary and secondary signals to obtain a power load interval prediction result.
The invention specifically adopts the following technical scheme:
a power load interval prediction method based on a high-performance CIG hybrid model is characterized by comprising the following steps: firstly, decomposing and reconstructing a power load signal by using a CEEMDAN-PE; then, interval presetting is carried out on the original signal based on an interval presetting criterion; then, an IGWO-GRU network is used for respectively carrying out interval prediction and point prediction on the main trend signal and the secondary signal; finally, the interval prediction and the point prediction results of the primary and secondary signals are added to obtain a power load interval prediction result;
the decomposing and reconstructing the power load signal based on the CEEMDAN-PE comprises: performing CEEMDAN decomposition on an original signal x to be decomposed, performing permutation entropy analysis on IMF and RES components obtained by decomposition, and reconstructing components with similar complexity into one component according to a PE structure; the reconstructed component is denoted as RC n (N ═ 1,2, …, N); wherein, RC N Being the main trend signal, RC n (N-1, 2, …, N-1) is a secondary signal;
the criterion used for performing interval presetting on the original signal based on the interval preset criterion is shown as the following formula and respectively corresponds to a holiday condition and a non-holiday condition:
wherein Interval * up_RC 、Interval * low_RC Upper and lower limits of a preset interval of the main trend signal, nor (-) and renor (-) represent the normalization function and the inverse normalization function, respectively, T act Is the true value of the power load, gamma and eta are E [0,1 ∈ ]]Are all scaling factors, μ e [0,1/2]Is a control coefficient;
the respectively performing interval prediction and point prediction on the main trend signal and the secondary signal by using the IGWO-GRU network specifically comprises the following steps: respectively inputting relevant characteristic data of an upper limit and a lower limit of a preset interval of the main trend signal into two groups of different GRU network models, respectively predicting the upper limit and the lower limit of the main trend signal interval, and setting GRU network parameters according to MAPE values of training results by an IGWO parameter optimization algorithm; then, an IGWO-GRU model is adopted to predict the load points of the remaining secondary signals;
the principle of the IGWO parameter optimization algorithm is as follows:
and (3) increasing the diversity of the initialized population by adopting a opponent learning algorithm: suppose random initialization population P' u×v Wherein u represents the population number of individuals, v represents the dimensionality of the parameter to be optimized, and the opposite population P 'of the initial population P' is calculated according to the following formula;
wherein lb v Represents the lower limit, ub, of the v-th parameter to be optimized v Representing the upper limit of the v parameter to be optimized;
taking the first U population individuals with the optimal fitness as the final initialization population and marking as P from the initial population and the opposite population which are 2U population individuals u×v ；
After the final initialization population is determined, updating the position of each individual in the population; updating the leader wolf by adopting a random walk method; the position updating mode is shown as the following formula:
p e (t+1)＝p e (t)+a rm (t)·cd(t)(e＝α,β,δ)
wherein p is e (t) is the position of the e-th wolf at the time of the t-th iteration, p e (t +1) is the position of the e-th wolf at the time of the t +1 iteration, cd (t) is a random number satisfying the Cauchy distribution, a rm (t) is a control factor;
for the remaining omega wolfs, the update is made according to the following equation:
A uv (t)＝2a(t)·rand 1 -a(t) (13)
C uv (t)＝2·rand 2 (14)
wherein A is uv (t)、C uv (t) is a vector of coefficients,
the v-dimension positions of the wolf at the t-th iteration time alpha, beta and delta respectively, rand 1 And rand 2 Is [0,1 ]]A (t) is a convergence factor;
the random number rand in IGWO is generated from Logistic mapping as shown in the following formula:
wherein the initial random number
Control parameter upsilon epsilon [0,4 ∈]；
Finally, determining the position updating result of each wolfsbane individual by adopting a greedy algorithm; and evaluating the fitness of the updated population every time, thereby selecting a new leading wolf, and then continuously updating the location of the population until the parameter optimization iteration stops.
Further, a function EMD is defined j (. to) is the j-th order IMF mode, δ, obtained by EMD of the signal to be decomposed i (I ═ 1,2, …, I) is the ith white noise satisfying N (0,1) distribution;
the procedure for CEEMDAN decomposition of the original signal x to be decomposed is as follows:
step 1: carrying out noise adding processing x on original signals to be decomposed i ＝x+ε 0 δ i EMD decomposition is carried out on the signals after noise addition to obtain a first-order IMF component of the original signal, and a first-order residual component r is obtained according to the first-order IMF component and the original signal x 1 The process is shown in formulas (1) and (2):
r 1 ＝x-IMF 1 (2)
step 2: according to Step1, after the signal is subjected to noise processing, solving is performed on a k-order IMF component and a residual component, and the process is shown in equations (3) - (5):
r k ＝r k-1 -IMF k (5)
where K is 2,3, …, K being the total number of IMF components;
step 3: repeating Step2 until the residual can not be decomposed further, the final residual component is shown in equation (6):
an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the power load interval prediction method based on the high-performance CIG hybrid model as described above when executing the program.
A computer-readable storage medium, on which a computer program is stored, wherein the program, when executed by a processor, implements a power load interval prediction method based on a high-performance CIG hybrid model as described above.
Compared with the prior art, the invention and the preferred scheme thereof have at least the following outstanding advantages:
1. an algorithm model for performing adaptive interval preset operation on power loads of different date types is designed, and prediction accuracy is obviously improved;
2. an IGWO algorithm is designed to carry out more optimal configuration on the parameters of the prediction model compared with the prior art;
3. the whole scheme can realize the prediction of the power load interval with high interval coverage rate and low interval width, and the performance is superior to that of the prior art.
Drawings
Fig. 1 is a schematic diagram of a load interval prediction logic structure of a CIG hybrid model according to an embodiment of the present invention.
FIG. 2 is a schematic flow chart of an IGWO-GRU hybrid model according to an embodiment of the present invention.
FIG. 3 is a schematic diagram of an optimization process and a result of a CIG hybrid model according to an embodiment of the present invention.
FIG. 4 is a schematic diagram of a prediction result of a CIG hybrid model interval according to an embodiment of the present invention.
Detailed Description
In order to make the features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail as follows:
it should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
As shown in fig. 1 to fig. 4, the method for predicting the power load interval based on the high-performance CIG hybrid model according to this embodiment may be described as the following steps:
step 1: and carrying out CEEMDAN-PE decomposition reconstruction on the load signal so as to reduce the influence of the load signal fluctuation on prediction. Now define function EMD j (. to) is the j-th order IMF mode, δ, obtained by EMD of the signal to be decomposed i (I ═ 1,2, …, I) is the ith white noise satisfying the N (0,1) distribution. The procedure for CEEMDAN decomposition of the original signal x to be decomposed is as follows:
step 1: carrying out noise adding processing x on original signals to be decomposed i ＝x+ε 0 δ i And performing EMD decomposition on the noisy signals respectively to obtain a first-order IMF component of the original signal, and obtaining a first-order residual component according to the first-order IMF component and the original signal x, wherein the process is shown as formulas (1) and (2):
r 1 ＝x-IMF 1 (2)
step 2: as in Step1, after the signal is subjected to noise processing, the solution of the IMF component and residual component of order k is performed, and the process is shown in equations (3) - (5).
r k ＝r k-1 -IMF k (5)
Where K is 2,3, …, K being the total number of IMF components.
Step 3: step2 is repeated until the residual can not be decomposed further, and the final residual component is as shown in equation (6):
and carrying out permutation entropy analysis on the IMF and RES components, and reconstructing the components with similar complexity into one component according to the PE structure. The reconstructed component is denoted as RC n (N-1, 2, …, N). Wherein, RC N Being the main trend signal, RC n (N-1, 2, …, N-1) is a secondary signal
Step 2: and presetting an interval of the main trend signal. Based on the particularity of the holiday load prediction, the embodiment performs interval presetting operation by establishing two sets of criteria, so that holidays and non-holidays are distinguished. This criterion is shown in equations (7) and (8), which correspond to the holiday case and the non-holiday case, respectively:
wherein Interval * up_RC 、Interval * low_RC Upper and lower limits of the preset interval of the main trend signal, nor (-) and renor (-) represent the normalization function and inverse regression, respectivelyA normalizing function, T act Is the true value of the power load, gamma, eta are formed to [0,1 ∈]Are all scaling factors, μ e [0,1/2]Is a control coefficient.
And step 3: and performing short-term power load interval prediction by using the IGWO-GRU hybrid model. And respectively inputting the relevant characteristic data of the upper limit and the lower limit of the preset interval of the main trend signal into two groups of different GRU network models, respectively predicting the upper limit and the lower limit of the main trend signal interval, and setting the GRU network parameters according to the MAPE value of the training result by an IGWO parameter optimization algorithm. Then, the IGWO-GRU model is adopted to predict the load point of the remaining secondary signals. Finally, the section prediction result of the main tendency signal and the point prediction result of the secondary signal are added, thereby obtaining the power load section prediction result. The IGWO-GRU hybrid model flow is shown in FIG. 2.
The GRU network integrates a forgetting gate and an input gate in the LSTM network into an updating gate, so that the GRU network only consists of a resetting gate, an updating gate and an output part, and the cell state and the hidden state in the LSTM network are combined. The simplified improvement of the GRU on the LSTM network not only reduces the parameters required to be trained by the network, shortens the time of model training, but also solves the problem of gradient dissipation. The mathematical expression of the GRU network model is shown in equation (9):
wherein x λ Is an input vector (i.e., a characteristic value of the load signal), h λ-1 For a state memory variable of the preceding moment, h λ For the state memory variable at the present moment, σ (-) is a sigmoid activation function, W r To reset the weighting parameters of the gate parts, r λ To reset the gate state, W u To update the weight parameter of the gate portion, z λ To update the gate state, tanh (-) is a hyperbolic tangent activation function, W o To calculate
The weight parameter of the portion(s),
and memorizing variables for the candidate state at the current moment.
Finally, the output h of the GRU network λ Obtaining the result Y of power load prediction after passing through the full connection layer pred 。
The GRU network parameter setting has a large influence on the prediction accuracy, but the network parameter setting does not have an objective standard, so that the embodiment proposes that the IGWO algorithm optimizes the number of hidden layer nodes and the number of iterations of the GRU.
Considering that the traditional GWO algorithm has the problems that the initial population lacks diversity, the leadership location updating principle has defects and the like, the GWO algorithm is easy to fall into local optimum in the global search process. Therefore, this embodiment proposes an IGWO algorithm, whose method principle is as follows:
and (3) increasing the diversity of the initialized population by adopting a opponent learning algorithm. Suppose random initialization population P' u×v (u represents the population number of individuals, v represents the dimension of the parameter to be optimized), and the opponent population P 'of the initial population P' is calculated according to the formula (10).
Wherein lb v Represents the lower limit, ub, of the v-th parameter to be optimized v Representing the upper limit of the v-th parameter to be optimized.
Taking the first U population individuals with the optimal fitness as a final initialization population and marking as P from the initial population and the opposite population which are 2U population individuals in total u×v 。
And after the final initialization population is determined, updating the position of each individual in the population. And updating the leader wolf by adopting a random walk method. The location update method is shown in equations (11) and (12).
p e (t+1)＝p e (t)+a rm (t)·cd(t) (e＝α,β,δ) (11)
Wherein p is e (t) is the position of the e-th wolf at the time of the t-th iteration, p e (t +1) is the position of the e-th wolf at the time of the t +1 iteration, cd (t) is a random number satisfying the Cauchy distribution, a rm And (t) is a control factor.
For the remaining ω wolfs, updating is performed according to equations (13) - (16).
A uv (t)＝2a(t)·rand 1 -a(t) (13)
C uv (t)＝2·rand 2 (14)
Wherein A is uv (t)、C uv (t) is a vector of coefficients,
the v-dimension positions of the wolf at the t-th iteration time alpha, beta and delta respectively, rand 1 And rand 2 Is [0,1 ]]A (t) is a convergence factor.
The random number rand in IGWO is generated from the Logistic mapping, which is shown as equation (17):
wherein the initial random number
Control parameter upsilon epsilon [0,4 ∈]。
And finally, determining the position updating result of each wolfsbane individual by adopting a greedy algorithm. And evaluating the fitness of the updated population every time, thereby selecting a new leading wolf, and then continuously updating the location of the population until the parameter optimization iteration stops.
The process of optimizing the prediction model parameters by the CIG proposed in this embodiment is shown in fig. 3, and the short-term power load interval prediction result of the CIG hybrid model is shown in fig. 4. It can be seen that it has a significant effect on the prediction performance.
The above program design scheme related to the algorithm provided in this embodiment can be stored in a computer readable storage medium in a coded form, and implemented in a computer program manner, and inputs basic parameter information required for calculation through computer hardware, and outputs a calculation result.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, apparatus, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations of methods, apparatus (devices), and computer program products according to embodiments of the invention. It will be understood that each flow of the flowcharts, and combinations of flows in the flowcharts, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows.
The foregoing is directed to preferred embodiments of the present invention, other and further embodiments of the invention may be devised without departing from the basic scope thereof, and the scope thereof is determined by the claims that follow. However, any simple modification, equivalent change and modification of the above embodiments according to the technical essence of the present invention are within the protection scope of the technical solution of the present invention.
The present invention is not limited to the above-mentioned preferred embodiments, and any other various methods for predicting the power load interval based on the high-performance CIG hybrid model can be obtained according to the teaching of the present invention, and all equivalent changes and modifications made according to the claims of the present invention shall fall within the scope of the present invention.
Claims (4)
1. A power load interval prediction method based on a high-performance CIG hybrid model is characterized by comprising the following steps: firstly, decomposing and reconstructing a power load signal by using a CEEMDAN-PE; then, interval presetting is carried out on the original signal based on an interval presetting criterion; then, an IGWO-GRU network is used for respectively carrying out interval prediction and point prediction on the main trend signal and the secondary signal; finally, the interval prediction and the point prediction results of the primary and secondary signals are added to obtain a power load interval prediction result;
the decomposing and reconstructing the power load signal based on the CEEMDAN-PE comprises: performing CEEMDAN decomposition on an original signal x to be decomposed, and performing permutation entropy division on IMF and RES components obtained by decompositionAnalyzing, and reconstructing components with similar complexity into one component according to the PE structure; the reconstructed component is denoted as RC n (N ═ 1,2, …, N); wherein, RC N Being the main trend signal, RC n (N-1, 2, …, N-1) is a secondary signal;
the criterion for performing interval presetting on the original signal based on the interval presetting criterion is shown as the following formula, and respectively corresponds to a holiday condition and a non-holiday condition:
wherein Interval * up_RC 、Interval * low_RC Upper and lower limits of a preset interval of the main trend signal, nor (-) and renor (-) represent the normalization function and the inverse normalization function, respectively, T act Is the true value of the power load, gamma, eta are formed to [0,1 ∈]Are all scaling factors, μ e [0,1/2]Is a control coefficient;
the respectively performing interval prediction and point prediction on the main trend signal and the secondary signal by using the IGWO-GRU network specifically comprises the following steps: respectively inputting relevant characteristic data of an upper limit and a lower limit of a preset interval of the main trend signal into two groups of different GRU network models, respectively predicting the upper limit and the lower limit of the main trend signal interval, and setting GRU network parameters according to MAPE values of training results by an IGWO parameter optimization algorithm; then, an IGWO-GRU model is adopted to predict the load points of the remaining secondary signals;
the principle of the IGWO parameter optimization algorithm is as follows:
and (3) increasing the diversity of the initialized population by adopting a opponent learning algorithm: suppose random initialization population P' u×v Wherein u represents the population number of individuals, v represents the dimensionality of the parameter to be optimized, and the opposite population P 'of the initial population P' is calculated according to the following formula;
wherein lb v Represents the lower limit, ub, of the v-th parameter to be optimized v Representing the upper limit of the v parameter to be optimized;
taking the first U population individuals with the optimal fitness as the final initialization population and marking as P from the initial population and the opposite population which are 2U population individuals u×v ；
After the final initialization population is determined, updating the position of each individual in the population; updating the leader wolf by adopting a random walk method; the position updating mode is shown as the following formula:
p e (t+1)＝p e (t)+a rm (t)·cd(t)(e＝α,β,δ)
wherein p is e (t) is the position of the e-th wolf at the time of the t-th iteration, p e (t +1) is the position of the e-th wolf at the time of the t +1 iteration, cd (t) is a random number satisfying the Cauchy distribution, a rm (t) is a control factor;
for the remaining omega wolfs, the update is made according to the following equation:
A uv (t)＝2a(t)·rand 1 -a(t) (13)
C uv (t)＝2·rand 2 (14)
wherein A is uv (t)、C uv (t) is a coefficient vector，p α v (t)、p β v (t)、p δ v (t) the v-dimension positions of the wolfs at the t-th iteration time alpha, beta and delta, rand 1 And rand 2 Is [0,1 ]]A (t) is a convergence factor;
the random number rand in IGWO is generated from Logistic mapping as shown in the following formula:
wherein the initial random number
Control parameter upsilon epsilon [0,4 ∈]；
Finally, determining the position updating result of each wolfsbane individual by adopting a greedy algorithm; and evaluating the fitness of the updated population every time, thereby selecting a new leading wolf, and then continuously updating the location of the population until the parameter optimization iteration stops.
2. The power load interval prediction method based on the high-performance CIG hybrid model according to claim 1, characterized in that:
defining a function EMD j (. to) is the j-th order IMF mode, δ, obtained by EMD of the signal to be decomposed i (I ═ 1,2, …, I) is the ith white noise satisfying N (0,1) distribution;
the procedure for CEEMDAN decomposition of the original signal x to be decomposed is as follows:
step 1: carrying out noise adding processing x on original signals to be decomposed i ＝x+ε 0 δ i EMD decomposition is carried out on the signals after the noise is added respectively to obtain a first-order IMF component of the original signals, and a first-order residual error component r is obtained according to the first-order IMF component and the original signals x 1 The process is shown in formulas (1) and (2):
r 1 ＝x-IMF 1 (2)
step 2: according to Step1, after the signal is subjected to noise processing, solving is performed on a k-order IMF component and a residual component, and the process is shown in equations (3) - (5):
r k ＝r k-1 -IMF k (5)
where K is 2,3, …, K being the total number of IMF components;
step 3: repeating Step2 until the residual can not be decomposed further, the final residual component is shown in equation (6):
3. an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the power load interval prediction method based on the high-performance CIG hybrid model as claimed in claim 1 or 2 when executing the program.
4. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements a power load interval prediction method based on a high-performance CIG hybrid model as set forth in claim 1 or 2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210597333.1A CN114971027A (en) | 2022-05-30 | 2022-05-30 | Power load interval prediction method based on high-performance CIG hybrid model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210597333.1A CN114971027A (en) | 2022-05-30 | 2022-05-30 | Power load interval prediction method based on high-performance CIG hybrid model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114971027A true CN114971027A (en) | 2022-08-30 |
Family
ID=82957259
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210597333.1A Pending CN114971027A (en) | 2022-05-30 | 2022-05-30 | Power load interval prediction method based on high-performance CIG hybrid model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114971027A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180284745A1 (en) * | 2016-05-09 | 2018-10-04 | StrongForce IoT Portfolio 2016, LLC | Methods and systems for self-organization of collected data using 3rd party data from a data marketplace in an industrial internet of things environment |
| CN109738519A (en) * | 2019-01-04 | 2019-05-10 | 国网四川省电力公司广安供电公司 | A kind of denoising method of transformer high-voltage bushing lead ultrasound detection |
| CN112882115A (en) * | 2021-04-30 | 2021-06-01 | 湖南师范大学 | Magnetotelluric signal denoising method and system based on GWO optimized wavelet threshold |
| CN114021847A (en) * | 2021-11-19 | 2022-02-08 | 福州大学 | Short-term power load prediction method and system based on ISGU (inverse transformation and aggregation) hybrid model |
-
2022
- 2022-05-30 CN CN202210597333.1A patent/CN114971027A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180284745A1 (en) * | 2016-05-09 | 2018-10-04 | StrongForce IoT Portfolio 2016, LLC | Methods and systems for self-organization of collected data using 3rd party data from a data marketplace in an industrial internet of things environment |
| CN109738519A (en) * | 2019-01-04 | 2019-05-10 | 国网四川省电力公司广安供电公司 | A kind of denoising method of transformer high-voltage bushing lead ultrasound detection |
| CN112882115A (en) * | 2021-04-30 | 2021-06-01 | 湖南师范大学 | Magnetotelluric signal denoising method and system based on GWO optimized wavelet threshold |
| CN114021847A (en) * | 2021-11-19 | 2022-02-08 | 福州大学 | Short-term power load prediction method and system based on ISGU (inverse transformation and aggregation) hybrid model |
Non-Patent Citations (1)
| Title |
|---|
| 陈坚等: "基于SOGWO 的电力系统稳定器参数优化", 电力系统保护与控制, vol. 48, no. 22, 16 November 2020 (2020-11-16), pages 159 - 164 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108900346B (en) | Wireless network flow prediction method based on LSTM network | |
| CN110909926A (en) | TCN-LSTM-based solar photovoltaic power generation prediction method | |
| CN110852515B (en) | Water quality index prediction method based on mixed long-time and short-time memory neural network | |
| CN112488415A (en) | Power load prediction method based on empirical mode decomposition and long-and-short-term memory network | |
| CN112990556A (en) | User power consumption prediction method based on Prophet-LSTM model | |
| CN112364975A (en) | Terminal operation state prediction method and system based on graph neural network | |
| CN111967688A (en) | Power load prediction method based on Kalman filter and convolutional neural network | |
| CN115935834B (en) | Historical fitting method based on deep autoregressive network and continuous learning strategy | |
| CN114219139B (en) | DWT-LSTM power load prediction method based on attention mechanism | |
| CN112381320A (en) | Community personnel flow prediction method, system, storage medium and equipment | |
| CN112116162A (en) | Power transmission line icing thickness prediction method based on CEEMDAN-QFAOA-LSTM | |
| CN111680786B (en) | Time sequence prediction method based on improved weight gating unit | |
| CN114547974A (en) | Dynamic soft measurement modeling method based on input variable selection and LSTM neural network | |
| CN115456245A (en) | Prediction method for dissolved oxygen in tidal river network area | |
| CN113935513A (en) | CEEMDAN-based short-term power load prediction method | |
| CN112434888A (en) | PM2.5 prediction method of bidirectional long and short term memory network based on deep learning | |
| CN111524348A (en) | Long-short term traffic flow prediction model and method | |
| CN114971027A (en) | Power load interval prediction method based on high-performance CIG hybrid model | |
| CN115660167A (en) | Short-term air quality prediction method based on sparrow search algorithm and decomposition error correction | |
| CN115389743A (en) | Method, medium and system for predicting content interval of dissolved gas in transformer oil | |
| CN114720129A (en) | Rolling bearing residual life prediction method and system based on bidirectional GRU | |
| Montiel et al. | The evolutionary learning rule for system identification | |
| CN114239418A (en) | Landslide displacement prediction method based on combination of multiple algorithms | |
| CN113255963A (en) | Road surface use performance prediction method based on road element splitting and deep learning model LSTM | |
| CN113688774B (en) | Advanced learning-based high-rise building wind induced response prediction and training method and device |
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 |