CN110210995B - Comprehensive energy load prediction method based on wavelet packet and neural network - Google Patents
Comprehensive energy load prediction method based on wavelet packet and neural network Download PDFInfo
- Publication number
- CN110210995B CN110210995B CN201910481870.8A CN201910481870A CN110210995B CN 110210995 B CN110210995 B CN 110210995B CN 201910481870 A CN201910481870 A CN 201910481870A CN 110210995 B CN110210995 B CN 110210995B
- Authority
- CN
- China
- Prior art keywords
- load
- prediction
- sequences
- data
- load data
- 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.)
- Active
Links
Images
Classifications
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- 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 discloses a comprehensive energy load prediction method based on a wavelet packet and a neural network, which belongs to the field of comprehensive energy system load prediction.
Description
Technical Field
The invention relates to the field of comprehensive energy system load prediction, in particular to a comprehensive energy load prediction method based on a wavelet packet and a neural network.
Background
The comprehensive energy system is a system integrating power, heat energy, cold energy and the like. The method is an important trend of energy development, and has important effects on promoting energy structure optimization, improving energy efficiency and promoting renewable energy consumption. The accurate load prediction of the comprehensive energy system is a basic premise of optimal design, operation, scheduling and energy management, and has important theoretical significance and practical value.
Currently, there are many studies on load prediction by scholars at home and abroad, and common load prediction methods include traditional algorithms based on statistics and machine learning algorithms. The traditional algorithms are further classified into a time series method, a grey prediction method, a regression analysis method and the like, and the change trend of the future load is generally predicted by fitting historical load data. The machine learning algorithm comprises deep learning, a support vector machine, a neural network and the like, has the characteristics of self-organization, self-learning and self-adaptation, can realize the nonlinear mapping between the load influence factor and the load, and has better prediction accuracy compared with the traditional load prediction method, so that the method is more widely applied.
However, the current load prediction model does not highlight the frequency band characteristic of the load, and the influence of high-frequency noise on the load prediction precision is ignored; and most of predictions only aim at the load of a certain energy source, the correlation among energy sources in the comprehensive energy source system is not considered, and the coupling relation among the energy source systems is split.
Disclosure of Invention
The invention aims to provide a comprehensive energy load prediction method based on a wavelet packet and a neural network, which improves the accuracy of load prediction by comprehensively considering the frequency band characteristics of loads and the correlation among various loads.
In order to achieve the above purpose, the invention adopts a comprehensive energy load prediction method based on a wavelet packet and a neural network, which comprises the following steps:
acquiring historical load data;
respectively carrying out wavelet packet decomposition on different types of load data in the historical load data to obtain load sequences of each type of load data under different frequency bands;
dividing the load sequence into a training sample and a prediction sample, and establishing different BP neural networks according to the type and frequency band of the load sequence;
inputting the training samples into a corresponding BP neural network to obtain a load prediction model;
and inputting the prediction sample into a corresponding load prediction model to obtain a load prediction result.
Further, after the obtaining of the historical load data, the method further includes:
filling missing data of the historical load data to obtain preprocessed historical load data;
correspondingly, the performing wavelet packet decomposition on different types of load data in the historical load data to obtain load sequences of each type of load data under different frequency bands includes:
and respectively carrying out wavelet packet decomposition on different types of load data in the preprocessed historical load data to obtain load sequences of each type of load data under different frequency bands.
Further, the performing wavelet packet decomposition on different types of load data in the historical load data to obtain load sequences of each type of load data in different frequency bands includes:
decomposing different types of load data in the historical load data by adopting wavelet basis functions respectively to obtain a wavelet packet decomposition coefficient of each type of load data;
and performing single-branch reconstruction on the wavelet packet decomposition coefficient to obtain a load sequence of each load data under different frequency bands.
Further, after the performing wavelet packet decomposition on different types of load data in the historical load data respectively to obtain load sequences of each type of load data in different frequency bands, the method further includes:
calculating the correlation coefficient between every two different types of load sequences under the same frequency band;
screening out load sequences with the absolute values of the relation numbers larger than a set threshold value under a certain frequency band;
correspondingly, the dividing the load sequence into training samples and prediction samples, and establishing different BP neural networks according to the type and frequency band of the load sequence includes:
dividing the screened load sequences into training samples and prediction samples, and establishing different BP neural networks according to the types and frequency bands of the screened load sequences.
Further, the dividing the screened load sequence into a training sample and a prediction sample, and establishing different BP neural networks according to the type and frequency band of the screened load sequence includes:
normalizing the screened load sequence;
dividing the load sequence after normalization into a training sample and a prediction sample;
and establishing different BP neural networks according to the types and frequency bands of the screened load sequences.
Further, the inputting the training samples into the corresponding BP neural network to obtain a load prediction model includes:
inputting the normalized load sequence obtained before the moment k in the training sample into a corresponding BP neural network to obtain a prediction result of the load sequence at the moment k;
and (4) carrying out iterative training on the BP neural network until the prediction error reaches a set threshold value, and stopping training to obtain a load prediction model.
Further, the inputting the prediction sample into a corresponding load prediction model to obtain a load prediction result includes:
inputting the prediction samples into corresponding load prediction models to obtain a prediction result of each load sequence;
and adding the prediction results of the same load sequence under different frequency bands to obtain a load prediction result.
Further, after the load prediction results are obtained by adding the prediction results of the load sequences of the same kind in different frequency bands, the method further includes:
and performing inverse normalization processing on the load prediction result to obtain a normalized load prediction result.
Compared with the existing load prediction method, the method has the following beneficial effects:
in the invention, considering that the frequency band characteristics of the load can influence the load prediction, the load data is decomposed into load sequences under a plurality of frequency bands by performing multi-layer wavelet packet transformation on historical load data, then load prediction models are respectively established, and the prediction results of the load sequences under different frequency bands are added to obtain a load prediction result, thereby improving the accuracy of the load prediction; in addition, a proper load sequence is screened out by calculating the correlation coefficients of different loads on each frequency band, so that the coupling between each energy source in the comprehensive energy source system is improved, and the accuracy of load prediction is further improved.
Drawings
The following detailed description of embodiments of the invention refers to the accompanying drawings in which:
FIG. 1 is a schematic flow diagram of the present invention;
FIG. 2 is a schematic diagram of a load data wavelet packet decomposition;
FIG. 3 is a graph of a predicted value versus an actual value of an electrical load;
FIG. 4 is a graph of a predicted value versus an actual value of heat load fitted to a load;
fig. 5 is a graph of a predicted value versus an actual value of the cooling load.
Detailed Description
To further illustrate the features of the present invention, refer to the following detailed description of the invention and the accompanying drawings. The drawings are for reference and illustration purposes only and are not intended to limit the scope of the present disclosure.
As shown in fig. 1, the invention adopts a comprehensive energy load prediction method based on a wavelet packet and a neural network, which comprises the following steps:
s1, acquiring historical load data;
s2, performing wavelet packet decomposition on different types of load data in the historical load data respectively to obtain load sequences of each type of load data under different frequency bands;
s3, dividing the load sequence into a training sample and a prediction sample, and establishing different BP neural networks according to the type and frequency band of the load sequence;
s4, inputting the training samples into the corresponding BP neural network to obtain a load prediction model;
and S5, inputting the prediction sample into a corresponding load prediction model to obtain a load prediction result.
Further, after S1, the method further includes:
filling missing data of the historical load data to obtain preprocessed historical load data;
accordingly, S2 includes:
and respectively carrying out wavelet packet decomposition on different types of load data in the preprocessed historical load data to obtain load sequences of each type of load data under different frequency bands.
Specifically, the present embodiment specifically describes using the electrical load data, the thermal load data, and the cooling load data in the historical load data:
first, electric load data l (T), heat load data h (T), and cold load data f (T) are acquired, where T is 1,2, …, and T is a history time. The embodiment collects the electricity, heat and cold load data of a college comprehensive energy system from 1 month and 1 day in 2018 to 12 months and 31 days in 2018, the sampling frequency is 1h, 24 times of data are collected every day, and 8760 points of data exist in each load.
In order to improve the accuracy of prediction, the initial historical load data needs to be preprocessed, missing data is filled, and the preprocessed historical load data is obtained, wherein the filling formula is as follows:
wherein x (t) represents the electrical, thermal and cold load data at time t after padding; beta is a repairing factor, beta belongs to (0.8,1.2),the value is generally 1; recording the preprocessed electric, heat and cold load data as
Further, S2 includes the following subdivision steps:
s21, decomposing different types of load data in the historical load data by adopting wavelet basis functions respectively to obtain wavelet packet decomposition coefficients of each type of load data;
and S22, performing single branch reconstruction on the wavelet packet decomposition coefficient to obtain a load sequence of each load data under different frequency bands.
Specifically, the present embodiment uses db3 wavelet basis function to process the preprocessed electrical, thermal and cold load dataMultiple wavelet packet decompositions, e.g. N layers, are performed separately to obtain 2 for each loadNGrouping wavelet packet decomposition coefficients; then, single reconstruction is carried out on the wavelet packet decomposition coefficient to obtain the electricity, heat and cold load data of 2NThe load sequences in each frequency band are respectively recorded asWherein i represents a frequency band, and i ═ 1,2N。
For illustration, in this embodiment, 3-layer wavelet packet decomposition is performed on the preprocessed electrical, thermal, and cold load data, and as shown in fig. 2, 8 frequency bands, that is, 8 load sequences, are obtained by decomposition and reconstruction.
Further, after S2, the method further includes:
calculating the correlation coefficient between every two different types of load sequences under the same frequency band;
screening out load sequences with the absolute values of the relation numbers larger than a set threshold value under a certain frequency band;
accordingly, S3 includes:
dividing the screened load sequences into training samples and prediction samples, and establishing different BP neural networks according to the types and frequency bands of the screened load sequences.
In particular, since there is a correlation between different loads, considering the correlation between different loads in predicting the load result may improve the accuracy of the prediction.
After 8 load sequences of each load data are obtained, calculating the correlation coefficient between every two load sequences, wherein the calculation formula is as follows:
wherein ρlh,i、ρlf,i、ρhf,iRespectively are correlation coefficients between an electric load sequence and a heat load sequence, an electric load sequence and a cold load sequence, and a heat load sequence and a cold load sequence; respectively electric load sequenceThermal load sequenceCold load sequenceAverage value of (d); the value of the correlation coefficient is between-1 and 1, and the larger the absolute value of the correlation coefficient is, the stronger the correlation is.
Comparing the correlation coefficients among the three load sequences with a set threshold delta respectively, and screening out min (| rho)lh,i|,|ρlf,i|,|ρhf,iI) is greater than the load sequence of delta; assuming deletion of m payload sequences, 2 remains after screeningN-m load sequences.
Table 1 shows the correlation coefficients between the electrical, thermal, and cold load sequences in 8 frequency bands, where the threshold δ is 0.05, and the load sequences with the absolute values of the correlation coefficients greater than the threshold are selected, that is, the load sequences in the frequency bands 1,2, 3, and 4 are used as the load sequences after screening.
Table 1: correlation coefficient corresponding to each load sequence under different frequency bands
Load sequence | Electric-cold correlation coefficient | Coefficient of electric-thermal correlation | Coefficient of cold- |
1 | 0.9490 | -0.7842 | -0.8236 |
2 | 0.3315 | -0.3613 | -0.4292 |
3 | 0.6850 | -0.1666 | -0.1563 |
4 | 0.4857 | 0.0543 | -0.1235 |
5 | 0.2259 | -0.0568 | 0.0212 |
6 | 0.2951 | -0.0489 | 0.0857 |
7 | 0.3029 | 0.0071 | 0.0110 |
8 | 0.3471 | -0.1883 | -0.0166 |
Further, dividing the screened load sequence into a training sample and a prediction sample, establishing different BP neural networks according to the type and frequency band of the screened load sequence, and subdividing the BP neural networks into the following steps:
normalizing the screened load sequence;
dividing the load sequence after normalization into a training sample and a prediction sample;
and establishing different BP neural networks according to the types and frequency bands of the screened load sequences.
Specifically, the screened load sequences are normalized, and the formula is as follows:
wherein, the first and the second end of the pipe are connected with each other,respectively are electric, heat and cold load sequences at the time t after normalization processing in the frequency band i;are respectively asMinimum value of (d);are respectively asIs measured.
Normalizing the load sequence obtained by the time S t 1, 2.. and S are used as training samples, and the load sequence after normalization processing acquired after the time S is used as a training sampleT is S +1, S +2, T as a prediction sample.
In this example, the selection of post-screening 2N8592 points of data in the first 358 days of the sequence of the electricity, heat and cold loads in the m frequency bands are used as training samples, and 168 points of data in the last 7 days are used as prediction samples.
According to 2 after screeningN-establishing BP neural network with different structures by electric, heat and cold load sequences in m frequency bands, wherein the number of input layer neurons is al,l=1,2,...,2NM, number of neurons in the hidden layer bl,l=1,2,...,2NM, number of neurons in the output layer cl,l=1,2,...,2NM, training algebra dl,l=1,2,...,2NM, and the neural network structure parameters of the load sequences under each frequency band are shown in the table 2.
Table 2: neural network structure parameter corresponding to different load sequences
Load sequence | Implicit layer number | Input-implicit-output layer neuron | Training algebra | |
1 | 1 | 20-12-1 | 20000 | |
2 | 1 | 16-12-1 | 15000 | |
3 | 1 | 12-12-1 | 10000 | |
4 | 1 | 8-12-1 | 8000 |
Further, S4 is subdivided into the following steps:
inputting the normalized load sequence obtained before the moment k in the training sample into a corresponding BP neural network to obtain a prediction result of the load sequence at the moment k;
and (4) carrying out iterative training on the BP neural network until the prediction error reaches a set threshold value, and stopping training to obtain a load prediction model.
Specifically, g normalized load sequences before the time k to be predicted And (d), inputting the BP neural networks of respective frequency bands and types, outputting load data at a time k to be predicted, and stopping iterative training when the prediction error of the output load data reaches a set threshold value beta, so as to obtain different BP neural network load prediction models.
In this embodiment, the load sequence after normalization processing 8 hours before the time k to be predicted is used as a training sample input to the BP neural network, the training sample is 8584 × 8 dimensional data, the output data is load data of the time k to be predicted, which is 8584 × 1 dimensional data, and the threshold β is set to 0.0001. And (3) realizing nonlinear mapping of input and output through iterative training, and storing a network structure to further obtain a corresponding load prediction model.
Further, S5 is subdivided into the following steps:
inputting the prediction samples into corresponding load prediction models to obtain a prediction result of each load sequence;
and adding the prediction results of the same load sequence under different frequency bands to obtain a load prediction result.
Further, after S5, the method further includes:
and performing inverse normalization processing on the load prediction result to obtain a normalized load prediction result.
Specifically, the processed load sequence is normalizedT, inputting the load prediction models of the frequency bands and the types where the load prediction models are located, and obtaining the prediction result of each load; respectively setting the electric, heat and cold load data at 2NAdding the predicted results of the load sequences under the m frequency bands to obtain a load predicted result; then, the load prediction results are subjected to inverse normalization processing to obtain final load prediction results of the electricity, heat and cold load data, which are respectively recorded as Lp(t),Hp(t),Fp(t),t=S+1,S+2,...,T。
As shown in FIGS. 3-5, experiments show that the curve of the prediction result of the electricity, heat and cold load data obtained by the method provided by the invention has high fitting degree with the curve of the actual electricity, heat and cold load data, and the effectiveness and feasibility of the method are verified.
Further, after performing inverse normalization processing on the load prediction result to obtain a normalized load prediction result, the method further includes:
and calculating the relative error and the average error of the load prediction result.
In order to further verify that the result of the load data predicted by the method provided by the invention has higher accuracy, the relative error and the average error of the load prediction result are respectively calculated, and the calculation formula is as follows:
wherein R isl(t)、Rh(t)、Rf(t) relative error of prediction of electrical, thermal and cold load data, Ml、Mh、MfThe average accuracy of the electrical, thermal and cold load data predictions, respectively.
Through calculation, the maximum relative errors of the electricity load data, the heat load data and the cold load data predicted by the method are respectively 4.82%, 7.10% and 6.99%, and the average accuracies of the electricity load data, the heat load data and the cold load data are respectively 97.57%, 97.42% and 96.87%. The prediction model provided by the present invention is compared with commonly used machine learning algorithm models (such as neural network models, support vector regression models), and the same prediction sample data is used, with the results as shown in table 3:
table 3: prediction accuracy comparison with machine learning
Then, the method provided by the invention is compared with a wavelet packet neural network method without considering the correlation in terms of time and precision, and the result is shown in the table 4:
table 4: comparison of various indexes of different prediction methods
The experimental result shows that the prediction method provided by the invention can effectively improve the accuracy of load prediction while shortening the training time.
The invention decomposes the load data into load sequences of different frequency bands by carrying out multi-layer wavelet packet transformation on the historical load data, screens out proper load sequences by calculating the correlation coefficient between different types of load sequences on each frequency band, and trains a neural network load prediction model by using the screened load sequences, thereby improving the coupling between energy sources in the comprehensive energy system and improving the accuracy of load prediction while considering the frequency characteristic of the load.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (7)
1. A comprehensive energy load prediction method based on a wavelet packet and a neural network is characterized by comprising the following steps:
acquiring historical load data;
respectively carrying out wavelet packet decomposition on different types of load data in the historical load data to obtain load sequences of each type of load data under different frequency bands;
calculating the correlation coefficient between every two different types of load sequences under the same frequency band;
screening out load sequences with relation numbers absolute values larger than a set threshold value under a certain frequency band;
dividing the screened load sequences into training samples and prediction samples, and establishing different BP neural networks according to the types and frequency bands of the screened load sequences;
inputting the training samples into a corresponding BP neural network to obtain a load prediction model;
and inputting the prediction sample into a corresponding load prediction model to obtain a load prediction result.
2. The method of claim 1, after said obtaining historical load data, further comprising:
filling missing data of the historical load data to obtain preprocessed historical load data;
correspondingly, the performing wavelet packet decomposition on different types of load data in the historical load data respectively to obtain load sequences of each type of load data under different frequency bands includes:
and respectively carrying out wavelet packet decomposition on different types of load data in the preprocessed historical load data to obtain load sequences of each type of load data under different frequency bands.
3. The method of claim 1, wherein the performing wavelet packet decomposition on different types of load data in the historical load data to obtain a load sequence of each type of load data in different frequency bands comprises:
decomposing different types of load data in the historical load data by adopting wavelet basis functions respectively to obtain a wavelet packet decomposition coefficient of each type of load data;
and performing single-branch reconstruction on the wavelet packet decomposition coefficient to obtain a load sequence of each load data under different frequency bands.
4. The method of claim 1, wherein the dividing the screened load sequences into training samples and prediction samples and establishing different BP neural networks according to the types and frequency bands of the screened load sequences comprises:
carrying out normalization processing on the screened load sequences;
dividing the load sequence after normalization into a training sample and a prediction sample;
and establishing different BP neural networks according to the types and frequency bands of the screened load sequences.
5. The method of claim 4, wherein inputting the training samples into a corresponding BP neural network to obtain a load prediction model comprises:
inputting the normalized load sequence obtained before the moment k in the training sample into a corresponding BP neural network to obtain a prediction result of the load sequence at the moment k;
and (4) carrying out iterative training on the BP neural network until the prediction error reaches a set threshold value, and stopping training to obtain a load prediction model.
6. The method of claim 1, wherein inputting the prediction samples into corresponding load prediction models to obtain load prediction results comprises:
inputting the prediction samples into corresponding load prediction models to obtain a prediction result of each load sequence;
and adding the prediction results of the same load sequence under different frequency bands to obtain a load prediction result.
7. The method of claim 6, wherein after adding the prediction results of the load sequences of the same kind in different frequency bands to obtain the load prediction results, the method further comprises:
and performing inverse normalization processing on the load prediction result to obtain a normalized load prediction result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910481870.8A CN110210995B (en) | 2019-06-04 | 2019-06-04 | Comprehensive energy load prediction method based on wavelet packet and neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910481870.8A CN110210995B (en) | 2019-06-04 | 2019-06-04 | Comprehensive energy load prediction method based on wavelet packet and neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110210995A CN110210995A (en) | 2019-09-06 |
CN110210995B true CN110210995B (en) | 2022-05-31 |
Family
ID=67790679
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910481870.8A Active CN110210995B (en) | 2019-06-04 | 2019-06-04 | Comprehensive energy load prediction method based on wavelet packet and neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110210995B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110909936B (en) * | 2019-11-21 | 2021-12-17 | 广东电网有限责任公司广州供电局 | Predicted load acquisition method and device, computer equipment and storage medium |
CN113052400A (en) * | 2021-04-23 | 2021-06-29 | 国网吉林省电力有限公司 | Short-term thermal load probability prediction method considering electrical-thermal coupling characteristics |
CN113516291A (en) * | 2021-05-24 | 2021-10-19 | 国网河北省电力有限公司经济技术研究院 | Power load prediction method, device and equipment |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104408529A (en) * | 2014-11-21 | 2015-03-11 | 广东工业大学 | Short-term load predicting method of power grid |
CN106446494A (en) * | 2016-05-11 | 2017-02-22 | 新疆大学 | Wavelet packet-neural network-based wind/photovoltaic power prediction method |
CN107392368A (en) * | 2017-07-17 | 2017-11-24 | 天津大学 | Meteorological forecast-based office building dynamic heat load combined prediction method |
WO2017212880A1 (en) * | 2016-06-09 | 2017-12-14 | 株式会社日立製作所 | Data prediction system and data prediction method |
CN109472404A (en) * | 2018-10-31 | 2019-03-15 | 山东大学 | A kind of Short-Term Load Forecasting of Electric Power System, model, apparatus and system |
-
2019
- 2019-06-04 CN CN201910481870.8A patent/CN110210995B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104408529A (en) * | 2014-11-21 | 2015-03-11 | 广东工业大学 | Short-term load predicting method of power grid |
CN106446494A (en) * | 2016-05-11 | 2017-02-22 | 新疆大学 | Wavelet packet-neural network-based wind/photovoltaic power prediction method |
WO2017212880A1 (en) * | 2016-06-09 | 2017-12-14 | 株式会社日立製作所 | Data prediction system and data prediction method |
CN107392368A (en) * | 2017-07-17 | 2017-11-24 | 天津大学 | Meteorological forecast-based office building dynamic heat load combined prediction method |
CN109472404A (en) * | 2018-10-31 | 2019-03-15 | 山东大学 | A kind of Short-Term Load Forecasting of Electric Power System, model, apparatus and system |
Non-Patent Citations (3)
Title |
---|
Heating,cooling,and electrical load forecasting for a large-scale district energy system;POWELL K M etal.;《Energy》;20141231;第877-885页 * |
基于Copula理论的综合能源系统负荷相关性研究;马建鹏 等;《青岛大学学报(工程技术版)》;20181130;第33卷(第4期);第56-60页 * |
基于小波包变换的信号去噪方法研究;范显峰 等;《哈尔滨工业大学学报》;20030731;第35卷(第7期);第809-811页 * |
Also Published As
Publication number | Publication date |
---|---|
CN110210995A (en) | 2019-09-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110210995B (en) | Comprehensive energy load prediction method based on wavelet packet and neural network | |
CN108416695B (en) | Power load probability density prediction method, system and medium based on deep learning | |
CN107315884B (en) | Building energy consumption modeling method based on linear regression | |
CN111337768A (en) | Deep parallel fault diagnosis method and system for dissolved gas in transformer oil | |
CN114091766B (en) | CEEMDAN-LSTM-based space load prediction method | |
CN110070228B (en) | BP neural network wind speed prediction method for neuron branch evolution | |
CN112861992B (en) | Wind power plant ultra-short term power prediction method based on independent sparse stacking self-encoder | |
CN112668611B (en) | Kmeans and CEEMD-PE-LSTM-based short-term photovoltaic power generation power prediction method | |
CN111105321A (en) | Power load prediction method based on EEMD secondary decomposition | |
CN108537359B (en) | Photovoltaic power generation power prediction based on wavelet analysis and BP neural network | |
CN104050547A (en) | Non-linear optimization decision-making method of planning schemes for oilfield development | |
CN114358185A (en) | Improved K-means clustering CCA-BilSTM-based multi-dimensional short-term power load prediction method | |
CN111783947A (en) | Energy consumption prediction method based on LSTM neural network | |
CN113673768A (en) | Wind power probability prediction model based on VMD and CNN-LSTM | |
CN112508279A (en) | Regional distributed photovoltaic prediction method and system based on spatial correlation | |
CN116151464A (en) | Photovoltaic power generation power prediction method, system and storable medium | |
CN116245030A (en) | Deep learning water demand prediction method with automatic parameter feedback adjustment | |
CN112381667B (en) | Distribution network electrical topology identification method based on deep learning | |
CN113434493A (en) | Non-invasive load decomposition method based on Transformer | |
Shang et al. | A model compression based framework for electrical equipment intelligent inspection on edge computing environment | |
CN111539558A (en) | Power load prediction method adopting optimized extreme learning machine | |
CN117034805A (en) | WSBLMA-based fuel cell centrifugal air compressor life prediction method | |
Chen et al. | Medium-term load forecast based on sequence decomposition and neural network | |
CN115456287A (en) | Long-and-short-term memory network-based multi-element load prediction method for comprehensive energy system | |
Mao et al. | Communication-efficient federated learning for power load forecasting in electric iots |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |