CN110222727A - A kind of short-term load forecasting method and device based on deep neural network - Google Patents
A kind of short-term load forecasting method and device based on deep neural network Download PDFInfo
- Publication number
- CN110222727A CN110222727A CN201910404028.4A CN201910404028A CN110222727A CN 110222727 A CN110222727 A CN 110222727A CN 201910404028 A CN201910404028 A CN 201910404028A CN 110222727 A CN110222727 A CN 110222727A
- Authority
- CN
- China
- Prior art keywords
- sample
- neural network
- deep neural
- data
- short
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 34
- 238000013277 forecasting method Methods 0.000 title claims abstract description 18
- 238000012549 training Methods 0.000 claims abstract description 55
- 238000005259 measurement Methods 0.000 claims abstract description 31
- 238000003062 neural network model Methods 0.000 claims abstract description 9
- 239000013598 vector Substances 0.000 claims description 52
- 238000005070 sampling Methods 0.000 claims description 24
- 238000010606 normalization Methods 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000012216 screening Methods 0.000 abstract 1
- 238000000034 method Methods 0.000 description 15
- 238000004422 calculation algorithm Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 240000002853 Nelumbo nucifera Species 0.000 description 2
- 235000006508 Nelumbo nucifera Nutrition 0.000 description 2
- 235000006510 Nelumbo pentapetala Nutrition 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000003313 weakening effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
- 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Artificial Intelligence (AREA)
- Human Resources & Organizations (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Tourism & Hospitality (AREA)
- Development Economics (AREA)
- Biomedical Technology (AREA)
- Power Engineering (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Public Health (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of short-term load forecasting methods based on deep neural network, comprising: obtains historical data and characteristic to be predicted;Calculate the similarity measurement of the historical data and the characteristic;Several data samples are picked out from the historical data according to the similarity measurement, and assign training weight to each data sample, obtain sample training collection;Deep neural network model is trained using the sample training collection, obtains prediction model;Prediction result is calculated by the prediction model in the characteristic.The present invention enables the prediction model based on deep neural network targetedly to learn to training sample, to promote the accuracy and precision of short-term load forecasting by screening and assigning sample different training weights to sample.
Description
Technical field
The present invention relates to Techniques for Prediction of Electric Loads fields, more particularly, to a kind of short-term negative based on deep neural network
Lotus prediction technique and device.
Background technique
Short-term load forecasting is the basis of power system security economical operation, continuous complete with China's electricity market system
Kind, the effect that short-term load forecasting plays is increasingly important.Currently, experts and scholars carry out in terms of short-term load forecasting both at home and abroad
Numerous studies, and a series of intelligent Forecasting methods are proposed based on machine learning, wherein deep neural network (Deep
Neural network, DNN) it is one of most common machine learning algorithm.But currently based on the short-term load forecasting of DNN
Method causes prediction result inaccurate since the training of machine learning algorithm does not reach requirement also.
Summary of the invention
In view of the above technical problems, the present invention provides a kind of short-term load forecasting method based on deep neural network and
Device, by being screened and being assigned sample different training weights to sample, so that the prediction mould based on deep neural network
Type can targetedly learn training sample, to promote the accuracy and precision of short-term load forecasting.The skill
Art scheme is as follows:
In a first aspect, the embodiment of the invention provides a kind of short-term load forecasting method based on deep neural network, packet
It includes:
Obtain historical data and characteristic to be predicted;
Calculate the similarity measurement of the historical data and the characteristic;
Several data samples are picked out from the historical data according to the similarity measurement and to each data sample
Training weight is assigned, sample training collection is obtained;
Deep neural network model is trained using the sample training collection, obtains prediction model;
Prediction result is calculated by the prediction model in the characteristic.
In a first possible implementation of the first aspect of the invention, the calculating historical data and the spy
The similarity measurement of data is levied, specifically:
It parses the historical data and obtains sampling feature vectors;
It calculates each sampling feature vectors and predicts the Euclidean distance of day character vector.
In a second possible implementation of the first aspect of the invention, described to assign training power to each data sample
Weight, specifically:
The Euclidean distance of a certain data sample of correspondence is compared with preset distance threshold, when the Euclidean distance is small
The weight coefficient of the data sample is set as constant 1 when the distance threshold;Described in being more than or equal to when the Euclidean distance
The weight coefficient of the data sample is set as to the ratio of the distance threshold Yu the Euclidean distance when distance threshold.
It is described based on the short-term of deep neural network in the third possible implementation of first aspect present invention
Load forecasting method, further includes: the prediction day character vector and sampling feature vectors that get are normalized, and
Anti-normalization processing is carried out to the prediction result of output.
Second aspect, the embodiment of the invention provides a kind of short-term load forecasting device based on deep neural network, packet
It includes:
Data acquisition module, for obtaining historical data and characteristic to be predicted;
Similarity measurement computing module, for calculating the similarity measurement of the historical data and the characteristic;
Training set generation module, for picking out several data samples from the historical data according to the similarity measurement
This, and training weight is assigned to each data sample, obtain sample training collection;
Model training module is obtained pre- for being trained using the sample training collection to deep neural network model
Survey model;
As a result prediction result is calculated by the prediction model for the characteristic in output module.
In a first possible implementation of the second aspect of the invention, the similarity measurement computing module, is also wrapped
It includes:
Data resolution module obtains sampling feature vectors for parsing the historical data;
Distance calculation module, for calculating each sampling feature vectors and predicting the Euclidean distance of day character vector.
In second of possible implementation of second aspect of the present invention, the training set generation module, further includes:
Weight calculation module, for carrying out pair the Euclidean distance and preset distance threshold that correspond to a certain data sample
Than the weight coefficient of the data sample is set as constant 1 when the Euclidean distance is less than the distance threshold;When the Europe
The weight coefficient of the data sample is set as the distance threshold and the Europe when family name's distance is more than or equal to the distance threshold
The ratio of family name's distance.
It is described based on the short-term of deep neural network in the third possible implementation of second aspect of the present invention
Load prediction device, further includes:
Normalized module, for place to be normalized to the prediction day character vector and sampling feature vectors that get
Reason, and anti-normalization processing is carried out to the prediction result of output.
Compared with the prior art, the embodiment of the present invention has the following beneficial effects:
A kind of short-term load forecasting method based on deep neural network of the invention has been recorded short-term negative by acquisition
Characteristic to be predicted needed for the historical data and short-term load forecasting of lotus prediction, and calculate the historical data and institute
The similarity measurement for stating characteristic, with the similitude between the similarity measurement characterize data.It is filtered out according to similitude
More targeted data are used for model training, to improve the quality of training sample, the data for rejecting interference model training are same
When reduce training data volume to mitigate calculating work load;Further, data sample is assigned according to similarity measurement and is instructed
Practice weight, is conducive to Optimized model training process, to keep deep neural network more intelligent, it is short-term to have more professional processing
The ability of load prediction work, and the prediction result for exporting prediction model are closer practical, and then it is pre- to promote short term
The accuracy and precision of survey.
Detailed description of the invention
Fig. 1 is the basic step of short-term load forecasting method of one of the embodiment of the present invention based on deep neural network
Flow chart;
Fig. 2 is the specific steps of short-term load forecasting method of one of the embodiment of the present invention based on deep neural network
Flow chart;
Fig. 3 is the structure chart of short-term load forecasting device of one of the embodiment of the present invention based on deep neural network.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Please refer to Fig. 1 it illustrates an illustrative embodiment of the invention provide it is a kind of based on the short of deep neural network
Phase load forecasting method, step include:
S101, historical data and characteristic to be predicted are obtained;Wherein, the historical data include have detected that it is negative
Charge values, historical forecast load value, history short-term load forecasting accuracy value etc.;The characteristic includes obtaining before this is predicted
Route variable, weather variable etc..
S102, the similarity measurement for calculating the historical data and the characteristic;
Wherein, it usually needs to estimate the similarity measurement between different samples when doing and classifying, the side at this moment generallyd use
Method is exactly " distance " between calculating sample, therefore calculates similarity measurement and can use these methods: Euclidean distance, Manhattan away from
With a distance from, Chebyshev, Minkowski Distance, standardization Euclidean distance, mahalanobis distance, included angle cosine, Hamming distance, outstanding card
Moral distance and Jie Kade similarity factor, related coefficient and correlation distance, comentropy.
S103, several data samples are picked out from the historical data according to the similarity measurement and to each data
Sample assigns training weight, obtains sample training collection;
S104, deep neural network model is trained using the sample training collection, obtains prediction model;Wherein,
The deep neural network model is the primary mold constructed, due to not having the ability of load prediction, it is therefore desirable to pass through
Special training is trained model and is conducive to improve load prediction precision.
Prediction result is calculated by the prediction model in S105, the characteristic.
A kind of short-term load forecasting method based on deep neural network of the embodiment of the present invention has been recorded by acquisition
Characteristic to be predicted needed for the historical data and short-term load forecasting of short-term load forecasting, and calculate the history number
According to the similarity measurement with the characteristic, with the similitude between the similarity measurement characterize data.According to similitude
More targeted data are filtered out for model training, to improve the quality of training sample, reject interference model training
Data reduce trained data volume simultaneously to mitigate calculating work load;Further, according to similarity measurement to data sample
Training weight is assigned, is conducive to Optimized model training process, to keep deep neural network more intelligent, has more professional place
The ability of short-term load forecasting work is managed, and the prediction result for exporting prediction model is closer practical, and then is promoted short-term
The accuracy and precision of load prediction.
Preferably, the similarity measurement for calculating the historical data and the characteristic, specifically:
It parses the historical data and obtains sampling feature vectors;
It calculates each sampling feature vectors and predicts the Euclidean distance of day character vector.
Wherein, the prediction day character vector and the sampling feature vectors select day meteorology as eigen vector, feature
Variable in vector comprising temperature, wind speed, humidity etc. with different dimensions.
It is understood that these meteorological index of the temperature of one day, wind speed, humidity are the feature of this day, if two days
Characteristic is more similar, then the level of this two days electric loads is also more similar.
And after being normalized, calculates each sampling feature vectors and predict the Euclidean distance of day character vector, use
Similarity degree between two features of measurement.Euclidean distance is a kind of method for calculating and clustering between two points in n-dimensional space,
Calculation formula is as follows:
Wherein, x=(x1,x2,…,xn) and y=(y1,y2,…,yn) it is corresponding be two n dimension sampling feature vectors with
Predict day character vector.
It is understood that the Euclidean distance between two points is closer, show that the similarity of two points is higher.According to above formula
Euclidean distance calculated is as a result, pick out with highest 150 data samples of day eigen vector similarity to be predicted as instruction
Practice sample, to guarantee that training sample and the degree of correlation for predicting day data are higher, and guarantees certain training samples number.
It is preferably, described to assign training weight to each data sample, specifically:
The Euclidean distance of a certain data sample of correspondence is compared with preset distance threshold, when the Euclidean distance is small
The weight coefficient of the data sample is set as constant 1 when the distance threshold;Described in being more than or equal to when the Euclidean distance
The weight coefficient of the data sample is set as to the ratio of the distance threshold Yu the Euclidean distance when distance threshold.
In the present embodiment, to the sample selected, conventional training method is that all samples are uniformly treated, and in fact,
The higher training sample with prediction day sample similarity degree, algorithm more answers selective learning, in view of considerations above, to the sample selected
Assign different weights.According to the bigger principle of the smaller weight of Euclidean distance, formulate following Weight Algorithm: Euclidean distance is less than c
Sample weight coefficient ωs=1;The weight coefficient ω of sample of the Euclidean distance more than or equal to cs=c/ds.Wherein, c is distance
Threshold value, dsIt is the Euclidean distance of s-th of sample data and prediction day data.
Preferably, the short-term load forecasting method based on deep neural network, further includes: to the prediction day got
Feature vector and sampling feature vectors are normalized, and carry out anti-normalization processing to the prediction result of output.
Wherein, place is normalized to sampling feature vectors and prediction day character vector using z-score method for normalizing
Reason, calculation formula are as follows:
In formula (2), x*For the feature vector after normalization, x is to normalized feature vector, and μ is sample average, and σ is
Sample standard deviation.
Meanwhile in order to make output quantity revert to actual value, need to carry out anti-normalization processing, meter to the load data of output
It is as follows to calculate formula:
X=x*σ+μ (3)
In formula (3), x*For the normalization predicted value of model output, x is to normalized sample value, and μ is sample average, σ
For sample standard deviation.
Please refer to Fig. 2 it illustrates an illustrative embodiment of the invention provide it is a kind of based on the short of deep neural network
Phase load forecasting method, specific steps are as follows:
Historical data and prediction day character vector are obtained, and the historical data is parsed to obtain sampling feature vectors;
The prediction day character vector described in each sample characteristics vector sum is normalized;
Calculate the Euclidean distance that day character vector is predicted described in each sample characteristics vector sum;
Pick out the smallest data sample of Euclidean distance of 150 sampling feature vectors and the prediction day character vector;
The weight that each data sample is calculated according to the Euclidean distance obtains the sample training collection with sample weights;
Deep neural network model is trained using the sample training collection, obtains prediction model;
Tentative prediction result is calculated by the prediction model in prediction day character vector after normalized;
Anti-normalization processing is carried out to the tentative prediction result, obtains final prediction result.
In the present embodiment, normalized can be improved precision, this is being related to the algorithm timeliness of some distance calculating
Fruit is significant, for example algorithm will calculate Euclidean distance.In the multiple criteria system, since the property of each evaluation index is different, lead to
Often with having different dimension and the order of magnitude.When the level between each index differs greatly, if directly carried out with original index value
Analysis, will protrude effect of the higher index of numerical value in comprehensive analysis, the opposite effect for weakening the horizontal lower index of numerical value.
Therefore, it in order to guarantee the reliability of result, needs that original index data are normalized.From empirically saying, normalize
Processing is that allow the feature between different dimensions numerically to have certain comparative, can greatly improve in deep neural network and classify
The accuracy of device, while normalized is conducive to the convergence rate and precision of lift scheme, and in deep learning, data
Normalization can prevent model gradient from exploding.
Please refer to Fig. 3 it illustrates an illustrative embodiment of the invention provide it is a kind of based on the short of deep neural network
Phase load prediction device, comprising:
Data acquisition module 201, for obtaining historical data and characteristic to be predicted;
Similarity measurement computing module 202, for calculating the similarity measurement of the historical data and the characteristic;
Training set generation module 203, for picking out several numbers from the historical data according to the similarity measurement
Training weight is assigned according to sample, and to each data sample, obtains sample training collection;
Model training module 204 is obtained for being trained using the sample training collection to deep neural network model
Prediction model;
As a result prediction result is calculated by the prediction model for the characteristic in output module 205.
Preferably, the similarity measurement computing module, further includes:
Data resolution module obtains sampling feature vectors for parsing the historical data;
Distance calculation module, for calculating each sampling feature vectors and predicting the Euclidean distance of day character vector.
Preferably, the training set generation module, further includes:
Weight calculation module, for carrying out pair the Euclidean distance and preset distance threshold that correspond to a certain data sample
Than the weight coefficient of the data sample is set as constant 1 when the Euclidean distance is less than the distance threshold;When the Europe
The weight coefficient of the data sample is set as the distance threshold and the Europe when family name's distance is more than or equal to the distance threshold
The ratio of family name's distance.
Preferably, the short-term load forecasting device based on deep neural network, further includes:
Normalized module, for place to be normalized to the prediction day character vector and sampling feature vectors that get
Reason, and anti-normalization processing is carried out to the prediction result of output.
The above is a preferred embodiment of the present invention, it is noted that for those skilled in the art
For, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also considered as
Protection scope of the present invention.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the program can be stored in a computer-readable storage medium
In, the program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, the storage medium can be magnetic
Dish, CD, read-only memory (Read-Only Memory, ROM) or random access memory (Random Access
Memory, RAM) etc..
Claims (8)
1. a kind of short-term load forecasting method based on deep neural network characterized by comprising
Obtain historical data and characteristic to be predicted;
Calculate the similarity measurement of the historical data and the characteristic;
Several data samples are picked out from the historical data according to the similarity measurement and each data sample is assigned
Training weight, obtains sample training collection;
Deep neural network model is trained using the sample training collection, obtains prediction model;
Prediction result is calculated by the prediction model in the characteristic.
2. as described in claim 1 based on the short-term load forecasting method of deep neural network, which is characterized in that the calculating
The similarity measurement of the historical data and the characteristic, specifically:
It parses the historical data and obtains sampling feature vectors;
It calculates each sampling feature vectors and predicts the Euclidean distance of day character vector.
3. as claimed in claim 2 based on the short-term load forecasting method of deep neural network, which is characterized in that described to every
One data sample assigns training weight, specifically:
The Euclidean distance of a certain data sample of correspondence is compared with preset distance threshold, when the Euclidean distance is less than institute
The weight coefficient of the data sample is set as constant 1 when stating distance threshold;When the Euclidean distance is more than or equal to the distance
The weight coefficient of the data sample is set as to the ratio of the distance threshold Yu the Euclidean distance when threshold value.
4. as claimed in claim 2 based on the short-term load forecasting method of deep neural network, which is characterized in that further include:
The prediction day character vector and sampling feature vectors that get are normalized, and the prediction result of output is carried out
Anti-normalization processing.
5. a kind of short-term load forecasting device based on deep neural network characterized by comprising
Data acquisition module, for obtaining historical data and characteristic to be predicted;
Similarity measurement computing module, for calculating the similarity measurement of the historical data and the characteristic;
Training set generation module, for picking out several data samples from the historical data according to the similarity measurement,
And training weight is assigned to each data sample, obtains sample training collection;
Model training module obtains prediction mould for being trained using the sample training collection to deep neural network model
Type;
As a result prediction result is calculated by the prediction model for the characteristic in output module.
6. the short-term load forecasting device based on deep neural network as claimed in claim 5, which is characterized in that described similar
Property metric calculation module, further includes:
Data resolution module obtains sampling feature vectors for parsing the historical data;
Distance calculation module, for calculating each sampling feature vectors and predicting the Euclidean distance of day character vector.
7. the short-term load forecasting device based on deep neural network as claimed in claim 6, which is characterized in that the training
Collect generation module, further includes:
Weight calculation module, for the Euclidean distance for corresponding to a certain data sample to be compared with preset distance threshold, when
The weight coefficient of the data sample is set as constant 1 when the Euclidean distance is less than the distance threshold;When the Euclidean away from
From when being more than or equal to the distance threshold by the weight coefficient of the data sample be set as the distance threshold and the Euclidean away from
From ratio.
8. the short-term load forecasting device based on deep neural network as claimed in claim 6, which is characterized in that further include:
Normalized module, for the prediction day character vector and sampling feature vectors that get to be normalized,
And anti-normalization processing is carried out to the prediction result of output.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910404028.4A CN110222727A (en) | 2019-05-15 | 2019-05-15 | A kind of short-term load forecasting method and device based on deep neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910404028.4A CN110222727A (en) | 2019-05-15 | 2019-05-15 | A kind of short-term load forecasting method and device based on deep neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110222727A true CN110222727A (en) | 2019-09-10 |
Family
ID=67821164
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910404028.4A Pending CN110222727A (en) | 2019-05-15 | 2019-05-15 | A kind of short-term load forecasting method and device based on deep neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110222727A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111461306A (en) * | 2020-03-31 | 2020-07-28 | 北京百度网讯科技有限公司 | Feature evaluation method and device |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102646198A (en) * | 2012-02-21 | 2012-08-22 | 温州大学 | Mode recognition method of mixed linear SVM (support vector machine) classifier with hierarchical structure |
JP2012244897A (en) * | 2011-05-13 | 2012-12-10 | Fujitsu Ltd | Apparatus and method for predicting short-term power load |
CN106600463A (en) * | 2016-12-21 | 2017-04-26 | 广东电网有限责任公司电力调度控制中心 | Local shape similarity ultra short-period load prediction method and apparatus |
CN107016469A (en) * | 2017-04-13 | 2017-08-04 | 重庆大学 | Methods of electric load forecasting |
CN107766995A (en) * | 2017-12-08 | 2018-03-06 | 国网山东省电力公司枣庄供电公司 | Power-system short-term load forecasting method based on depth recurrent neural network |
WO2018124918A1 (en) * | 2016-12-27 | 2018-07-05 | Общество с ограниченной ответственностью "ВижнЛабс" | Training of deep neural networks on the basis of distributions of paired similarity measures |
CN108416307A (en) * | 2018-03-13 | 2018-08-17 | 北京理工大学 | A kind of Aerial Images road surface crack detection method, device and equipment |
CN108985323A (en) * | 2018-06-04 | 2018-12-11 | 江苏大学 | A kind of short term prediction method of photovoltaic power |
CN109389238A (en) * | 2017-08-14 | 2019-02-26 | 中国电力科学研究院 | A kind of short-term load forecasting method and device based on ridge regression |
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 |
CN109636010A (en) * | 2018-11-23 | 2019-04-16 | 国网湖北省电力有限公司 | Provincial power network short-term load forecasting method and system based on correlative factor matrix |
-
2019
- 2019-05-15 CN CN201910404028.4A patent/CN110222727A/en active Pending
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2012244897A (en) * | 2011-05-13 | 2012-12-10 | Fujitsu Ltd | Apparatus and method for predicting short-term power load |
CN102646198A (en) * | 2012-02-21 | 2012-08-22 | 温州大学 | Mode recognition method of mixed linear SVM (support vector machine) classifier with hierarchical structure |
CN106600463A (en) * | 2016-12-21 | 2017-04-26 | 广东电网有限责任公司电力调度控制中心 | Local shape similarity ultra short-period load prediction method and apparatus |
WO2018124918A1 (en) * | 2016-12-27 | 2018-07-05 | Общество с ограниченной ответственностью "ВижнЛабс" | Training of deep neural networks on the basis of distributions of paired similarity measures |
CN107016469A (en) * | 2017-04-13 | 2017-08-04 | 重庆大学 | Methods of electric load forecasting |
CN109389238A (en) * | 2017-08-14 | 2019-02-26 | 中国电力科学研究院 | A kind of short-term load forecasting method and device based on ridge regression |
CN107766995A (en) * | 2017-12-08 | 2018-03-06 | 国网山东省电力公司枣庄供电公司 | Power-system short-term load forecasting method based on depth recurrent neural network |
CN108416307A (en) * | 2018-03-13 | 2018-08-17 | 北京理工大学 | A kind of Aerial Images road surface crack detection method, device and equipment |
CN108985323A (en) * | 2018-06-04 | 2018-12-11 | 江苏大学 | A kind of short term prediction method of photovoltaic power |
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 |
CN109636010A (en) * | 2018-11-23 | 2019-04-16 | 国网湖北省电力有限公司 | Provincial power network short-term load forecasting method and system based on correlative factor matrix |
Non-Patent Citations (6)
Title |
---|
何琬等: "基于深层神经网络的电力负荷预测", 《环境与可持续发展》 * |
何琬等: "基于深层神经网络的电力负荷预测", 《环境与可持续发展》, no. 01, 30 January 2016 (2016-01-30), pages 83 - 87 * |
金海峰等: "基于相似性原理的短期负荷预测方法", 《电力系统自动化》 * |
金海峰等: "基于相似性原理的短期负荷预测方法", 《电力系统自动化》, no. 23, 15 December 2001 (2001-12-15), pages 45 - 48 * |
黄宇腾等: "深度神经网络在配电网公变短期负荷预测中的应用研究", 《浙江电力》 * |
黄宇腾等: "深度神经网络在配电网公变短期负荷预测中的应用研究", 《浙江电力》, vol. 37, no. 05, 28 May 2018 (2018-05-28), pages 1 - 6 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111461306A (en) * | 2020-03-31 | 2020-07-28 | 北京百度网讯科技有限公司 | Feature evaluation method and device |
CN111461306B (en) * | 2020-03-31 | 2023-04-18 | 北京百度网讯科技有限公司 | Feature evaluation method and device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108734355B (en) | Short-term power load parallel prediction method and system applied to power quality comprehensive management scene | |
CN108333468B (en) | The recognition methods of bad data and device under a kind of active power distribution network | |
CN110991786A (en) | 10kV static load model parameter identification method based on similar daily load curve | |
CN112732871B (en) | Multi-label classification method for acquiring client intention labels through robot induction | |
CN111028100A (en) | Refined short-term load prediction method, device and medium considering meteorological factors | |
CN112735097A (en) | Regional landslide early warning method and system | |
CN111929633A (en) | Electric energy meter detection system and method based on fusion ant colony algorithm | |
CN113408341B (en) | Load identification method and device, computer equipment and storage medium | |
CN113361785A (en) | Power distribution network short-term load prediction method and device, terminal and storage medium | |
CN109657884A (en) | Power grid power supply optimization method, apparatus, equipment and computer readable storage medium | |
CN112529638A (en) | Service demand dynamic prediction method and system based on user classification and deep learning | |
CN109257383A (en) | A kind of BGP method for detecting abnormality and system | |
CN115660182A (en) | Photovoltaic output prediction method based on maximum expected sample weighted neural network model | |
CN115204698A (en) | Real-time analysis method for power supply stability of low-voltage transformer area | |
Ullah et al. | Adaptive data balancing method using stacking ensemble model and its application to non-technical loss detection in smart grids | |
CN110222727A (en) | A kind of short-term load forecasting method and device based on deep neural network | |
CN109858667A (en) | It is a kind of based on thunder and lightning weather to the short term clustering method of loading effects | |
CN110222610A (en) | A kind of method and device of Modulation recognition | |
CN113298318A (en) | Novel overload prediction method for distribution transformer | |
CN113704389A (en) | Data evaluation method and device, computer equipment and storage medium | |
CN110991638B (en) | Generalized load modeling method based on clustering and neural network | |
CN112115984A (en) | Tea garden abnormal data correction method and system based on deep learning and storage medium | |
CN113837266B (en) | Software defect prediction method based on feature extraction and Stacking ensemble learning | |
CN115409132A (en) | Method and system for processing power distribution network data | |
CN114118812A (en) | Hydropower station energy efficiency analysis and evaluation method and device based on improved fuzzy mean clustering |
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 |