CN113673579B - Small sample-based electricity load classification algorithm - Google Patents
Small sample-based electricity load classification algorithm Download PDFInfo
- Publication number
- CN113673579B CN113673579B CN202110853474.0A CN202110853474A CN113673579B CN 113673579 B CN113673579 B CN 113673579B CN 202110853474 A CN202110853474 A CN 202110853474A CN 113673579 B CN113673579 B CN 113673579B
- Authority
- CN
- China
- Prior art keywords
- sample
- samples
- classification
- training
- algorithm
- 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
- 238000007635 classification algorithm Methods 0.000 title claims abstract description 12
- 230000005611 electricity Effects 0.000 title description 6
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 39
- 238000012549 training Methods 0.000 claims abstract description 35
- 230000008859 change Effects 0.000 claims abstract description 16
- 238000012706 support-vector machine Methods 0.000 claims abstract description 14
- 238000012360 testing method Methods 0.000 claims abstract description 13
- 238000003066 decision tree Methods 0.000 claims abstract description 12
- 238000000034 method Methods 0.000 claims abstract description 7
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 238000005457 optimization Methods 0.000 claims abstract description 4
- 238000011156 evaluation Methods 0.000 claims description 15
- 230000007547 defect Effects 0.000 claims description 3
- 238000005315 distribution function Methods 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000002194 synthesizing effect Effects 0.000 claims description 3
- 238000004904 shortening Methods 0.000 claims description 2
- 230000001747 exhibiting effect Effects 0.000 claims 1
- 238000001514 detection method Methods 0.000 description 3
- 230000011218 segmentation Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/285—Selection of pattern recognition techniques, e.g. of classifiers in a multi-classifier system
-
- 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
- G06F18/2148—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the process organisation or structure, e.g. boosting cascade
-
- 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
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides an electric load classification algorithm based on a small sample, which comprises the steps of firstly extracting the characteristics of different non-steady change load samples, preprocessing, expanding and evaluating the characteristics through network training, carrying out mixed training on simplified data through K adjacent algorithm, support vector machine and decision tree mixed training, and finally adjusting the weight of each algorithm on the classification result precision through a weighted optimization mode, evaluating the weight, wherein the method can be used for testing an actual sample when the classification precision of a mixed model meets the condition. Has the following advantages: 1. the original sample has good representativeness and general applicability; 2. expanding the consistency of samples before and after expansion; 3. the limitations of a single algorithm can be overcome by adopting a mixed classifier and a weighting mode.
Description
Technical Field
The invention relates to an electric load classification algorithm, in particular to an electric load classification algorithm based on a small sample.
Background
Along with large-scale use of large-scale integration of clean energy sources such as wind power, photovoltaic and the like and electric automobiles in the electric power system, the electric load components of the electric power system are more complex, analysis of the electric load components of the electric power system is favorable for power grid personnel to formulate reasonable demands according to the electric load rules and implement corresponding measures, once the electric load is not clear in classification, the load of an electric power department is mispredicted due to light weight, the electric power system is disordered in scheduling, the electric power marketing rule is destroyed due to heavy weight, and huge economic loss is caused. Therefore, the development of the electric load classification research in the electric power system has very important significance.
At present, the power load classification adopts a data-driven mode for analysis, has the characteristics of high efficiency, simple operation, operation cost reduction and the like, but has the problems of unbalanced data classification, low analysis result precision and the like, and is specifically expressed as follows: 1. in an actual power grid, a sample with normal power consumption load is far larger than a sample with fluctuation of the power consumption load, so that data unbalance caused by the sample is capable of negatively influencing the analysis performance of data driving; 2. the analysis algorithm is mainly a single algorithm and an improved algorithm thereof, has obvious property, is easy to sink into a local optimal solution, and greatly reduces the reliability of an analysis result.
Disclosure of Invention
The invention mainly solves the technical problems existing in the prior art: the sample expansion algorithm is provided for expanding a small amount of power load samples which show non-steady change, and the problem of unbalanced sample distribution is solved.
Still another object of the present invention is to solve the technical problems of the prior art: the state classification algorithm is provided, load data is classified through mixed analysis and comprehensive decision of multiple intelligent algorithms, the limitation of a single algorithm is overcome, and efficient identification of multiple power utilization load states is realized.
The technical problems of the invention are mainly solved by the following technical proposal:
An electrical load classification algorithm based on a small sample comprises the following steps:
Step 1, sample expansion: extracting characteristics of different non-steady change load samples, preprocessing the samples, expanding the processed samples in a network training mode, judging the generated samples according to evaluation indexes, outputting the samples when the conditions are met, and forming a database of a state classification link together with other small quantity of non-steady change power consumption load samples and steady change power consumption load samples.
Step 2, classifying states: the method comprises the steps of obtaining training data from a database, extracting features of different data, simplifying and shortening a classification flow, adopting a K-nearest algorithm, a support vector machine and a decision tree for mixed training, making up the defects of a single algorithm, adjusting weights of the algorithms on classification result precision in a weighted optimization mode, providing classification precision, judging by combining evaluation indexes, and being capable of being used for actual sample testing when the classification precision of a mixed model meets the condition.
In the above-mentioned small sample-based power load classification algorithm, in the step 1, the specific operation method of sample expansion is as follows:
And step 1.1, acquiring samples under different non-stationary change electricity load conditions in a certain time period in different time periods of 0-24 hours a day.
And 1.2, preprocessing samples with different characteristics, and eliminating influences caused by differences of dimensions, measurement intervals and the like in a normalization mode.
And 1.3, adding noise to the preprocessed sample set, and establishing a linear model based on a normal distribution function, wherein a proper sample can be considered to be generated when constraint conditions are met.
And 1.4, introducing an evaluation index to quantitatively analyze the expanded sample set, and considering that the expanded sample has higher reliability when the condition is met.
In the above-mentioned small sample-based power load classification algorithm, in the step 2, the specific operation method of the state classification is as follows:
and 2.1, acquiring data from a database, dividing the data into training data and test data according to a proportion, and extracting entropy characteristic values of the data.
And 2.2, performing mixed training by KNN, SVM, DT according to the entropy characteristic value of the sample, and acquiring training errors when the training accuracy of a single algorithm meets the requirement.
And 2.3, distributing weights of the algorithms according to training parameters of the hybrid classifier, and determining a hybrid training result by the weighted results of the algorithms.
And 2.4, synthesizing the training data and the test data, quantitatively analyzing the precision of the mixed classification according to the evaluation index, and if the precision meets the condition, using the mixed classification for testing an actual sample.
Therefore, the invention has the following advantages: 1. the diversity of non-stationary samples is considered, common types of power users are simplified, and the samples have good representativeness and general applicability and are closer to the actual use condition; 2. the small samples are expanded, preprocessing, noise adding and evaluation are adopted, and when the characteristics of different small samples are considered, the difficulty of expanding the large samples by the small samples is also reduced, and the consistency of the samples before and after expansion is ensured; 3. the load data is classified by adopting a hybrid classifier and a weighting mode through hybrid analysis and comprehensive decision of a plurality of intelligent algorithms, the limitation of a single algorithm is overcome, and the efficient identification of a plurality of power utilization load states is realized.
Drawings
Fig. 1 is a main flow chart of the implementation of the present invention.
FIG. 2 is a main flow chart of the electrical load sample expansion of the present invention.
Fig. 3 is a main flow chart of the classification of the electric load state of the present invention.
Detailed Description
The technical scheme of the invention is further specifically described below through examples and with reference to the accompanying drawings.
As shown in figure 1, the invention sequentially adopts sample expansion and state classification modes to realize accurate classification of the electric load based on small samples, wherein the sample expansion expands a small amount of electric load samples with unstable change, solves the problem of unbalanced sample distribution, classifies load data through mixed analysis and comprehensive decision of multiple intelligent algorithms, overcomes the limitation of a single algorithm, and realizes efficient identification of multiple electric load states.
As shown in fig. 2, in the sample expansion link, the characteristics of different non-steady change load samples are extracted, the samples are preprocessed, the processed samples are expanded in a network training mode, the generated samples are judged according to evaluation indexes, and the samples are output when the evaluation indexes meet the conditions, and the samples, other small quantity of non-steady change power consumption load samples and steady change power consumption load samples form a database of a state classification link. The main steps of the link are as follows:
And step 1.1, acquiring samples under different non-stationary change electricity load conditions in a certain time period in different time periods of 0-24 hours a day.
The invention divides the sample set of the user load of the power system into four main categories: community resident ([I1,U1,P1,T1]、[I2,U2,P2,T2]、[I3,U3,P3,T3])、 individual business unit ([I4,U4,P4,T4]、[I5,U5,P5,T5]、[I6,U6,P6,T6])、 business unit ([I7,U7,P7,T7]、[I8,U8,P8,T8]、[I9,U9,P9,T9])、 large enterprise ([I10,U10,P10,T10]、[I11,U11,P11,T11]、[I12,U12,P12,T12]), wherein ,[I1,U1,P1,T1]、[I4,U4,P4,T4]、[I7,U7,P7,T7]、[I10,U10,P10,T10] represents community resident, individual business unit, large enterprise normal power load in period T 1、T4、T7、T10, I 1、I4、I7、I10 represents current set of user in the period, U 1、U4、U7、U10 represents voltage set of user in the period, P 1、P4、P7、P10 represents power set ,[I2,U2,P2,T2]、[I5,U5,P5,T5]、[I8,U8,P8,T8]、[I11,U11,P11,T11] of user in the period, community resident, Individual merchants, institutions and large enterprises peak electricity loads in a period T 2、T5、T8、T11, I 2、I5、I8、I11 represents the current set of users in the period, U 2、U5、U8、U11 represents the voltage set of users in the period, P 2、P5、P8、P11 represents the power set ,[I3,U3,P3,T3]、[I6,U6,P6,T6]、[I9,U9,P9,T9]、[I12,U12,P12,T12] of users in the period, community residents, individual merchants, institutions and large enterprises valley electricity loads in a period T 3、T6、T9、T12, I 3、I6、I9、I12 represents the current set of users in the period, U 3、U6、U9、U12 represents the voltage set of users in the period, p 3、P6、P9、P12 represents the power set of the user during this period.
And 1.2, preprocessing samples with different characteristics, and eliminating influences caused by differences of dimensions, measurement intervals and the like in a normalization mode.
B represents the normalized sample set, a i represents the user load set, (a i)max represents the maximum load in the set, and a i)min represents the minimum load in the set.
And 1.3, adding noise to the preprocessed sample set, and establishing a linear model based on a normal distribution function, wherein a proper sample can be considered to be generated when constraint conditions are met.
B′=B+α (3)
B 'represents the set of samples after noise is mixed in, α represents the set of noise, f (B') represents the linear model, B 1 represents the gaussian function coefficient, B 2 represents the function bias, B represents the dimension of the set of samples B,Representing the mean of the sample set B ', δ B' represents the variance of the sample set B ', and f ' (B ') represents differentiating the sample set B '.
To reduce the negative impact of sample set, noise set magnitude spacing, when the constraint of equation (6) is satisfied, it is considered that samples are generated that have consistency with the input samples. And when the constraint condition of the formula (6) is not satisfied, returning to the formulas (4) and (5) and adjusting the b 1、b2 function coefficients until the constraint condition of the formula (6) is satisfied. In particular, the present invention is not limited to a particular mode of function coefficient adjustment.
According to the invention, on the basis of acquiring different characteristic sample sets in step 1.1, a sample with fluctuation change is expanded in a mixed noise mode, a linear model based on a Gaussian function is established in view of universality of probability density distribution, a new sample is formed in an inverse solving mode, and when constraint conditions are met, the new sample with consistency with an input sample is considered to be generated for evaluating and judging whether the conditions are met in step 1.4.
And 1.4, introducing an evaluation index to quantitatively analyze the expanded sample set, and considering that the expanded sample has higher reliability when the condition is met.
C3=||f'(B')||2 (9)
C 1 represents the overlapping degree of the sample set f '(B') in the global range, C 4 represents the ratio of the average value of the sample set f '(B') to the minimum value thereof in the global range, C 5 represents the ratio of the average value of the sample set f '(B') to the maximum value thereof in the global range, C 2 represents the contour width of the sample set f '(B'), and C 3 represents the sum of the squares of the distances of the sample set f '(B').
The invention provides a global overlap degree C 1, a contour width C 2 and a distance square sum C 3 as evaluation indexes, wherein a floating interval of the global overlap degree C 1 is [0,1], the similarity degree of an extended sample set and an original sample set is represented, the closer the value is to 1, and the extended sample generated by inverse solution and the original sample have consistency in a dimension range; the profile width C 2 is [ -1,1], which represents the approaching degree of the extended sample set curve and the original sample set curve, the closer the value is to 1, the more the variation trend of the load curves of the extended sample and the original sample is the same; the floating interval of the square sum of distances C 3 is [0, ++ infinity ], which represents the tightness degree between different characteristic samples and the original sample of the drink after expansion, and the closer the value is to 0, the more compact the different characteristic objects in the expanded samples are. In particular, the invention is not limited to the criteria for qualifying a particular sample, i.e., making decisions with reference to actual conditions.
As shown in figure 3, in the state classification link, training data are acquired from a database, characteristics of different data are extracted, classification flow is shortened, K-nearest neighbor (KNN) algorithm, support Vector Machine (SVM) and Decision Tree (DT) are adopted for mixed training, the defects of a single algorithm are overcome, the weight of each algorithm on classification result precision is adjusted in a weighted optimization mode, classification accuracy is provided, and judgment is carried out by combining with evaluation indexes, so that the method can be used for actual sample test when the classification precision of a mixed model meets the condition. The main steps of the link are as follows:
and 2.1, acquiring data from a database, dividing the data into training data and test data according to a proportion, and extracting entropy characteristic values of the data.
D(B')=∑f'(B')logf'(B') (10)
D (B') is the entropy feature value set of the data.
Through the inverse solution of the step 1, the complexity of the data information is greatly simplified, and the invention provides that the dimension characteristics of the data can be further represented in an information entropy mode.
And 2.2, performing mixed training by KNN, SVM, DT according to the entropy characteristic value of the sample, and acquiring training errors when the training accuracy of a single algorithm meets the requirement.
ESVM=β(xj,yj)+e1 (12)
EDT=∑E1(xj,yj)=e2 (13)
xj、yj∈D(B') (14)
Samples x j、yj are derived from an entropy feature value set D (B'), E KNN represents the sum of Euclidean distances of different sample points, E SVM represents a hyperplane segmentation curve of the samples, beta represents an SVM kernel function, E 1 represents a bias term of the SVM kernel function, the bias term is constant, E DT represents an entropy decision judging function of the samples, and E 2 represents the types of different feature sample sets.
The invention provides training by using a hybrid classifier, namely training entropy feature samples by adopting a KNN algorithm, an SVM algorithm and a DT algorithm, improving the reliability of the algorithm, and is characterized in that the invention is not limited to defining specific spacing curved surfaces, segmentation curves and decision functions, and the types of feature sample sets depend on the samples and can be adjusted according to actual conditions.
And 2.3, distributing weights of the algorithms according to training parameters of the hybrid classifier, and determining a hybrid training result by the weighted results of the algorithms.
Χ k corresponds to the weight coefficient of KNN, SVM, DT respectively, δ k corresponds to the output error of KNN, SVM, DT respectively, R represents the voting result, when it is 1, the KNN detection result is based, when it is 2, the SVM detection result is based, when it is 3, the DT detection result is based.
The invention optimizes the data result by a weighting adjustment mode on the basis of mixed classification, further improves the accuracy of the algorithm, and is particularly not limited to the error calculation mode of a single algorithm.
And 2.4, synthesizing the training data and the test data, quantitatively analyzing the precision of the mixed classification according to the evaluation index, and if the precision meets the condition, using the mixed classification for testing an actual sample.
While the basic principles and advantages of the present invention have been described above, it will be understood by those skilled in the art that the present invention is not limited to the above-described embodiments, and that the above-described embodiments and descriptions merely illustrate the structural relationships and principles of the present invention, and that various changes and modifications may be made therein without departing from the spirit and scope of the invention as defined in the appended claims. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (3)
1. An electrical load classification algorithm based on a small sample is characterized by comprising the following steps:
Step 1, sample expansion: extracting characteristics of different non-steady change load samples, preprocessing the samples, expanding the processed samples in a network training mode, judging the generated samples according to evaluation indexes, outputting the samples when the conditions are met, and forming a database of a state classification link together with other small quantity of non-steady change power consumption load samples and steady change power consumption load samples;
Step 2, classifying states: acquiring training data from a database, extracting features of different data, simplifying and shortening a classification flow, adopting a K-nearest algorithm, a support vector machine and a decision tree for mixed training, making up the defects of a single algorithm, adjusting the weight of each algorithm on the classification result precision in a weighted optimization mode, providing the classification precision, judging by combining with an evaluation index, and being used for actual sample test when the classification precision of a mixed model meets the condition;
The sample expansion specific operation method comprises the following steps:
step 1.1, in different time periods of 0-24 hours a day, obtaining samples under different non-steady change power load conditions in a certain time period,
Step 1.2, preprocessing samples with different characteristics, eliminating influences caused by differences of dimension, measurement interval and the like in a normalization mode,
Step 1.3, adding noise to the preprocessed sample set, establishing a linear model based on a normal distribution function, and considering that a proper sample can be generated when the constraint condition is met, specifically,
B′=B+α
B 'represents the set of samples after noise is mixed in, α represents the set of noise, f (B') represents the linear model, B 1 represents the gaussian function coefficient, B 2 represents the function bias, B represents the dimension of the set of samples B,Representing the mean of the sample set B ', δ B' representing the variance of the sample set B ', f ' (B ') representing differentiating the sample set B ',
To reduce the negative impact of sample set, noise set magnitude interval, when the constraint condition is satisfied, then consider to generate samples with consistency with the input samples, when the constraint condition is not satisfied, return to adjust b 1、b2 function coefficients until the constraint condition is satisfied,
Step 1.4, introducing an evaluation index to quantitatively analyze the expanded sample set, and considering that the expanded sample has higher reliability when the conditions are met,
The specific operation method of the state classification is as follows:
step 2.1, obtaining data from a database, dividing the data into training data and test data according to a proportion, extracting entropy characteristic values of the data,
Step 2.2, adopting KNN, SVM, DT to perform mixed training according to the entropy characteristic value of the sample, obtaining training error when the training precision of a single algorithm reaches the requirement,
Step 2.3, according to the training parameters of the mixed classifier, the weights of the algorithms are distributed, the mixed training result is determined by the weighted results of the algorithms,
And 2.4, synthesizing the training data and the test data, quantitatively analyzing the precision of the mixed classification according to the evaluation index, and if the precision meets the condition, using the mixed classification for testing an actual sample.
2. The small sample-based electrical load classification algorithm of claim 1, wherein the sample expansion expands a small number of electrical load samples exhibiting non-stationary changes, solving the problem of unbalanced sample distribution.
3. The small sample-based power load classification algorithm according to claim 1, wherein the state classification classifies load data through hybrid analysis and comprehensive decision of multiple intelligent algorithms, overcomes the limitation of a single algorithm, and realizes efficient identification of multiple power load states.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110853474.0A CN113673579B (en) | 2021-07-27 | 2021-07-27 | Small sample-based electricity load classification algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110853474.0A CN113673579B (en) | 2021-07-27 | 2021-07-27 | Small sample-based electricity load classification algorithm |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113673579A CN113673579A (en) | 2021-11-19 |
CN113673579B true CN113673579B (en) | 2024-05-28 |
Family
ID=78540532
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110853474.0A Active CN113673579B (en) | 2021-07-27 | 2021-07-27 | Small sample-based electricity load classification algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113673579B (en) |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108416366A (en) * | 2018-02-06 | 2018-08-17 | 武汉大学 | A kind of power-system short-term load forecasting method of the weighting LS-SVM based on Meteorological Index |
CN108537281A (en) * | 2018-04-13 | 2018-09-14 | 贵州电网有限责任公司 | A kind of power consumer feature recognition sorting technique based on random forest |
WO2019033636A1 (en) * | 2017-08-16 | 2019-02-21 | 哈尔滨工业大学深圳研究生院 | Method of using minimized-loss learning to classify imbalanced samples |
CN109447364A (en) * | 2018-11-08 | 2019-03-08 | 国网湖南省电力有限公司 | Power customer based on label complains prediction technique |
CN110334661A (en) * | 2019-07-09 | 2019-10-15 | 国网江苏省电力有限公司扬州供电分公司 | Infrared power transmission and transformation abnormal heating point target detecting method based on deep learning |
WO2019237492A1 (en) * | 2018-06-13 | 2019-12-19 | 山东科技大学 | Semi-supervised learning-based abnormal electricity utilization user detection method |
CN111832615A (en) * | 2020-06-04 | 2020-10-27 | 中国科学院空天信息创新研究院 | Sample expansion method and system based on foreground and background feature fusion |
CN112801123A (en) * | 2020-09-09 | 2021-05-14 | 华北电力大学 | Small sample user electricity consumption data expansion method with frequency domain distribution consistency |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9898811B2 (en) * | 2015-05-08 | 2018-02-20 | Kla-Tencor Corporation | Method and system for defect classification |
-
2021
- 2021-07-27 CN CN202110853474.0A patent/CN113673579B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019033636A1 (en) * | 2017-08-16 | 2019-02-21 | 哈尔滨工业大学深圳研究生院 | Method of using minimized-loss learning to classify imbalanced samples |
CN108416366A (en) * | 2018-02-06 | 2018-08-17 | 武汉大学 | A kind of power-system short-term load forecasting method of the weighting LS-SVM based on Meteorological Index |
CN108537281A (en) * | 2018-04-13 | 2018-09-14 | 贵州电网有限责任公司 | A kind of power consumer feature recognition sorting technique based on random forest |
WO2019237492A1 (en) * | 2018-06-13 | 2019-12-19 | 山东科技大学 | Semi-supervised learning-based abnormal electricity utilization user detection method |
CN109447364A (en) * | 2018-11-08 | 2019-03-08 | 国网湖南省电力有限公司 | Power customer based on label complains prediction technique |
CN110334661A (en) * | 2019-07-09 | 2019-10-15 | 国网江苏省电力有限公司扬州供电分公司 | Infrared power transmission and transformation abnormal heating point target detecting method based on deep learning |
CN111832615A (en) * | 2020-06-04 | 2020-10-27 | 中国科学院空天信息创新研究院 | Sample expansion method and system based on foreground and background feature fusion |
CN112801123A (en) * | 2020-09-09 | 2021-05-14 | 华北电力大学 | Small sample user electricity consumption data expansion method with frequency domain distribution consistency |
Non-Patent Citations (2)
Title |
---|
基于LS-SVM的重要用户供用电安全评价分析方法;潘明明;刘连光;田世明;徐震;;发电与空调;20160615(第03期);全文 * |
考虑类别不平衡的海量负荷用电模式辨识方法;李想;王鹏;刘洋;许立雄;;中国电机工程学报(第01期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN113673579A (en) | 2021-11-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109919353B (en) | Distributed photovoltaic prediction method of ARIMA model based on spatial correlation | |
CN112561156A (en) | Short-term power load prediction method based on user load mode classification | |
CN106845717B (en) | Energy efficiency evaluation method based on multi-model fusion strategy | |
CN111324642A (en) | Model algorithm type selection and evaluation method for power grid big data analysis | |
CN111723876B (en) | Load curve integrated spectrum clustering method considering double-scale similarity | |
CN104573879A (en) | Photovoltaic power station output predicting method based on optimal similar day set | |
CN115099511A (en) | Photovoltaic power probability estimation method and system based on optimized copula | |
CN110909977A (en) | Power grid fault diagnosis method based on ADASYN-DHSD-ET | |
CN103049651A (en) | Method and device used for power load aggregation | |
CN110738232A (en) | grid voltage out-of-limit cause diagnosis method based on data mining technology | |
CN112418476A (en) | Ultra-short-term power load prediction method | |
CN105678406A (en) | Short-term load prediction method based on cloud model | |
CN115829105A (en) | Photovoltaic power prediction method based on historical data feature search | |
CN112417308A (en) | User portrait label generation method based on electric power big data | |
CN111415049A (en) | Power failure sensitivity analysis method based on neural network and clustering | |
CN115099296A (en) | Sea wave height prediction method based on deep learning algorithm | |
CN111062539B (en) | Total electric quantity prediction method based on secondary electric quantity characteristic cluster analysis | |
CN112288157A (en) | Wind power plant power prediction method based on fuzzy clustering and deep reinforcement learning | |
CN109271424A (en) | A kind of parameter adaptive clustering method based on density | |
CN115660182A (en) | Photovoltaic output prediction method based on maximum expected sample weighted neural network model | |
CN116470491A (en) | Photovoltaic power probability prediction method and system based on copula function | |
CN113673579B (en) | Small sample-based electricity load classification algorithm | |
CN111553434A (en) | Power system load classification method and system | |
CN116070458A (en) | New wind power plant scene generation method based on RAC-GAN | |
Mao et al. | Naive Bayesian algorithm classification model with local attribute weighted based on KNN |
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 |