CN109829571A - A kind of short-term electricity demand forecasting method of user based on multilist fused data - Google Patents
A kind of short-term electricity demand forecasting method of user based on multilist fused data Download PDFInfo
- Publication number
- CN109829571A CN109829571A CN201910029664.3A CN201910029664A CN109829571A CN 109829571 A CN109829571 A CN 109829571A CN 201910029664 A CN201910029664 A CN 201910029664A CN 109829571 A CN109829571 A CN 109829571A
- Authority
- CN
- China
- Prior art keywords
- day
- fuzzy
- matrix
- vector
- multilist
- 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
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of short-term electricity demand forecasting methods of user based on multilist fused data, belong to technical field of electric power.There is certain correlations between existing user's energy.If the correlation between user's energy can be excavated fully, so that it may effectively improve the precision of prediction.The present invention chooses similar day by fuzzy clustering method first, then uses these similar days as the training sample of support vector machines, optimizes to support vector machines parameter, the model after being trained, for predicting user power consumption;The present invention can fully excavate the correlation between user's energy, rationally utilize multilist fused data, effectively improve the precision of prediction.The present invention has fully considered the influence factor of various electricity consumptions, effectively improves the precision of prediction, reduces training samples number, has great importance to user power utilization unusual checking.
Description
Technical field
The present invention relates to a kind of short-term electricity demand forecasting methods of user based on multilist fused data, belong to power technology neck
Domain.
Background technique
With society development and the arrival in energy revolution epoch, the development of smart grid and improve receive it is more and more
Attention.The development of intelligent power business can satisfy that user is novel, diversified demand, be Utilities Electric Co., power grid transformation hair
The important component of exhibition mode.
Load prediction is a basic content of Power System Analysis and management, by analysis power grid integral load or individually
The inherent trend and rule of user power consumption, it is established that suitable model predicts power load, for electric system
High-efficiency operation is of great significance.However, due to lacking user's water, with destiny evidence, previous research was all only based on electricity consumption
It measures to user's prediction that can be carried out, in practice, there is certain correlations between user's energy, if it is possible to fully dig
Dig the correlation between user's energy, so that it may effectively improve the precision of the short-term electricity demand forecasting of user.
Summary of the invention
In view of the drawbacks of the prior art, the purpose of the present invention is to provide one kind fully to excavate between user's energy
Correlation, rationally utilize multilist fused data, the user based on multilist fused data for effectively improving the precision of prediction is short
Phase electricity demand forecasting method.
To achieve the above object, the technical solution of the present invention is as follows:
A kind of short-term electricity demand forecasting method of user based on multilist fused data, calculates shadow first with path analysis
The weight for ringing the day character vector of user power consumption, obtains fuzzy similarity matrix;Then it is selected by fuzzy clustering Transitive Closure Method
Similar day is taken, and using these similar days as sample training support vector machines, user power consumption is predicted;
Specifically includes the following steps:
The first step, the determination and quantization of day character vector choose the factor for influencing user power consumption as day character vector
Element, formed sample set X={ xi};
Second step, path analysis obtain day character vector weight matrix A, and different factors is to electricity consumption in day character vector
Influence degree be different, it is therefore desirable to different weights is assigned to these elements;
Third step calculates the similarity R (x between day character vectori, xj), in order to be compared different characteristic vector to select
Similar day is taken, needs to define the similarity between two feature vectors, and then similarity calculated construction obscures similar square
Battle array is simultaneously used for fuzzy cluster analysis;
4th step, uses fuzzy clustering method to classify sample set to carry out the selection of similar day, and one fuzzy etc.
The foundation of valence relationship can determine a fuzzy classification, and fuzzy equivalence relation is indicated by fuzzy equivalent matrix;
5th step, supporting vector machine model is established and parameter initialization;
6th step inputs similar day data, Support Vector Machines Optimized parameter;
Trained model is used for electricity consumption short-term forecast by the 7th step.
The present invention chooses similar day by fuzzy clustering method first, then uses these similar days as support vector machines
Training sample, optimizes support vector machines parameter, the model after being trained, for predicting user power consumption;
The present invention can fully excavate the correlation between user's energy, rationally utilize multilist fused data, effectively improve prediction
Precision.The present invention has fully considered the influence factor of various electricity consumptions, effectively improves the precision of prediction, reduces training sample
Quantity has great importance to user power utilization unusual checking.For power grid overall structure, short-term Electric Load Forecasting
The balance that can promote power supply and demand is surveyed, power department is helped rationally to work out production plan, high effective integration resource, improve
Economic benefit makes the operation more safety and stability of power grid.Meanwhile effective load prediction can also help us to use energy to abnormal
Situation makes detection.
As optimization technique measure, the first step,
Choose element of the factor for influencing user power consumption as day character vector, including water consumption, gas consumption, week class
Type, precipitation, air pressure, wind speed, relative humidity, mean temperature, maximum temperature, minimum temperature, fully consider various factors, can
Improve precision of prediction.
As optimization technique measure, day character vector is represented by xi=(x1i, x2i..., xni)T, wherein n is of element
Number;Training data includes history day data and the characteristic vector data on the day of prediction day, and all training datas constitute sample
Set X={ xi, wherein i=1,2 ..., m, m indicate the number of data.
As optimization technique measure, the second step, influence degree of the different factors to electricity consumption in day character vector
It is different, it is therefore desirable to different weights is assigned to these elements;With diagonal matrix A=diag (λ1, λ2..., λn) indicate
Weight matrix, then the day character vector after weighting is
x′i=Axi=(λ1x1i, λ2x2i..., λnxni)T (1)。
As optimization technique measure, for the sample set matrix of aforementioned day character vector composition
Every a line indicates a feature, can regard an independent variable sequence x ask=(xk1, xk2..., xkm), k=1,
2 ... n, the length of sequence are m;Corresponding dependent variable sequence is y=(y1, y2..., ym), then independent variable xkTo dependent variable y's
Direct path coefficient is
Wherein bkFor the partial regression coefficient of multiple regression equation,For independent variable xkStandard deviation, syFor the mark of dependent variable y
It is quasi- poor;Each factor can be found out according to direct path coefficient is to the weighing factor of dependent variable
The present invention determines the weight of element in day character vector using method of path analysis.Path analysis is based on polynary
Linear regression analysis, with the method for correlativity between path coefficient analysis variable, can handle more complicated variable relation.It is logical
Diameter coefficient includes direct path coefficient and indirect path coefficients, reflects independent variable directly affecting and pass through other to dependent variable
The indirect influence of independent variable.Therefore the weight vectors of day character vector element are calculated using direct path coefficient herein.
As optimization technique measure, the third step needs to be compared to different characteristic vector to choose similar day
The similarity between two feature vectors is defined, and then similarity calculated constructs fuzzy similarity matrix and for obscuring
Clustering;To two sampling feature vectors xi、xjBetween similarity be defined as
R(xi, xj)=α D (xi, xj)+βS(xi, xj), (5)
Wherein
D(xi, xj) reflect the distance between feature vector, S (xi, xj) reflect variation tendency between feature vector
Consistency, λkFor the weight of k-th of the element of feature vector found out above, α and β are D (xi, xj) and S (xi, xj) weight, take α
=β=0.5;By calculating any two sampling feature vectors xi、xjBetween similarity R (xi, xj), fuzzy phase can be obtained
Like matrix Rij={ R (xi, xj)}。
As optimization technique measure,
4th step uses fuzzy clustering method to classify to carry out the selection of similar day sample set;It is fuzzy poly-
Alanysis is a kind of mathematical method that things is described and is classified by certain requirement using fuzzy mathematics language;One mould
The foundation of paste equivalence relation can determine a fuzzy classification, and fuzzy equivalence relation is indicated by fuzzy equivalent matrix;It is fuzzy
Similar matrix R and fuzzy equivalent matrix R*It is not necessarily the same;Herein, it can be solved by method gradually square fuzzy
The transitive closure t (R) of similar matrix, obtains R*=t (R).Cluster level (threshold value) T appropriate is chosen at this time to cut R*
It cuts, a corresponding T Level Matrix can be obtainedByIt can determine the classification of a T level.When T changes from 1 to 0
When, classification also changes therewith, forms dynamic clustering.
As optimization technique measure, threshold value T is chosen using F statistic;If r is the value class number of T, miIt is
The number of i class sample is by the average value that original sample matrix X can obtain j-th of feature of all samplesI-th
The average value of j-th of feature of class sample isFollowing F statistic can be then constructed, which obeys freedom degree
F for r-1, m-r is distributed;
Its molecule is able to reflect the distance between class and class, and denominator is able to reflect the distance between sample in class, then F unites
It is better to measure more macrotaxonomy effect;For given confidence alpha, table look-up to obtain critical value Fα, different corresponding F values are faced with this
Dividing value compares, if F > FαThen classifying quality is significant;Meeting F > FαIn the case where, take difference F-FαThe conduct of T is corresponded to when maximum
Optimal choice threshold value;Cutting is carried out to fuzzy equivalent matrix using the threshold value, to extract similar day.
As optimization technique measure, in order to be compared to different factors, they can be normalized to [0,1] section;
Furthermore, it is contemplated that week type itself is a non-quantitation index, need to carry out it quantification treatment, working day, (Monday was extremely
Friday) and day off (Saturday and Sunday) have larger difference with electrical feature;When quantifying value, by day off
Biggish weighted value is assigned, to distinguish with working day.
Compared with prior art, the invention has the following advantages:
The present invention chooses similar day by fuzzy clustering method first, then uses these similar days as support vector machines
Training sample, optimizes support vector machines parameter, the model after being trained, for predicting user power consumption;
The present invention can fully excavate the correlation between user's energy, rationally utilize multilist fused data, effectively improve prediction
Precision.The present invention has fully considered the influence factor of various electricity consumptions, effectively improves the precision of prediction, reduces training sample
Quantity has great importance to user power utilization unusual checking.
Detailed description of the invention
Fig. 1 is that the present invention is based on the similar day prediction techniques of multilist fused data;
Fig. 2 is that the result predicted using the present invention user I is shown;
Fig. 3 is that the present invention predicts that percentage ratio error is shown to user I;
Fig. 4 is that the result predicted using the present invention multi-user is shown;
Fig. 5 is that the present invention predicts that percentage ratio error is shown to multi-user.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
On the contrary, the present invention covers any substitution done on the essence and scope of the present invention being defined by the claims, repairs
Change, equivalent method and scheme.Further, in order to make the public have a better understanding the present invention, below to of the invention thin
It is detailed to describe some specific detail sections in section description.Part without these details for a person skilled in the art
The present invention can be also understood completely in description.
As shown in Figure 1, a kind of short-term electricity demand forecasting method of user based on multilist fused data, first with latus rectum point
Analysis calculates the weight for influencing the day character vector of user power consumption, obtains fuzzy similarity matrix;Then it is passed by fuzzy clustering
It passs Closure and chooses similar day, and using these similar days as sample training support vector machines, user power consumption is predicted;
Specifically includes the following steps:
The first step, the determination and quantization of day character vector choose the factor for influencing user power consumption as day character vector
Element, formed sample set X={ xi};
Second step, path analysis obtain day character vector weight matrix A, and different factors is to electricity consumption in day character vector
Influence degree be different, it is therefore desirable to different weights is assigned to these elements;
Third step calculates the similarity R (x between day character vectori, xj), in order to be compared different characteristic vector to select
Similar day is taken, needs to define the similarity between two feature vectors, and then similarity calculated construction obscures similar square
Battle array is simultaneously used for fuzzy cluster analysis;
4th step, uses fuzzy clustering method to classify sample set to carry out the selection of similar day, and one fuzzy etc.
The foundation of valence relationship can determine a fuzzy classification, and fuzzy equivalence relation is indicated by fuzzy equivalent matrix;
5th step, supporting vector machine model is established and parameter initialization;
6th step inputs similar day data, Support Vector Machines Optimized parameter;
Trained model is used for electricity consumption short-term forecast by the 7th step.
The present invention chooses similar day by fuzzy clustering method first, then uses these similar days as support vector machines
Training sample, optimizes support vector machines parameter, the model after being trained, for predicting user power consumption;
The present invention can fully excavate the correlation between user's energy, rationally utilize multilist fused data, effectively improve prediction
Precision.The present invention has fully considered the influence factor of various electricity consumptions, effectively improves the precision of prediction, reduces training sample
Quantity has great importance to user power utilization unusual checking.
The first step chooses element of the factor for influencing user power consumption as day character vector, including water consumption, use
Tolerance, week type, precipitation, air pressure, wind speed, relative humidity, mean temperature, maximum temperature, minimum temperature fully consider each
Kind factor, can be improved precision of prediction.
Day character vector is represented by xi=(x1i, x2i..., xni)T, wherein n is the number of element;Training data includes
History day the data and characteristic vector data on the day of prediction day, all training datas constitute sample set X={ xi, wherein
I=1,2 ..., m, m indicate the number of data.
In order to be compared to different factors, they can be normalized to [0,1] section;Furthermore, it is contemplated that week class
Type itself is a non-quantitation index, needs to carry out it quantification treatment, working day (Monday to Friday) and day off (star
Phase six and Sunday) have larger difference with electrical feature;When quantifying value, by assigning biggish weighted value to day off,
To be distinguished with working day.
The second step, different factors is different the influence degree of electricity consumption in day character vector, it is therefore desirable to
Different weights is assigned to these elements;With diagonal matrix A=diag (λ1, λ2..., λn) indicate weight matrix, then weighting
Day character vector afterwards is
x′i=Λ xi=(λ1x1i, λ2x2i..., λnxni)T (1)。
For the sample set matrix of aforementioned day character vector composition
Every a line indicates a feature, can regard an independent variable sequence x ask=(xk1, xk2..., xkm), k=1,
2 ... n, the length of sequence are m;Corresponding dependent variable sequence is y=(y1, y2..., ym), then independent variable xkTo dependent variable y's
Direct path coefficient is
Wherein bkFor the partial regression coefficient of multiple regression equation,For independent variable xkStandard deviation, syFor the mark of dependent variable y
It is quasi- poor;Each factor can be found out according to direct path coefficient is to the weighing factor of dependent variable
The present invention determines the weight of element in day character vector using method of path analysis.Path analysis is based on polynary
Linear regression analysis, with the method for correlativity between path coefficient analysis variable, can handle more complicated variable relation.It is logical
Diameter coefficient includes direct path coefficient and indirect path coefficients, reflects independent variable directly affecting and pass through other to dependent variable
The indirect influence of independent variable.Therefore the weight vectors of day character vector element are calculated using direct path coefficient herein.
The third step, in order to be compared different characteristic vector to choose similar day, need to define two features to
Similarity between amount, and then similarity calculated constructs fuzzy similarity matrix and is used for fuzzy cluster analysis;To two
Sampling feature vectors xi、xjBetween similarity be defined as
R(xi, xj)=α D (xi, xj)+βS(xi, xj), (5)
Wherein
D(xi, xj) reflect the distance between feature vector, S (xi, xi) reflect variation tendency between feature vector
Consistency, λkFor the weight of k-th of the element of feature vector found out above, α and β are D (xi, xj) and S (xi, xj) weight, take α
=β=0.5;By calculating any two sampling feature vectors xi、xjBetween similarity R (xi, xj), fuzzy phase can be obtained
Like matrix Rij={ R (xi, xj)}。
4th step uses fuzzy clustering method to classify to carry out the selection of similar day sample set;It is fuzzy poly-
Alanysis is a kind of mathematical method that things is described and is classified by certain requirement using fuzzy mathematics language;One mould
The foundation of paste equivalence relation can determine a fuzzy classification, and fuzzy equivalence relation is indicated by fuzzy equivalent matrix;It is fuzzy
Similar matrix R and fuzzy equivalent matrix R*It is not necessarily the same;Herein, it can be solved by method gradually square fuzzy
The transitive closure t (R) of similar matrix, obtains R*=t (R).Cluster level (threshold value) T appropriate is chosen at this time to R*Cutting is carried out,
It can obtain a corresponding T Level MatrixByIt can determine the classification of a T level.When T changes from 1 to 0, point
Class also changes therewith, forms dynamic clustering.
Threshold value T is chosen using F statistic;If r is the value class number of T, miFor the number of the i-th class sample, by
The average value that original sample matrix X can obtain j-th of feature of all samples isI-th j-th of feature of class sample
Average value beFollowing F statistic can be then constructed, it is r-1 which, which obeys freedom degree, and the F of m-r divides
Cloth;
Its molecule is able to reflect the distance between class and class, and denominator is able to reflect the distance between sample in class, then F unites
It is better to measure more macrotaxonomy effect;For given confidence alpha, table look-up to obtain critical value Fα, different corresponding F values are faced with this
Dividing value compares, if F > FαThen classifying quality is significant;Meeting F > FαIn the case where, take difference F-FαThe conduct of T is corresponded to when maximum
Optimal choice threshold value;Cutting is carried out to fuzzy equivalent matrix using the threshold value, to extract similar day.
In order to verify the performance of electricity demand forecasting method of the present invention, we use the user 2016 of Hangzhou cell 74
The three table fused data of water, electricity, gas in October to December is tested.
Embodiment 1
Certain typical user I is chosen, on December 15,1 day to 2016 October in 2016 totally 76 days electricity consumption data and day
Characteristic vector data is trained as sample data the set pair analysis model;Then, we were to 16 days to 2016 12 December in 2016
Months 25 days totally 10 days electricity consumptions predicted.Firstly, calculating day character vector element weights using path analysis, analysis is different
Influence degree of the factor for electricity consumption.The corresponding weight of each influence factor of user I is as shown in table 1.From experimental result, we
It can be seen that temperature is to influence the main factor of electricity consumption, wherein the weight of mean temperature is maximum.Secondly, we it can also be seen that
Water consumption has a certain impact to electricity consumption tool, therefore it is anticipated that considers that water consumption can improve prediction to a certain extent
Precision.However, correlation is weaker before the two since combustion gas and electric power can be substituted for each other under certain conditions.
Each influence factor weight of 1 user I of table
In order to assess the performance set forth herein method, we are to single table data (electricity consumption data) and multilist fused data
The performance predicted in the case of two kinds is compared.Fig. 2 gives and predicts electricity consumption under the practical electricity consumption of user I, single table data
With the curve for predicting electricity consumption under multilist data.From experimental result, we can be found that prediction result and reality under multilist data
The electricity consumption on border is closer to, hence it is evident that better than the prediction result of single table data.It is biggish especially for electricity consumption fluctuating change
On the date (such as December 19), single table prediction result is it is possible that very big deviation, and the support vector machine method based on multilist
Due to comprehensively considering water consumption and gas consumption information, electricity consumption variation tendency can be preferably predicted.In addition, Fig. 3 is shown
The relative percentage of prediction 10 days predicts error curve.From experimental result it was found that the consensus forecast under single table data misses
Difference is 8.64%, and the mean error under multilist data is 1.99%, is reduced compared to the mean error under single table data
6% or more.
Embodiment 2
We can not only predict the electricity consumption of single user that electricity consumption that can also be total to multiple users carries out
Prediction.The corresponding weight of each influence factor of total electricity consumption is as shown in table 2, it is clear that and temperature is still most important influence factor,
Wherein the weight of mean temperature is maximum.Compared to single user, since intra-cell users total electricity consumption maintains one relatively steadily
State (referring to fig. 4), water consumption is relatively small to the influence degree of gas consumption.
Each influence factor weight of certain cell multi-user of table 2
Fig. 4 is the prediction result curve to above-mentioned 74 user's total electricity consumptions, and Fig. 5 shows that daily relative percentage is pre-
Survey error curve.It can be seen that prediction result and the variation tendency of practical electricity consumption are almost the same, prediction error is smaller, single table number
Mean error under is 1.87%, and the average forecasting error under multilist data is 0.63%, reduces 1% than single table data
Left and right.
The present invention is directed to the demand of power consumer electricity consumption short-term forecast, has used the prediction technique based on similar day, leads to
It crosses building day character vector, carry out fuzzy cluster analysis selection similar day, then use these similar days as training data training
Supporting vector machine model is predicted in real data using the model later.This method has fully considered various electricity consumptions
The influence factor of amount effectively improves the precision of prediction, reduces training samples number, has to user power utilization unusual checking
Important meaning.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.
Claims (9)
1. a kind of short-term electricity demand forecasting method of user based on multilist fused data, which is characterized in that first with latus rectum point
Analysis calculates the weight for influencing the day character vector of user power consumption, obtains fuzzy similarity matrix;Then it is passed by fuzzy clustering
It passs Closure and chooses similar day, and using these similar days as sample training support vector machines, user power consumption is predicted;
Specifically includes the following steps:
Member of the factor for influencing user power consumption as day character vector is chosen in the first step, the determination and quantization of day character vector
Element forms sample set X={ xi};
Second step, path analysis obtain day character vector weight matrix A, shadow of the different factors to electricity consumption in day character vector
The degree of sound is different, it is therefore desirable to different weights is assigned to these elements;
Third step calculates the similarity R (x between day character vectori, xj), in order to be compared different characteristic vector to choose phase
Like day, need to define the similarity between two feature vectors, and then similarity calculated construction fuzzy similarity matrix is simultaneously
For fuzzy cluster analysis;
4th step uses fuzzy clustering method to classify sample set to carry out the selection of similar day, and a fuzzy equivalence closes
The foundation of system can determine a fuzzy classification, and fuzzy equivalence relation is indicated by fuzzy equivalent matrix;
5th step, supporting vector machine model is established and parameter initialization;
6th step inputs similar day data, Support Vector Machines Optimized parameter;
Trained model is used for electricity consumption short-term forecast by the 7th step.
2. a kind of short-term electricity demand forecasting method of user based on multilist fused data as described in claim 1, feature exist
In, the first step,
Choose element of the factor as day character vector for influencing user power consumption, including water consumption, gas consumption, week type,
Precipitation, air pressure, wind speed, relative humidity, mean temperature, maximum temperature, minimum temperature.
3. a kind of short-term electricity demand forecasting method of user based on multilist fused data as claimed in claim 2, feature exist
In day character vector is represented by xi=(x1i, x2i..., xni)T, wherein n is the number of element;Training data includes history day
Data and the characteristic vector data on the day of prediction day, all training datas constitute sample set X={ xi, wherein i=1,
The number of 2 ..., m, m expression data.
4. a kind of short-term electricity demand forecasting method of user based on multilist fused data as claimed in claim 3, feature exist
Different factors is different the influence degree of electricity consumption in, the second step, day character vector, it is therefore desirable to give these
Element assigns different weights;With diagonal matrix A=diag (λ1, λ2..., λn) indicate weight matrix, then the day after weighting
Feature vector is
x′i=Axi=(λ1x1i, λ2x2i..., λnxni)T (1)。
5. a kind of short-term electricity demand forecasting method of user based on multilist fused data as claimed in claim 4, feature exist
In for the sample set matrix of aforementioned day character vector composition
Every a line indicates a feature, can regard an independent variable sequence x ask=(xk1, xk2..., xkm), k=1,2 ... n,
The length of sequence is m;Corresponding dependent variable sequence is y=(y1, y2..., ym), then independent variable xkTo the directly logical of dependent variable y
Diameter coefficient is
Wherein bkFor the partial regression coefficient of multiple regression equation,For independent variable xkStandard deviation, syFor the standard deviation of dependent variable y;
Each factor can be found out according to direct path coefficient is to the weighing factor of dependent variable
6. a kind of short-term electricity demand forecasting method of user based on multilist fused data as claimed in claim 5, feature exist
In the third step needs to define between two feature vectors to be compared to choose similar day different characteristic vector
Similarity, and then similarity calculated construction fuzzy similarity matrix and be used for fuzzy cluster analysis;To two sample spies
Levy vector xi、xjBetween similarity be defined as
R(xi, xj)=α D (xi, xj)+βS(xi, xj), (5)
Wherein
D(xi, xj) reflect the distance between feature vector, S (xi, xj) reflect the consistent of variation tendency between feature vector
Property, λkFor the weight of k-th of the element of feature vector found out above, α and β are D (xi, xj) and S (xi, xj) weight, take α=β
=0.5;By calculating any two sampling feature vectors xi、xjBetween similarity R (xi, xj), it can obtain and obscure similar square
Battle array Rij={ R (xi, xj)}。
7. a kind of short-term electricity demand forecasting method of user based on multilist fused data as claimed in claim 6, feature exist
In,
4th step uses fuzzy clustering method to classify to carry out the selection of similar day sample set;Fuzzy clustering point
Analysis is a kind of mathematical method that things is described and is classified by certain requirement using fuzzy mathematics language;One fuzzy etc.
The foundation of valence relationship can determine a fuzzy classification, and fuzzy equivalence relation is indicated by fuzzy equivalent matrix;It obscures similar
Matrix R and fuzzy equivalent matrix R*It is not necessarily the same;Herein, it can be obscured by method solution gradually square similar
The transitive closure t (R) of matrix, obtains R*=t (R);The horizontal T of cluster appropriate is chosen at this time to R*Cutting is carried out, can be obtained
One corresponding T Level MatrixByIt can determine the classification of a T level;When f changes from 1 to 0, classification also becomes therewith
Change, forms dynamic clustering.
8. a kind of short-term electricity demand forecasting method of user based on multilist fused data as claimed in claim 7, feature exist
In being chosen using F statistic to threshold value T;If r is the value class number of T, miFor the number of the i-th class sample, by original sample
The average value that this matrix X can obtain j-th of feature of all samples isI-th j-th of feature of class sample is averaged
Value isFollowing F statistic can be then constructed, it is r-1 which, which obeys freedom degree, and the F of m-r is distributed;
Its molecule is able to reflect the distance between class and class, and denominator is able to reflect the distance between sample in class, then F statistic
More macrotaxonomy effect is better;For given confidence alpha, table look-up to obtain critical value Fα, by different corresponding F values and the critical value
Compare, if F > FαThen classifying quality is significant;Meeting F > FαIn the case where, take difference F-FαThe conduct that T is corresponded to when maximum is best
Selected threshold;Cutting is carried out to fuzzy equivalent matrix using the threshold value, to extract similar day.
9. a kind of short-term electricity demand forecasting method of user based on multilist fused data a method as claimed in any one of claims 1-8,
It is characterized in that, in order to be compared to different factors, they can be normalized to [0,1] section;In addition, working day and rest
There is larger difference in day with electrical feature;When quantifying value, by assigning biggish weighted value to day off, thus and working day
It distinguishes.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910029664.3A CN109829571A (en) | 2019-01-11 | 2019-01-11 | A kind of short-term electricity demand forecasting method of user based on multilist fused data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910029664.3A CN109829571A (en) | 2019-01-11 | 2019-01-11 | A kind of short-term electricity demand forecasting method of user based on multilist fused data |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109829571A true CN109829571A (en) | 2019-05-31 |
Family
ID=66861669
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910029664.3A Pending CN109829571A (en) | 2019-01-11 | 2019-01-11 | A kind of short-term electricity demand forecasting method of user based on multilist fused data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109829571A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110222180A (en) * | 2019-06-04 | 2019-09-10 | 江南大学 | A kind of classification of text data and information mining method |
CN110781494A (en) * | 2019-10-22 | 2020-02-11 | 武汉极意网络科技有限公司 | Data abnormity early warning method, device, equipment and storage medium |
CN111881190A (en) * | 2020-08-05 | 2020-11-03 | 厦门力含信息技术服务有限公司 | Key data mining system based on customer portrait |
CN113449793A (en) * | 2021-06-28 | 2021-09-28 | 国网北京市电力公司 | Method and device for determining power utilization state |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106251001A (en) * | 2016-07-18 | 2016-12-21 | 南京工程学院 | A kind of based on the photovoltaic power Forecasting Methodology improving fuzzy clustering algorithm |
CN108171369A (en) * | 2017-12-21 | 2018-06-15 | 国家电网公司 | Short term combination forecasting method based on customer electricity differentiation characteristic |
-
2019
- 2019-01-11 CN CN201910029664.3A patent/CN109829571A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106251001A (en) * | 2016-07-18 | 2016-12-21 | 南京工程学院 | A kind of based on the photovoltaic power Forecasting Methodology improving fuzzy clustering algorithm |
CN108171369A (en) * | 2017-12-21 | 2018-06-15 | 国家电网公司 | Short term combination forecasting method based on customer electricity differentiation characteristic |
Non-Patent Citations (1)
Title |
---|
郑国和 等: "基于多表融合数据的用户短期用电量预测", 《电力系统及其自动化学报》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110222180A (en) * | 2019-06-04 | 2019-09-10 | 江南大学 | A kind of classification of text data and information mining method |
CN110781494A (en) * | 2019-10-22 | 2020-02-11 | 武汉极意网络科技有限公司 | Data abnormity early warning method, device, equipment and storage medium |
CN111881190A (en) * | 2020-08-05 | 2020-11-03 | 厦门力含信息技术服务有限公司 | Key data mining system based on customer portrait |
CN113449793A (en) * | 2021-06-28 | 2021-09-28 | 国网北京市电力公司 | Method and device for determining power utilization state |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109829571A (en) | A kind of short-term electricity demand forecasting method of user based on multilist fused data | |
Nahmmacher et al. | Carpe diem: A novel approach to select representative days for long-term power system modeling | |
Park et al. | Apples or oranges? Identification of fundamental load shape profiles for benchmarking buildings using a large and diverse dataset | |
Yildiz et al. | A review and analysis of regression and machine learning models on commercial building electricity load forecasting | |
Xie et al. | A new method for crude oil price forecasting based on support vector machines | |
Yu et al. | A decision tree method for building energy demand modeling | |
Huang et al. | Data mining for electricity price classification and the application to demand-side management | |
Dunford et al. | Exploring climate change vulnerability across sectors and scenarios using indicators of impacts and coping capacity | |
Huang et al. | A clustering based grouping method of nearly zero energy buildings for performance improvements | |
CN110109971A (en) | A kind of low-voltage platform area user power utilization Load Characteristic Analysis method | |
CN110930198A (en) | Electric energy substitution potential prediction method and system based on random forest, storage medium and computer equipment | |
CN109634942A (en) | A kind of energy data exception judgment method and device | |
Nwulu et al. | Modelling and predicting electricity consumption using artificial neural networks | |
CN109409561A (en) | The construction method of Multiple Time Scales time series collaborative forecasting model | |
CN112330077B (en) | Power load prediction method, power load prediction device, computer equipment and storage medium | |
Muzumdar et al. | Designing a robust and accurate model for consumer-centric short-term load forecasting in microgrid environment | |
CN112149890A (en) | Comprehensive energy load prediction method and system based on user energy label | |
Raghavendra et al. | Artificial humming bird with data science enabled stability prediction model for smart grids | |
Kaur et al. | A hybrid energy management approach for home appliances using climatic forecasting | |
Omar et al. | Seasonal clustering forecasting technique for intelligent hourly solar irradiance systems | |
CN117332288A (en) | Building energy saving technology evaluation system and method | |
Liu et al. | An inexact stochastic–fuzzy jointed chance-constrained programming for regional energy system management under uncertainty | |
Sayegh | Holistic optimization of buildings based on the evaluation of annual performances from short simulation sequences | |
CN116308456A (en) | User electricity consumption behavior analysis method, system, device and medium in big data environment | |
CN114707422A (en) | Intelligent power check method based on load prediction |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190531 |
|
RJ01 | Rejection of invention patent application after publication |