CN109508820A - Campus electricity demand forecasting modeling method based on differentiation modeling - Google Patents
Campus electricity demand forecasting modeling method based on differentiation modeling Download PDFInfo
- Publication number
- CN109508820A CN109508820A CN201811296952.7A CN201811296952A CN109508820A CN 109508820 A CN109508820 A CN 109508820A CN 201811296952 A CN201811296952 A CN 201811296952A CN 109508820 A CN109508820 A CN 109508820A
- Authority
- CN
- China
- Prior art keywords
- data
- differentiation
- distance
- point
- modeling
- 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
- 230000005611 electricity Effects 0.000 title claims abstract description 73
- 238000000034 method Methods 0.000 title claims abstract description 50
- 230000004069 differentiation Effects 0.000 title claims abstract description 49
- 239000013598 vector Substances 0.000 claims description 20
- 238000013480 data collection Methods 0.000 claims description 6
- 230000002950 deficient Effects 0.000 claims description 5
- JEIPFZHSYJVQDO-UHFFFAOYSA-N ferric oxide Chemical compound O=[Fe]O[Fe]=O JEIPFZHSYJVQDO-UHFFFAOYSA-N 0.000 claims description 5
- 230000003321 amplification Effects 0.000 claims description 3
- 230000006399 behavior Effects 0.000 claims description 3
- 238000001514 detection method Methods 0.000 claims description 3
- 238000000556 factor analysis Methods 0.000 claims description 3
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 230000002093 peripheral effect Effects 0.000 claims 1
- 230000010287 polarization Effects 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- 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)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Tourism & Hospitality (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of campus electricity demand forecasting modeling method based on differentiation modeling, this method using campus weather conditions, whether festivals or holidays and history electricity consumption are as raw data set;Bad data recognition is carried out to raw data set using the method based on FCM and differentiation distance, training set data library is determined, for the interference for avoiding noise data, classifies again by FCM to training set, generate sub- training set;Differentiation modeling is carried out to electricity demand forecasting using Lazy Learning algorithm.This method passes through to bad data recognition in practical electricity consumption data and true fitting, reasonably electricity consumption is predicted, it can be according to different buildings, the different dates to be measured establish different electricity demand forecasting models, example shows that this method can carry out rationally accurately prediction to campus building electricity consumption, and is conducive to power department and obtains clearly electricity consumption pattern information, reasonable arrangement dispatches the electricity consumption in campus difference building.
Description
Technical field
The present invention relates to a kind of campus electricity demand forecasting modeling methods based on differentiation modeling.
Background technique
Accurate electricity demand forecasting is the important component of Energy Management System, has reasonably accurate predicted electricity consumption situation both
It runs with being conducive to electric system economic and reliable, and is beneficial to making rational planning for for campus power grid.Difference is carried out for electrical feature
Change electricity demand forecasting modeling, can help to better meet the power demand under different situations, be conducive to the optimal scheduling of electric power.
However, since the capacity of electricity consumption data under smart grid background is big, the features such as complexity height and formation speed are fast, so that traditional
Prediction technique can not effectively excavate the validated user electricity consumption pattern information under the conditions of different times, different work.How big
It is directed under data environment and does not have to user power utilization mode to carry out differentiation modeling to be a challenge for researcher.
Summary of the invention
Technical problem to be solved by the invention is to provide a kind of campus electricity demand forecasting modelings based on differentiation modeling
Method, this method establish meter and weather, festivals or holidays spy to bad data recognition in the practical electricity consumption data in campus and true fitting
The campus differentiation electricity demand forecasting model of sign is conducive to carry out rationally accurately prediction to each building electricity consumption in campus
Power department obtains clearly electricity consumption pattern information, and reasonable arrangement dispatches the electricity consumption in each building in campus.
In order to solve the above technical problems, including such as the present invention is based on the campus electricity demand forecasting modeling method of differentiation modeling
Lower step:
Step 1: obtain the weather conditions in campus location to be measured, by school calendar obtain whether festivals or holidays, pass through campus electricity
Net obtains history electricity consumption as raw data set;
Step 2: being quantified weather conditions using factor analysis, used when to build model as parameter;
Step 3: raw data set exist due to equipment record or electricity consumption behavior change suddenly caused by umber of defectives
According to, using based on FCM clustering algorithm and differentiation distance method to raw data set carry out bad data recognition, screen out isolated data
And bad data, determine training set data library;
Step 4: selecting modeling point advantageously to reduce the interference of noise data, FCM clustering algorithm is used again
Classify to training set, generates sub- training set;
Step 5: being concentrated through k-vnn using Lazy Learning algorithm in son training and calculating data point in training set
With the Euclidean distance between point to be predicted, the nearest k number strong point of selected distance passes through Regression Model Simulator k as modeling point
A modeling point establishes prediction modeling to campus electricity consumption, predicts the electricity consumption of point to be predicted.
Further, the FCM clustering algorithm handles database data, according to weather conditions, whether festivals or holidays, most
Four high-temperature, minimum temperature features obtain electricity consumption classification;Break up distance method to the electricity consumption obtained through FCM clustering algorithm
Classified calculating calculates separately middle any two data point x of all categoriesiAnd xjBetween practical Euclidean distance D (i, j) and it is maximum away from
From distancemax, the numerical value in one [0,1] is chosen as degree of differentiation μ, and differentiation calculating is carried out to whole D (i, j), that is, is amplified
The distance between data object, x after being handlediAnd xjBetween differentiation distance be r (i, j), mathematic(al) representation are as follows:
Compare D (i, j) and distancemaxRatio and compared with μ, if ratio be less than μ, reduce ratio and to break up
Distance is less than actual range;Conversely, amplification differentiation distance;The difference of ratio and μ become with differentiation distance in the inverse ratio that slope increases
Gesture, i.e. difference are bigger, break up apart from smaller;Difference is smaller, and differentiation distance is bigger;It is handled by differentiation, is associated in cluster
Property big data it is even closer, while separating, there are the data of otherness to reach the two poles of the earth so that outlier is farther apart from normal value
The purpose of differentiation.
Further, the bad data recognition assumes the data set mass center handled through FCM clustering algorithm to all data points
The average distance of distance is Adistance, the size that R represents data object peripheral extent is defined, i.e. average distance accounts for maximum distance
Ratio by degree of differentiation treated numerical value,
Then
The number of other data points within the scope of the neighbor density of data collection i.e. object R is set when neighbours' point number is less than
Fixed number mesh Knum, then it is assumed that it is Outlier Data, i.e. umber of defectives strong point.
Further, according to formula (1) and formula (2), compare the relationship of r (i, j) and R, judge the number of object neighbours' point, calculate
Steps are as follows for method:
1) input cluster the data obtained collection, degree of differentiation μ, minimum neighbours' number Knum;
2) it calculates and clusters middle mass center and A of all categoriesdistance;
3) data set distance R is calculated1With data object xiDifferentiation distance r1(i,j)。
R1=μ Adistance(3)
4) compare r1(i, j) and R1Relationship, if r1(i, j) < R1, then xiNeighbours' point number K add 1, once K > Knum,
The K for interrupting the data object into next data object is calculated, and otherwise traverses all neighbours' point numbers;
5) data object for being unsatisfactory for K > Knum is rejected as bad data.
Further, the campus electricity demand forecasting model carries out differentiation modeling for different tested points, by every
One sampled point all selects most matched data vector to establish estimation model from historical data base, and each model is only to current
Sampled point is effective, and differentiation modeling specifically comprises the following steps:
1. K-VNN reconnaissance, in K-VNN method, the importation of each point is seen as a vector, by considering simultaneously
The distance between two input vectors and angle judge whether two vectors are neighbour;
Assuming that the input vector of a data point in database is xp=[xp1,xp2,…,xpr]T, current sampling point it is defeated
Incoming vector is x (t)=[x1,x2,…,xr]T, wherein r is the dimension of sampled point importation, then Euclidean distance between the two
It is defined as follows with angle:
Select the process of K Neighbor Points as follows from database using K-VNN method:
(1) as cos (xp, x (t)) < 0, xpIn the opposite direction with x (t), x is abandonedp;
(2) as cos (xp, x (t)) >=0, Euclidean distance and angle are considered using following formula simultaneously;
Wherein, α is weight, 0≤α≤1, D (xp, x (t)) value and two vectors similarity it is proportional, by D (xp, x
(t)) in magnitude order, preceding K data point is selected for constructing partial model modeling data collection { (x at this time1,y1),…,
(xk,yk)};
2. the foundation of prediction model
After K data point is chosen, establishes first-order linear and return multinomial to represent the regression model of current data point, i.e.,
Model of the sampled point output par, c relative to importation is established,
Wherein: β=[β0,β1,…,βr] it is model parameter, k=1,2 ..., K,
Using the weighted sum of squares of reality output and model output residual error as the interpretational criteria of model accuracy,
Model parameter can be obtained by minimizing JThen the model output of current point x (t) can indicate are as follows:
Must arrive school the prediction model of electricity consumption.
Further, by data detection, precision of prediction with higher and speed when 5 are chosen in k number strong point.
Since the campus electricity demand forecasting modeling method modeled the present invention is based on differentiation uses above-mentioned technical proposal, i.e.,
This method using campus weather conditions, whether festivals or holidays and campus power grid history electricity consumption are as raw data set;Day is vaporous
Condition quantization, uses when to build model as parameter;Raw data set is carried out using the method based on FCM and differentiation distance
Bad data recognition screens out isolated data and bad data, determines training set data library, for the interference for avoiding noise data, choosing
Modeling point advantageously is selected, classifies again by FCM to training set, generates sub- training set;Utilize Lazy Learning
Algorithm carries out differentiation modeling to the prediction of electricity consumption.This method passes through to bad data recognition in practical electricity consumption data and really
Fitting, reasonably predicts electricity consumption, establishes the campus otherness electricity demand forecasting mould of meter and weather, festivals or holidays feature
Type;This method, can be according to different buildings as a kind of method for predicting campus electricity consumption, and the different dates to be measured establish different use
Power quantity predicting model, example shows that this method can carry out rationally accurately prediction to campus building electricity consumption, and is conducive to
Power department obtains clearly electricity consumption pattern information, and reasonable arrangement dispatches the electricity consumption in campus difference building.
Detailed description of the invention
The present invention will be further described in detail below with reference to the accompanying drawings and embodiments:
Fig. 1 is that the present invention is based on the campus electricity demand forecasting modeling method flow diagrams that differentiation models;
Fig. 2 is the result schematic diagram that bad data recognition is carried out using real data;
Fig. 3 is that this method is broken up by FCM- apart from combination bad data recognition flow diagram figure;
Fig. 4 is the prediction result figure based on Lazy Learning algorithm.
Specific embodiment
As shown in Figure 1, the present invention is based on the campus electricity demand forecasting modeling methods of differentiation modeling to include the following steps:
Step 1: obtain the weather conditions in campus location to be measured, by school calendar obtain whether festivals or holidays, pass through campus electricity
Net obtains history electricity consumption as raw data set, wherein it is to be characterized value 0 that whether festivals or holidays, which use, it is no to be characterized 1 table of value
Show;
Step 2: being quantified weather conditions using factor analysis, used when to build model as parameter;Weather is special
Value indicative is as shown in table 1;
1 weather characteristics value numerical tabular of table
Step 3: raw data set exist due to equipment record or electricity consumption behavior change suddenly caused by umber of defectives
According to, using based on FCM clustering algorithm and differentiation distance method to raw data set carry out bad data recognition, screen out isolated data
And bad data, determine training set data library;
Step 4: selecting modeling point advantageously to reduce the interference of noise data, FCM clustering algorithm is used again
Classify to training set, generates sub- training set;
Step 5: being concentrated through k-vnn using Lazy Learning algorithm in son training and calculating data point in training set
With the Euclidean distance between point to be predicted, the nearest k number strong point of selected distance passes through Regression Model Simulator k as modeling point
A modeling point establishes prediction modeling to campus electricity consumption, predicts the electricity consumption of point to be predicted.LazyLearning algorithm is a kind of
It can be a kind of flexible modeling method that essence is adaptive based on the algorithm of " similar input generates similar output " principle.
Preferably, consider bad data adverse effect caused by predicted distortion, the FCM clustering algorithm is to database number
According to being handled, according to feature weather conditions, whether four festivals or holidays, maximum temperature, minimum temperature acquisition electricity consumptions classification;Point
Change distance method to the electricity consumption classified calculating obtained through FCM clustering algorithm, calculates separately middle any two data point of all categories
xiAnd xjBetween practical Euclidean distance D (i, j) and maximum distance distancemax, choose the numerical value conduct in one [0,1]
Degree of differentiation μ carries out differentiation calculating, i.e. the distance between amplification data object, x after being handled to whole D (i, j)iAnd xjBetween
Differentiation distance be r (i, j), mathematic(al) representation are as follows:
Compare D (i, j) and distancemaxRatio and compared with μ, if ratio be less than μ, reduce ratio and to break up
Distance is less than actual range;Conversely, amplification differentiation distance;The difference of ratio and μ become with differentiation distance in the inverse ratio that slope increases
Gesture, i.e. difference are bigger, break up apart from smaller;Difference is smaller, and differentiation distance is bigger;It is handled by differentiation, is associated in cluster
Property big data it is even closer, while separating, there are the data of otherness to reach the two poles of the earth so that outlier is farther apart from normal value
The purpose of differentiation.
Preferably, the bad data recognition assumes the data set mass center handled through FCM clustering algorithm to all data points
The average distance of distance is Adistance, the size that R represents data object peripheral extent is defined, i.e. average distance accounts for maximum distance
Ratio by degree of differentiation treated numerical value,
Then
The number of other data points within the scope of the neighbor density of data collection i.e. object R is set when neighbours' point number is less than
Fixed number mesh Knum, then it is assumed that it is Outlier Data, i.e. umber of defectives strong point.
Preferably, according to formula (1) and formula (2), compare the relationship of r (i, j) and R, judge the number of object neighbours' point, calculate
Steps are as follows for method:
1) input cluster the data obtained collection, degree of differentiation μ, minimum neighbours' number Knum;
2) it calculates and clusters middle mass center and A of all categoriesdistance;
3) data set distance R is calculated1With data object xiDifferentiation distance r1(i,j)。
R1=μ Adistance(3)
4) compare r1(i, j) and R1Relationship, if r1(i, j) < R1, then xiNeighbours' point number K add 1, once K > Knum,
The K for interrupting the data object into next data object is calculated, and otherwise traverses all neighbours' point numbers;
5) data object for being unsatisfactory for K > Knum is rejected as bad data.
As shown in Fig. 2, carrying out bad data recognition using the practical electricity consumption data of certain colleges and universities, the bad data picked out will
It can reject, then be predicted from data set.
As shown in Figure 3, it is preferred that the campus electricity demand forecasting model carries out differentiation for different tested points and builds
Mould, by selecting most matched data vector to establish estimation model, Mei Gemo from historical data base each sampled point
Type is only effective to current sampled point, and differentiation modeling specifically comprises the following steps:
1. K-VNN reconnaissance, in K-VNN method, the importation of each point is seen as a vector, by considering simultaneously
The distance between two input vectors and angle judge whether two vectors are neighbour;
Assuming that the input vector of a data point in database is xp=[xp1,xp2,…,xpr]T, current sampling point it is defeated
Incoming vector is x (t)=[x1,x2,…,xr]T, wherein r is the dimension of sampled point importation, then Euclidean distance between the two
It is defined as follows with angle:
Select the process of K Neighbor Points as follows from database using K-VNN method:
(1) as cos (xp, x (t)) < 0, xpIn the opposite direction with x (t), x is abandonedp;
(2) as cos (xp, x (t)) >=0, Euclidean distance and angle are considered using following formula simultaneously;
Wherein, α is weight, 0≤α≤1, D (xp, x (t)) value and two vectors similarity it is proportional, by D (xp, x
(t)) in magnitude order, preceding K data point is selected for constructing partial model modeling data collection { (x at this time1,y1),…,
(xk,yk)};
2. the foundation of prediction model
After K data point is chosen, establishes first-order linear and return multinomial to represent the regression model of current data point, i.e.,
Model of the sampled point output par, c relative to importation is established,
Wherein: β=[β0,β1,…,βr] it is model parameter, k=1,2 ..., K,
Using the weighted sum of squares of reality output and model output residual error as the interpretational criteria of model accuracy,
Model parameter can be obtained by minimizing JThen the model output of current point x (t) can indicate are as follows:
Must arrive school the prediction model of electricity consumption.
Preferably.By data detection, precision of prediction with higher and speed when k number strong point chooses 5.
As shown in figure 4, to carry out the matched curve of electricity demand forecasting using certain colleges and universities' real data.
This method carries out electricity consumption curve matching using campus actual measurement electricity consumption data and corresponding weather data etc., utilizes FCM
Bad data is recognized apart from combined method with differentiation;Differentiation modeling is carried out with power mode to different, is made an uproar first with FCM rejecting
Sound data reduce training set, then establish electricity demand forecasting model using Lazy learning algorithm.
Claims (6)
1. a kind of campus electricity demand forecasting modeling method based on differentiation modeling, it is characterised in that this method includes following step
It is rapid:
Step 1: obtain the weather conditions in campus location to be measured, by school calendar obtain whether festivals or holidays, obtained by campus power grid
Take history electricity consumption as raw data set;
Step 2: being quantified weather conditions using factor analysis, used when to build model as parameter;
Step 3: raw data set exist due to equipment record or electricity consumption behavior change suddenly caused by bad data, adopt
Bad data recognition is carried out to raw data set with based on FCM clustering algorithm and differentiation distance method, screens out isolated data and not
Good data determine training set data library;
Step 4: modeling point advantageously is selected to reduce the interference of noise data, again using FCM clustering algorithm to instruction
Practice collection to classify, generates sub- training set;
Step 5: using Lazy Learning algorithm, son training be concentrated through k-vnn calculate in training set data point with to
Euclidean distance between future position, the nearest k number strong point of selected distance are built as modeling point by Regression Model Simulator k
Mould point establishes prediction modeling to campus electricity consumption, predicts the electricity consumption of point to be predicted.
2. the campus electricity demand forecasting modeling method according to claim 1 based on differentiation modeling, it is characterised in that: institute
FCM clustering algorithm is stated to handle database data, according to weather conditions, whether festivals or holidays, maximum temperature, minimum temperature four
A feature obtains electricity consumption classification;Break up distance method to the electricity consumption classified calculating obtained through FCM clustering algorithm, counts respectively
Calculate middle any two data point x of all categoriesiAnd xjBetween practical Euclidean distance D (i, j) and maximum distance distancemax, choosing
It takes the numerical value in one [0,1] as degree of differentiation μ, differentiation calculating is carried out to whole D (i, j), i.e., between amplification data object
Distance, x after being handlediAnd xjBetween differentiation distance be r (i, j), mathematic(al) representation are as follows:
Compare D (i, j) and distancemaxRatio and compared with μ, if ratio be less than μ, reduce ratio make break up distance
Less than actual range;Conversely, amplification differentiation distance;The inverse ratio trend that the difference of ratio and μ and differentiation distance increase in slope, i.e.,
Difference is bigger, breaks up apart from smaller;Difference is smaller, and differentiation distance is bigger;It is handled by differentiation, relevance is big in cluster
Data are even closer, while separating, there are the data of otherness to reach polarization so that outlier is farther apart from normal value
Purpose.
3. the campus electricity demand forecasting modeling method according to claim 1 or 2 based on differentiation modeling, feature exist
In: the bad data recognition assumes the data set mass center handled through FCM clustering algorithm to the average departure of all data point distances
From for Adistance, the size that R represents data object peripheral extent is defined, i.e. average distance accounts for the ratio of maximum distance by differentiation
Degree treated numerical value,
Then
The number of other data points within the scope of the neighbor density of data collection i.e. object R, when neighbours' point number is less than setting number
Mesh Knum, then it is assumed that it is Outlier Data, i.e. umber of defectives strong point.
4. the campus electricity demand forecasting modeling method according to claim 3 based on differentiation modeling, it is characterised in that: root
According to formula (1) and formula (2), compare the relationship of r (i, j) and R, judge the number of object neighbours' point, algorithm steps are as follows:
1) input cluster the data obtained collection, degree of differentiation μ, minimum neighbours' number Knum;
2) it calculates and clusters middle mass center and A of all categoriesdistance;
3) data set distance R is calculated1With data object xiDifferentiation distance r1(i,j)。
R1=μ Adistance (3)
4) compare r1(i, j) and R1Relationship, if r1(i, j) < R1, then xiNeighbours' point number K add 1, once K > Knum, interrupt
The K that the data object enters next data object is calculated, and otherwise traverses all neighbours' point numbers;
5) data object for being unsatisfactory for K > Knum is rejected as bad data.
5. the campus electricity demand forecasting modeling method according to claim 3 based on differentiation modeling, it is characterised in that: institute
State campus electricity demand forecasting model and carry out differentiation modeling for different tested point, by each sampled point from history
Most matched data vector is selected to establish estimation model, each model, differentiation effective to current sampled point in database
Modeling specifically comprises the following steps:
1. K-VNN reconnaissance, in K-VNN method, the importation of each point is seen as a vector, by considering two simultaneously
The distance between input vector and angle judge whether two vectors are neighbour;
Assuming that the input vector of a data point in database is xp=[xp1,xp2,…,xpr]T, the input of current sampling point to
Amount is x (t)=[x1,x2,…,xr]T, wherein r is the dimension of sampled point importation, then Euclidean distance and angle between the two
Degree is defined as follows:
Select the process of K Neighbor Points as follows from database using K-VNN method:
(1) as cos (xp, x (t)) < 0, xpIn the opposite direction with x (t), x is abandonedp;
(2) as cos (xp, x (t)) >=0, Euclidean distance and angle are considered using following formula simultaneously;
Wherein, α is weight, 0≤α≤1, D (xp, x (t)) value and two vectors similarity it is proportional, by D (xp, x (t)) and it presses
It sorts according to size, preceding K data point is selected for constructing partial model modeling data collection { (x at this time1,y1),…,(xk,
yk)};
2. the foundation of prediction model
After K data point is chosen, establishes first-order linear and return multinomial to represent the regression model of current data point, that is, establish
Model of the sampled point output par, c relative to importation,
Wherein: β=[β0,β1,…,βr] it is model parameter, k=1,2 ..., K,
Using the weighted sum of squares of reality output and model output residual error as the interpretational criteria of model accuracy,
Model parameter can be obtained by minimizing JThen the model output of current point x (t) can indicate are as follows:
Must arrive school the prediction model of electricity consumption.
6. the campus electricity demand forecasting modeling method according to claim 5 based on differentiation modeling, it is characterised in that: warp
Cross data detection, precision of prediction with higher and speed when 5 are chosen in k number strong point.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811296952.7A CN109508820A (en) | 2018-11-01 | 2018-11-01 | Campus electricity demand forecasting modeling method based on differentiation modeling |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811296952.7A CN109508820A (en) | 2018-11-01 | 2018-11-01 | Campus electricity demand forecasting modeling method based on differentiation modeling |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109508820A true CN109508820A (en) | 2019-03-22 |
Family
ID=65747381
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811296952.7A Pending CN109508820A (en) | 2018-11-01 | 2018-11-01 | Campus electricity demand forecasting modeling method based on differentiation modeling |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109508820A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110084158A (en) * | 2019-04-15 | 2019-08-02 | 杭州拓深科技有限公司 | A kind of electrical equipment recognition methods based on intelligent algorithm |
CN110689250A (en) * | 2019-09-20 | 2020-01-14 | 深圳供电局有限公司 | Method and system for processing user electricity consumption data and computer readable medium |
CN114553538A (en) * | 2022-02-22 | 2022-05-27 | 国网山东省电力公司电力科学研究院 | Active protection method and system for power grid information safety |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106779219A (en) * | 2016-12-19 | 2017-05-31 | 北京中电普华信息技术有限公司 | A kind of electricity demand forecasting method and system |
CN108171379A (en) * | 2017-12-28 | 2018-06-15 | 无锡英臻科技有限公司 | A kind of electro-load forecast method |
CN108510006A (en) * | 2018-04-08 | 2018-09-07 | 重庆邮电大学 | A kind of analysis of business electrical amount and prediction technique based on data mining |
-
2018
- 2018-11-01 CN CN201811296952.7A patent/CN109508820A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106779219A (en) * | 2016-12-19 | 2017-05-31 | 北京中电普华信息技术有限公司 | A kind of electricity demand forecasting method and system |
CN108171379A (en) * | 2017-12-28 | 2018-06-15 | 无锡英臻科技有限公司 | A kind of electro-load forecast method |
CN108510006A (en) * | 2018-04-08 | 2018-09-07 | 重庆邮电大学 | A kind of analysis of business electrical amount and prediction technique based on data mining |
Non-Patent Citations (3)
Title |
---|
HAIYING LI 等: "Daily Electricity Consumption Forecasting Based on Lazy Learning", 《PROCEEDINGS OF 2018 2ND INTERNATIONAL CONFERENCEON ELECTRONIC INFORMATION TECHNOLOGY AND COMPUTER ENGINEERING》 * |
刘欢 等: "基于分化距离的离群点检测算法", 《计算机应用研究》 * |
翟坤: "基于数据挖掘的成本管理方法研究-以钢铁企业为例", 《中国博士学位论文全文数据库 信息科技辑》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110084158A (en) * | 2019-04-15 | 2019-08-02 | 杭州拓深科技有限公司 | A kind of electrical equipment recognition methods based on intelligent algorithm |
CN110084158B (en) * | 2019-04-15 | 2021-01-29 | 杭州拓深科技有限公司 | Electric equipment identification method based on intelligent algorithm |
CN110689250A (en) * | 2019-09-20 | 2020-01-14 | 深圳供电局有限公司 | Method and system for processing user electricity consumption data and computer readable medium |
CN110689250B (en) * | 2019-09-20 | 2023-04-14 | 深圳供电局有限公司 | Method and system for processing user electricity consumption data and computer readable medium |
CN114553538A (en) * | 2022-02-22 | 2022-05-27 | 国网山东省电力公司电力科学研究院 | Active protection method and system for power grid information safety |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106373025B (en) | Power information acquisition system based on outlier detection anti-electricity-theft monitoring method in real time | |
Zhang et al. | A new index and classification approach for load pattern analysis of large electricity customers | |
CN110990461A (en) | Big data analysis model algorithm model selection method and device, electronic equipment and medium | |
CN109871976A (en) | A kind of prediction technique of power quality containing distributed power distribution network based on cluster and neural network | |
CN109630095B (en) | A kind of rod-pumped well operating mode's switch method and system based on multi-angle of view study | |
CN109508820A (en) | Campus electricity demand forecasting modeling method based on differentiation modeling | |
CN105868853A (en) | Method for predicting short-term wind power combination probability | |
CN113255900A (en) | Impulse load prediction method considering improved spectral clustering and Bi-LSTM neural network | |
CN105930900B (en) | The Forecasting Methodology and system of a kind of hybrid wind power generation | |
CN113901977A (en) | Deep learning-based power consumer electricity stealing identification method and system | |
CN111563682A (en) | Test evaluation method for distribution automation equipment | |
CN109359665A (en) | A kind of family's electric load recognition methods and device based on support vector machines | |
CN110401977A (en) | A kind of more floor indoor orientation methods returning more Classification and Identification devices based on Softmax | |
CN116308883A (en) | Regional power grid data overall management system based on big data | |
CN116090757A (en) | Method for evaluating capability demand satisfaction of information guarantee system | |
CN112287980A (en) | Power battery screening method based on typical feature vector | |
Miraftabzadeh et al. | K-means and alternative clustering methods in modern power systems | |
CN109886488A (en) | Consider the distributing wind power plant layering mixing short term prediction method of wind speed time lag | |
CN111062539A (en) | Total electric quantity prediction method based on secondary electric quantity characteristic clustering analysis | |
CN110533243A (en) | A kind of building energy consumption prediction technique based on fractal theory | |
Ullah et al. | Adaptive data balancing method using stacking ensemble model and its application to non-technical loss detection in smart grids | |
CN114021837A (en) | Regional power consumption prediction method based on hybrid machine learning and spatial address matching | |
CN110321390A (en) | Based on the load curve data visualization method for thering is supervision and unsupervised algorithm to combine | |
CN104821854B (en) | A kind of many primary user's multidimensional frequency spectrum sensing methods based on random set | |
CN117458480A (en) | Photovoltaic power generation power short-term prediction method and system based on improved LOF |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190322 |