CN105989420A - Method of determining user electricity consumption behavior features, method of predicting user electricity consumption load and device - Google Patents
Method of determining user electricity consumption behavior features, method of predicting user electricity consumption load and device Download PDFInfo
- Publication number
- CN105989420A CN105989420A CN201510076866.5A CN201510076866A CN105989420A CN 105989420 A CN105989420 A CN 105989420A CN 201510076866 A CN201510076866 A CN 201510076866A CN 105989420 A CN105989420 A CN 105989420A
- Authority
- CN
- China
- Prior art keywords
- user
- electricity consumption
- load
- history
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Supply And Distribution Of Alternating Current (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The method of determining user electricity consumption behavior features comprises steps: S10, according to the similarity of multiple historical daily electricity consumption load curves of the user, and clustering is carried out on the multiple historical daily electricity consumption load curves according to different time intervals; S20, statistics processing is carried out on the historical electricity consumption load data of the user and historical current data to obtain an electricity consumption feature model for the user at each time interval; S30, controllable factors and uncontrollable factors for influencing the user electricity consumption behaviors are acquired; and S40, the following data serve as input of an initial electricity consumption load prediction model, the initial electricity consumption load prediction model of the user is trained, and a standard electricity consumption load prediction model reflecting the user electricity consumption behavior features is obtained, wherein the data comprise the controllable factors and the uncontrollable factors for influencing the user electricity consumption behaviors, the historical electricity consumption load data of the user and historical electricity consumption load values obtained according to the electricity consumption feature model. The standard electricity consumption load prediction model can accurately reflect the electricity consumption behavior features of the user.
Description
Technical field
The present invention relates to power domain, particularly relate to the prediction of electric load.
Background technology
The electricity consumption behavior (mainly including power load) of power consumer is often change, not only different in the one day time
Change, and be also sensitive to factors such as season, temperature, weather, different seasons, the weather of different regions, Yi Jiwen
Load all can be caused significantly impact by the change of degree.In addition, the electric load of different industries is also different.Such as commercial user,
Including food and drink, show business, owing to commercial activity can increase the business hours in festivals or holidays, thus become and festivals or holidays affect electric power
One of key factor of load.The proportion of general industry load occupy the first in electricity consumption composition, and industrial user (mainly includes
Factory) power load depend not only on the working method (including the shift etc. of machine utilization, enterprise) of industrial user,
And be closely related with the industry characteristic of every profession and trade.User is predicted in realization by electricity consumption behavioural characteristic exactly that accurately determine user
Power load extremely important.
Electro-load forecast includes peak load, minimum load, year (or day) average load, power consumption, load curve etc.
Prediction.Peak load prediction is very important for the capacity of determination power system generating equipment and power transmission and transforming equipment.In order to select
Select suitable machine set type and rational power supply architecture and determine fuel planning etc., it is necessary to averagely bearing in prediction a period of time
Lotus and electricity.The prediction of load curve can be the peak value of research power system, the capacity of hydroenergy storage station and a transmission facility
Coordinated operation provide data support.
With the development of electricity market, the importance of load forecast manifests day by day, and the requirement to load prediction precision is got over
Come higher.Traditional Forecasting Methodology comparative maturity, it was predicted that result has certain reference value, but to improve prediction essence further
Degree, it is necessary to carry out some to conventional method and improve, simultaneously with the continuous progress of modern science and technology, theoretical research is progressively
Deeply, occur in that the emerging interdisciplinary theory with gray theory, expert system theory, fuzzy mathematics etc. as representative, be load
Developing rapidly of prediction provides theoretical foundation and Fundamentals of Mathematics.
Content of the invention
An embodiment of the invention provides a kind of method determining user power utilization behavioural characteristic, and the method includes: obtain this use
The historical load data at family, obtain a plurality of history daily load curve of this this user, according to the similarity of a plurality of daily load curve,
Realize a plurality of history daily load curve is carried out clustering (S10) according to different time intervals.Obtain with this user in each time
The historical load data and historical current data obtaining are carried out statistical procedures, obtain this use by the historical current data on interval
Electricity consumption characteristic model on each time interval for the family, the combination of the electricity consumption characteristic model of each time interval represents the use of this user
Electricity behavioural characteristic (S20).Obtain user's controllable factor and the uncontrollable factor (S30) affecting user power utilization behavior；With by following
Data, as the input of an initial electro-load forecast model, train this user initial electro-load forecast model, and obtaining reflection should
Standard electro-load forecast model (S40) of user power utilization behavioural characteristic, described data include: affect this user power utilization behavior
The electricity consumption characteristic model of user's controllable factor, uncontrollable factor, the history power load data of this user and this user or foundation should
The history power load value that the electricity consumption characteristic model of user obtains.
For updating the electricity consumption behavioural characteristic of user, following steps can be periodically executed: obtain the historical load data of this user,
To a plurality of history daily load curve of this this user, according to the similarity of a plurality of daily load curve, it is achieved by a plurality of history daily load
Curve carries out clustering (S10) according to different time intervals.Obtain and historical current data on each time interval for this user,
Statistical procedures is carried out to the historical load data and historical current data obtaining, obtains use on each time interval for this user
Electrical feature model, the combination of the electricity consumption characteristic model of each time interval represents the electricity consumption behavioural characteristic (S20) of this user.
The another one embodiment of the present invention provides a kind of method being used for and predicting user power utilization load, obtains the use to this user
Electricity behavior produces user's controllable factor and the uncontrollable factor (S50) of impact；Obtain history on each time interval for this user
The historical load data and historical current data obtaining are carried out statistical procedures, obtain this user in each time by current data
Electricity consumption characteristic model (S100) on interval；With user's controllable factor of acquisition, uncontrollable factor and this user used electrical feature
The history power load value that model or the electricity consumption characteristic model according to this user obtain is input in standard electro-load forecast model,
Predict the power load (S60) of this user.
An embodiment of the invention provides a kind of power load standard electro-load forecast model making to obtain in aforementioned manners
The method predicting new user power utilization load.The method includes: the user that the electricity consumption behavior on new user that obtains produces impact is controlled
Factor and uncontrollable factor (S70)；Determine electricity consumption behavior and the immediate old user of new user (S80)；Acquisition is determined
The history power load data of old user, obtain a plurality of history day power load curve of old user, according to a plurality of day power load
The similarity of curve, it is achieved a plurality of history day power load curve of old user is clustered according to different time intervals
(S90)；Obtain historical current data on each time interval for the old user, to the history power load data obtaining and history
Current data carries out statistical procedures, obtains electricity consumption characteristic model (S200) on each time interval for the old user；With will obtain
The electricity consumption characteristic model of the new user's controllable factor, uncontrollable factor and the old user that take or the electricity consumption characteristic model according to old user
The standard power load that the historical load value obtaining is input to this old user of reflection that method according to claim 1 obtains is pre-
Survey in model, it is achieved predict the power load (S300) of new user；Wherein new user refers to that history power load data are less than first
Preset time length user, old user refer to history power load data more than second preset time length user.
An embodiment of the invention provides a kind of device determining user power utilization behavioural characteristic, comprising: cluster cell, is used for
Obtain the historical load data of this user, obtain a plurality of history daily load curve of this this user, according to a plurality of daily load curve
Similarity, it is achieved a plurality of history daily load curve is clustered according to different time intervals；Electricity consumption behavioural characteristic determining unit,
For obtaining and historical current data on each time interval for this user, to the historical load data obtaining and historical current number
According to carrying out statistical procedures, obtaining electricity consumption characteristic model on each time interval for this user, the electricity consumption of each time interval is special
The combination levying model represents the electricity consumption behavioural characteristic of this user；Influence factor acquiring unit, being used for obtaining affects user power utilization behavior
User's controllable factor and uncontrollable factor；Model training unit, for using following data as the initial power load of user
The input of forecast model, trains initial electro-load forecast model, obtains reflecting the standard power load of this user power utilization behavioural characteristic
Forecast model, described data include: affect user's controllable factor of this user power utilization behavior, uncontrollable factor, the going through of this user
History power load value that history load data and the electricity consumption characteristic model according to this user obtain or the electricity consumption characteristic model of this user.
Brief description
Fig. 1 shows the flow chart of a kind of method of exemplary determination user power utilization behavioural characteristic；
Fig. 2 A-2B shows the cluster result to 6 history daily load curves at 6 days for the user；
Fig. 3 A-3B shows that the history daily load curve to the user shown in Fig. 1 carries out returning the result obtaining；
Fig. 4 shows the flow chart of a kind of method of exemplary prediction user power utilization load；
Fig. 5 shows the flow chart of a kind of method of exemplary new user power utilization load of prediction；With
Fig. 6 shows a kind of device determining user power utilization behavioural characteristic.
Detailed description of the invention
In order to the purpose, technical scheme and the advantage that make the embodiment of the present invention are clearer, below illustrate and the embodiment of the present invention is carried out
Further describe in detail.
Fig. 1 shows the flow chart of a kind of method of exemplary determination user power utilization behavioural characteristic.This determines user power utilization behavior
The method of feature obtains the historical load data of this user, obtains a plurality of history daily load curve of this this user, according to a plurality of day
The similarity of load curve, it is achieved a plurality of history daily load curve is carried out clustering (S10) according to different time intervals.Afterwards,
Obtain and historical current data on each time interval for this user, the historical load data and historical current data obtaining are entered
Row statistical procedures, obtains electricity consumption characteristic model on each time interval for this user, the electricity consumption character modules of each time interval
The combination of type represents the electricity consumption behavioural characteristic (S20) of this user.Obtain affect user's controllable factor of user power utilization behavior with can not
Control factor (S30)；Finally, using following data as the input of an initial electro-load forecast model, train this user initial
Electro-load forecast model, obtains reflecting standard electro-load forecast model (S40) of this user power utilization behavioural characteristic, described number
According to including: affect user's controllable factor of this user power utilization behavior, uncontrollable factor, this user history power load data and
The electricity consumption characteristic model of this user or the history power load value obtaining according to the electricity consumption characteristic model of this user.It is appreciated that
The multiple steps stated in method can perform in other orders.Affect the user of user power utilization behavior for example, it is possible to obtain at first
Controllable factor and user's uncontrollable factor (S30), then perform other steps again.
Below in conjunction with the accompanying drawings said method is described in detail.Needing exist for explanation, in Fig. 2-Fig. 3, history day is born
Lotus curve is the historical load value by user of sampling in every 15 minutes, divide into 96 points by one day 24 hours.Horizontal stroke in figure
These time points of coordinate representation, ordinate represents the history power load value of user.
User in the method refers to use people or the unit of the electric power of Utilities Electric Co.'s offer.Owing to the need for electricity of unit user is big,
And unit user is due to the difference of place industry etc., it will usually show feature total in a certain or some industry, therefore,
Embodiments of the invention are more suitable for unit user, but also can apply personal user.Following citing is enter with unit user
Row citing.The data of the power of actual loading when power load is to show that user uses electric power.
The method obtains the historical load data of this user, obtains a plurality of history daily load curve of this this user, according to a plurality of day
The similarity of load curve, it is achieved a plurality of history daily load curve is carried out clustering (S10) according to different time intervals.From with
The power supply administration at place, family can obtain the history power load data of user, can generate a plurality of going through according to history power load data
History daily load data and curves.For example can obtain 365 daily load data and curves that user was at 2014 365 days.
Industrial user changes greatly in the interval power load of different time.For example, a lot of factories day Saturday and holiday no collection,
Comparing with workaday power load, the power consumption of its day Saturday and festivals or holidays can lack a lot.For another example, the little steel mill having in order to
Reduce electricity cost, be simply engaged in production relatively low night in electricity price.The power load of this little steel mill night is far longer than it
The power load on daytime.In view of this power consumption characteristics of industrial user, this method by the history daily load curve of a user according to
Different time intervals clusters.So can improve the accuracy of cluster, user can be determined more subtly at different time
Interval electricity consumption behavioural characteristic.Fig. 2 A-2B shows the cluster result to 6 daily load curves at 6 days for the user.From
It will be seen that 3 three workaday daily load curve similarities are higher in Fig. 2 A and 2B, the daily load of three days Saturday is bent
Line similarity is higher, therefore, this 6 daily load curves be clustered respectively on weekdays with on day Saturday two time intervals.For
For the sake of Jian Dan, Fig. 2-Fig. 3 has been merely illustrative 6 daily load curves, actually generally requires and bears use more day
Lotus curve, the time interval marking off also can be more.The such as similarity according to daily load curve may be divided into: workaday in vain
My god, weeknight, day Saturday and four time intervals festivals or holidays.
After a plurality of history daily load curve of user is clustered on time interval, obtain with this user at each time interval
On historical current data.The historical current data each time interval can being divided into after power supply administration obtains corresponding to cluster.
Then, statistical procedures is carried out to the historical load data and historical current data obtaining, obtain this user at each time interval
On electricity consumption characteristic model, the combination of the electricity consumption characteristic model of each time interval represents the electricity consumption behavioural characteristic (S20) of this user.
After statistical procedures is carried out to historical load data and current data, user can be deviateed normal electricity consumption behavior and revert to
Normal electricity consumption behavior.For example, the unusual ground in certain area in certain summer is hot, and a refrigeration plant of this area is in this summer
Power load can deviate normal range (NR), by statistical procedures, can be by the electricity consumption data regression in this summer to normal summer
Power load scope.
Multiple existing statistical algorithms can be used to carry out processing to the historical load data obtaining and current data, and (i.e. data are dug
Pick), obtain electricity consumption characteristic model on each time interval for this user.Different systems can be used according to the power consumption characteristics of user
Algorithm learned by meter.A kind of simple statistical algorithms is to return each section of history daily load curve on each time interval respectively,
Using the functional relation of regression curve that obtains as the electricity consumption characteristic model on each time interval.Fig. 3 shows and shows in Fig. 1
The history daily load curve of the user going out carries out returning the result obtaining.It is used herein a kind of simple regression algorithm, i.e. ask for
The mean value of the daily load curve on each time interval.Wherein Fig. 3 A is for asking for user 3 workaday history daily loads
The average load curve on working day that the mean value of curve obtains, Fig. 3 B is the history daily load song asking for user 3 days Saturday
Per day load curve Saturday that the mean value of line obtains.The function representation of this two sections of per day load curves is the use of this user
Electrical feature model.The combination of these electricity consumption characteristic models just represents the electricity consumption behavioural characteristic of this user.For example, the use of a user
Being combined as of electrical feature model:
Wherein, f1 and f2 representative function relation, P represents the power load value of user.
Obtain user's controllable factor and the user's uncontrollable factor (S30) affecting user power utilization behavior.The controllable factor of user is
The factor that user can regulate and control.On the contrary, the factor that the uncontrollable factor user of user cannot regulate and control.For example to a steel mill
Speech, uncontrollable factor includes: meteorologic factor, labour market fluctuating factor, steel market fluctuating factor, electrical network power cuts to limit consumption
Factor and factor festivals or holidays etc..Controllable factor includes: the production device type of use and quantity, production scale and the production schedule,
Plant maintenance time, arrangement festivals or holidays etc. of enterprise oneself.
Finally, using following data as the input of an initial electro-load forecast model, the initial electro-load forecast of this user is trained
Model, obtains reflecting standard electro-load forecast model (S40) of this user power utilization behavioural characteristic, and described data include: impact
User's controllable factor of this user power utilization behavior, uncontrollable factor, the historical load data of this user and the electricity consumption according to this user
History power load value that characteristic model obtains or the electricity consumption characteristic model of this user.This standard electro-load forecast model can be used
Predict short-term, the power load of such as some day, it is also possible to medium-term and long-term for predicting, the power load in such as a certain year.
Meanwhile, power load can be the maximum power load of user, minimum power load or average power load etc..
User's initial electro-load forecast model can be to utilize existing time series algorithm, neural network algorithm or support vector
The model that machine algorithm (Support vector machines, SVM) scheduling algorithm is set up.Time series method is a kind of common load
Forecasting Methodology, it is the characteristic of certain random process presenting for whole observation sequence, goes to set up and estimate to produce actual sequence
The model of the random process of row, then goes to be predicted with these models.It make use of the inertia characteristics that electric load changes and when
Continuity between, by analyzing and processing historical data seasonal effect in time series, determines its essential characteristic and Changing Pattern, it was predicted that not
Carry out load.Neural network algorithm is the learning functionality utilizing neutral net, allows computer learning be included in historical load data
Mapping relations, recycle this mapping relations prediction future load.This algorithm has very strong robustness, memory capability, non-thread
Property mapping ability and powerful self-learning capability.Algorithm of support vector machine is similar with neutral net, is all learning-oriented mechanism,
But use mathematical method and optimisation technique from algorithm of support vector machine unlike neutral net.Algorithm of support vector machine is not only
Be suitable to predict more stable data, be also suitably for predicting the data that fluctuation ratio is bigger.But this algorithm is complicated, it is difficult to realize, and
And it is slow to calculate speed.These characteristics based on algorithm of support vector machine, it is possible to use the electricity consumption of this algorithm prediction Wave crest and wave trough period
Load, it was predicted that ratio of precision time series precision is high.For the sequence of nonlinearity, time series models need to set it
Non-linear form, this is relatively difficult.Use the initial model that neural network algorithm or algorithm of support vector machine are set up in such cases
More preferably.Neural network model (initial model i.e. set up by neural network algorithm) can be with arbitrary accuracy Nonlinear Function Approximation.
Should be noted that when using neural network model at the beginning of structure (the mainly number of hidden neuron) and the parameter of setting neutral net
Initial value, it is to avoid over-fitting and local minimum.Those skilled in the art can according to the These characteristics of these existing algorithms, in conjunction with
The electricity consumption behavioral characteristic of user determines which algorithm specifically used.
User's initial electro-load forecast model can also is that the combination of the multiple initial predicted models set up at times.Due to a lot
User differed greatly in festivals or holidays and non-festivals or holidays, daytime and night, the power load of Various Seasonal.So in order to carry further
High electro-load forecast precision, first can be divided into festivals or holidays and non-festivals or holidays the history power load data in such as dining room
Two classes.One can be used to set up the initial electro-load forecast model when festivals or holidays for this dining room with time series algorithm, use
Algorithm of support vector machine sets up the initial electro-load forecast model in non-festivals or holidays for this dining room.
It is comprehensive that initial user electro-load forecast model can also is that the combination of multiple initial predicted model with different weight is formed
Matched moulds type, thus can the advantage of comprehensive different forecast models, improve precision of prediction.For example, it is possible to be applied in combination existing
Gray model algorithm, exponential smoothing algorithm and rolling average algorithm, based on the understanding to these three algorithm precision of prediction, Ke Yifu
Giving its different weight, such as weight is followed successively by the 0.5th, the 0.3rd, 0.2.The collective model of these three algorithm groups synthesis can be as user
The initial predicted model of power load.
Set up the initial electro-load forecast model (including the built-up pattern that the combination of the aforementioned plurality of forecast model is formed) of user
After, using following data as the input of initial electro-load forecast model, train initial electro-load forecast model, i.e. can determine that
Go out the parameter in initial electro-load forecast model, obtain reflecting the standard electro-load forecast model of this user power utilization behavioural characteristic
(S40), above-mentioned input data include: affect user's controllable factor of this user power utilization behavior, uncontrollable factor, this user
History power load value that history power load data and the electricity consumption characteristic model according to this user obtain or this user use electrical feature
Model.
For example, the initial electro-load forecast model of foundation can be P=Ax1+Bx2+Cx3+Dx4.Variable X 1 can be defined
For representing user's controllable factor, variable X 2 being defined as user's uncontrollable factor, the history electricity consumption that X3 definition is represented as user is born
Lotus data, X4 definition is represented as the history power load value obtaining according to the electricity consumption characteristic model of this user, the i.e. history of user
Power load value.For example, the history power load value that the electricity consumption characteristic model of foundation user obtains is for the example above, just
It is based on electricity consumption characteristic model f1(t) and f2The history power load value of t user that () calculates.For another example, the initial electricity consumption of foundation
Load forecasting model can be P=g (f1(t), f2(t))+h (z)+i (n)+j (m), i.e. electricity consumption characteristic model f1(t) and f2(t)
It is the variable of initial electro-load forecast model.H (z), i (n) and j (m) can be respectively defined as representing and affect user power utilization
The historical load data of the controllable factor of load, uncontrollable factor and user, wherein g, h, i and j represent a kind of function respectively
Relation.
The method of above-mentioned determination user power utilization behavioural characteristic, can by realizing from cluster to the history daily load curve of a user
Determine the different electricity consumption behavioural characteristics in different time interval for the user.So that the standard electro-load forecast model energy training
Enough electricity consumption behavioural characteristics reflecting this user more accurately.Meanwhile, said method affects the controlled of user power utilization load by using
The electricity consumption behavioural characteristic curve of factor and user, the electricity consumption characteristic model of user and the history electricity consumption obtaining according to electricity consumption characteristic model
Load value can reflect that user's self uses electrical feature.For example, for industrial user, it can reflect that the production of user sets
Standby type and equipment use frequency, the use time, the inspection of production equipment repair the users such as cycle self use electrical feature.By using
The history power load value that the electricity consumption characteristic model of user and foundation electricity consumption characteristic model obtain is to train initial electro-load forecast mould
Type, it becomes possible to enable the master pattern that trains closer to the power consumption characteristics of user self.In addition, by history such as load
Data and current data carry out statistical procedures, can filter out the abnormal history power load data of user and current data, this
Sample can further improve the precision of prediction.Finally, due to use the history power load data training initial user electricity consumption of user
Load forecasting model so that the standard electro-load forecast model of generation can reflect the electricity consumption behavioural characteristic of user more accurately.
Especially when user has nearest history power load data, it is possible to use nearest history power load data training initial user
Electro-load forecast model, the standard electro-load forecast model so training can reflect the nearest electricity consumption of this user exactly
Behavioural characteristic.
Determine in the method that user determines user power utilization behavioural characteristic at another, be wherein periodically executed above-mentioned cluster (S10) and
Return (S20) step, so can update the electricity consumption behavioural characteristic of user so that electricity consumption behavioural characteristic can reflect that user is nearest
Electricity consumption behavioural characteristic.Update frequency can be one month once or half a year once, this is mainly determined by the power consumption characteristics of user
Fixed.
Fig. 4 shows the flow chart of a kind of method of exemplary prediction user power utilization load.The side of this prediction user power utilization load
Method is capable of utilizing the history power load data of a user to calculate to a nicety the following power load of this user.The method bag
Include: obtain this user and realize the historical load data on each time interval of determination during cluster and historical current in the above-mentioned methods
The historical load data and historical current data obtaining are carried out statistical procedures, obtain this user at each time interval by data
On electricity consumption characteristic model (S100).The historical load value that these factors of acquisition and the electricity consumption characteristic model according to this user are obtained
Or the electricity consumption characteristic model of this user is input in the standard electro-load forecast model of this user that foundation said method obtains, in advance
Survey the power load (S60) of this user.
Owing to needing to use the historical load data of user, historical current data and special according to the two electricity consumption determining in the method
Levy model, so for not or lack history daily load data and the new user of historical current data cannot use the method pre-
Survey its power load.
Fig. 5 shows the flow chart of a kind of method of exemplary new user power utilization load of prediction.The method includes: obtain to newly
The electricity consumption behavior of user produces user's controllable factor and the uncontrollable factor (S70) of impact；Determine special with new user power utilization behavior
Levy immediate old user (S80)；Obtain the history power load data of the old user determining, obtain a plurality of of old user and go through
History day power load curve, according to the similarity of a plurality of day power load curve, it is achieved press a plurality of history day power load curve
Carry out clustering (S90) according to different time intervals.Obtain historical current data on each time interval for the old user, to acquisition
History power load data and historical current data carry out statistical procedures, obtain electricity consumption on each time interval for the old user
Characteristic model (S200).By the to be predicted new user's controllable factor obtaining, uncontrollable factor and old user electricity consumption characteristic model or
Person according to the historical load value that the electricity consumption characteristic model of old user obtains be input to that method according to claim 1 obtains anti-
Reflect in the standard electro-load forecast model of old user, it is achieved predict the power load (S300) of new user.Here, new user is
Refer to history power load data less than first preset time length user, old user refer to history power load data be more than second
Preset time length user.For example, it is possible to the user setting history power load data time length less than half a year is as new user,
The user more than 1 year for the history power load data time length is old user.Very first time length and the second time span can phases
Deng.It should be noted that history electricity consumption data are referred to as new user less than the user of above-mentioned time span.Though the user for example having
The right electricity consumption time is longer, but its historical data may lose or do not gather, and such user falls within institute in this method
The new user stating.
Can be by the determination of following two ways and the electricity consumption behavioural characteristic immediate history power load data abundance of new user
Old user.When new user has certain historical load data, can history of forming daily load curve when, can be by this new user
History daily load curve and the daily load curve of multiple old user compare, the highest old with the daily load curve similarity of new user
User is the immediate old user of electricity consumption behavioural characteristic with new user.When the historical load data of new user are considerably less, it is impossible to
During history of forming daily load curve, can be according to power consumption parameter such as the power transformation capacity of new user, rated voltage and rated current, really
Make the immediate old user of these parameters with new user.Utilize the standard electro-load forecast model of this old user determining,
By user's controllable factor of the electricity consumption behavior generation impact on this new user to be predicted, uncontrollable factor and the electricity consumption behavior determined
Feature and the electricity consumption characteristic model of this immediate old user of new user or according to going through that the electricity consumption characteristic model of this old user obtains
History load value is input in the standard electro-load forecast model of old user, can realize predicting the power load of new user.
Fig. 6 shows a kind of device determining user power utilization behavioural characteristic.This device includes cluster cell the 12nd, electricity consumption behavior
Characteristics determining unit the 14th, influence factor acquiring unit 16 and model training unit 18.The function of these unit can pass through software
Realize.Cluster cell 12, for obtaining the historical load data of this user, obtains a plurality of history daily load curve of this this user,
Similarity according to a plurality of daily load curve, it is achieved a plurality of history daily load curve is clustered according to different time intervals.
Electricity consumption behavioural characteristic determining unit 14, for obtaining and historical current data on each time interval for this user, to obtain
Historical load data and historical current data carry out statistical procedures, obtain electricity consumption character modules on each time interval for this user
Type, the combination of the electricity consumption characteristic model of each time interval represents the electricity consumption behavioural characteristic of this user.Influence factor acquiring unit 16
For obtaining the user's controllable factor affecting user power utilization behavior and uncontrollable factor.Model training unit 18 is for by following data
As the input of the initial electro-load forecast model of user, train initial electro-load forecast model, obtain reflecting this user
The standard electro-load forecast model of electricity consumption behavioural characteristic, described data include: affect the user of this user power utilization behavior controlled because of
Element, the history power load data of uncontrollable factor and this user and the electricity consumption characteristic model of this user or the electricity consumption according to this user
The history power load value that characteristic model obtains.
The above, only presently preferred embodiments of the present invention, it is not intended to limit protection scope of the present invention.Concrete
Can be to being improved appropriately according to a preferred embodiment of the invention in implementation process, to adapt to the concrete needs of concrete condition.Cause
This is appreciated that the detailed description of the invention of invention as described herein simply plays an exemplary role, not in order to limit the guarantor of the present invention
Protect scope.
Claims (5)
1. the method determining user power utilization behavioural characteristic, comprising:
Obtain the history power load data of this user, obtain a plurality of history day power load curve of this this user, according to a plurality of
The similarity of day power load curve, it is achieved a plurality of history day power load curve is clustered according to different time intervals
(S10)；
Obtain and historical current data on each time interval for this user, to the history power load data obtaining and history electricity
Flow data carries out statistical procedures, obtains electricity consumption characteristic model on each time interval for this user, the use of each time interval
The combination of electrical feature model represents the electricity consumption behavioural characteristic (S20) of this user；
Obtain user's controllable factor and the uncontrollable factor (S30) affecting user power utilization behavior；With
Using following data as the input of an initial electro-load forecast model, train this user initial electro-load forecast model,
Obtaining reflecting the standard electro-load forecast model of this user power utilization behavioural characteristic, described data include: affect this user power utilization row
For user's controllable factor, uncontrollable factor, the electricity consumption characteristic model of the history power load data of this user and this user or depend on
History power load value (S40) obtaining according to the electricity consumption characteristic model of this user.
2. method according to claim 1, wherein, for updating the electricity consumption behavioural characteristic of user, is periodically executed following steps:
Obtain the history power load data of this user, obtain a plurality of history daily load curve of this this user, bear according to a plurality of day
The similarity of lotus curve, it is achieved a plurality of history daily load curve is carried out clustering (S10) according to different time intervals；
Obtain and historical current data on each time interval for this user, to the historical load data obtaining and historical current number
According to carrying out statistical procedures, obtaining electricity consumption characteristic model on each time interval for this user, the electricity consumption of each time interval is special
The combination levying model represents the electricity consumption behavioural characteristic (S20) of this user.
3. the user power utilization load criterion forecast model that a kind uses the method according to any one of claim 1-3 to obtain is to predict this use
The method of family power load, comprising:
The electricity consumption behavior on this user that obtains produces user's controllable factor and the uncontrollable factor (S50) of impact；
Obtain historical current data on each time interval for this user, the historical load data and historical current data obtaining are entered
Row statistical procedures, obtains electricity consumption characteristic model (S100) on each time interval for this user；With
By the electricity consumption characteristic model of user's controllable factor of acquisition, uncontrollable factor and this user or use electrical feature according to this user
The history power load value that model obtains is input in standard electro-load forecast model, it was predicted that the power load (S60) of this user.
4. the method predicting new user power utilization load, comprising:
The electricity consumption behavior on new user that obtains produces user's controllable factor and the uncontrollable factor (S70) of impact；
Determine electricity consumption behavioural characteristic and the immediate old user of new user (S80)；
Obtain the history power load data of old user determined, obtain a plurality of history day power load curve of old user, root
Similarity according to a plurality of day power load curve, it is achieved by old user a plurality of history day power load curve according to the different time
Interval carries out clustering (S90)；
Obtain historical current data on each time interval for the old user, to the history power load data obtaining and historical current
Data carry out statistical procedures, obtain electricity consumption characteristic model (S200) on each time interval for the old user；With
By electricity consumption characteristic model or the electricity consumption according to old user of the new user's controllable factor obtaining, uncontrollable factor and old user
The standard that the historical load value that characteristic model obtains is input to this old user of reflection that method according to claim 1 obtains is used
In electric load forecast model, it is achieved predict the power load (S300) of new user；
Wherein new user refer to history power load data less than first preset time length user, old user refers to history electricity consumption
Load data more than second preset time length user.
5. the device determining user power utilization behavioural characteristic, comprising:
Cluster cell (12), for obtaining the historical load data of this user, a plurality of history daily load obtaining this this user is bent
Line, according to the similarity of a plurality of daily load curve, it is achieved gather a plurality of history daily load curve according to different time intervals
Class；
Electricity consumption behavioural characteristic determining unit (14), for obtaining and historical current data on each time interval for this user,
Statistical procedures is carried out to the historical load data and historical current data obtaining, obtains use on each time interval for this user
Electrical feature model, the combination of the electricity consumption characteristic model of each time interval represents the electricity consumption behavioural characteristic of this user；
Influence factor acquiring unit (16), for obtaining user's controllable factor and the uncontrollable factor affecting user power utilization behavior；
Model training unit (18), for using following data as the input of the initial electro-load forecast model of user, training
Initial electro-load forecast model, obtains reflecting the standard electro-load forecast model of this user power utilization behavioural characteristic, described data
Including: affect user's controllable factor of this user power utilization behavior, uncontrollable factor, this user history power load data and should
The electricity consumption characteristic model of user or the history power load value obtaining according to the electricity consumption characteristic model of this user.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510076866.5A CN105989420B (en) | 2015-02-12 | 2015-02-12 | Method for determining electricity utilization behavior characteristics of user, and method and device for predicting electricity utilization load of user |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510076866.5A CN105989420B (en) | 2015-02-12 | 2015-02-12 | Method for determining electricity utilization behavior characteristics of user, and method and device for predicting electricity utilization load of user |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105989420A true CN105989420A (en) | 2016-10-05 |
| CN105989420B CN105989420B (en) | 2020-07-17 |
Family
ID=57042325
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510076866.5A Active CN105989420B (en) | 2015-02-12 | 2015-02-12 | Method for determining electricity utilization behavior characteristics of user, and method and device for predicting electricity utilization load of user |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105989420B (en) |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106570581A (en) * | 2016-10-26 | 2017-04-19 | 东北电力大学 | Attribute association based load prediction system and method in energy Internet environment |
| CN107146015A (en) * | 2017-05-02 | 2017-09-08 | 联想(北京)有限公司 | Multivariate Time Series Forecasting Methodology and system |
| CN107274025A (en) * | 2017-06-21 | 2017-10-20 | 国网山东省电力公司诸城市供电公司 | A kind of system and method realized with power mode Intelligent Recognition and management |
| CN108364120A (en) * | 2018-01-17 | 2018-08-03 | 华北电力大学 | Intelligent residential district demand response cutting load method based on user power utilization irrelevance |
| CN108776939A (en) * | 2018-06-07 | 2018-11-09 | 上海电气分布式能源科技有限公司 | The analysis method and system of user power utilization behavior |
| CN109149644A (en) * | 2018-09-29 | 2019-01-04 | 南京工程学院 | A kind of integrated strategy of on-line matching of light storage based on big data analysis and cooperative optimization method |
| CN109376971A (en) * | 2018-12-29 | 2019-02-22 | 北京中电普华信息技术有限公司 | A kind of load curve forecasting method and system towards power consumer |
| CN109948820A (en) * | 2017-12-20 | 2019-06-28 | 新智数字科技有限公司 | The optimization method and device of user's energy |
| CN110097261A (en) * | 2019-04-17 | 2019-08-06 | 三峡大学 | A method of judging user power utilization exception |
| CN110309982A (en) * | 2019-07-09 | 2019-10-08 | 南方电网科学研究院有限责任公司 | Power load prediction method, device and equipment based on matrix decomposition |
| CN110503232A (en) * | 2019-06-28 | 2019-11-26 | 国网浙江省电力有限公司湖州供电公司 | A kind of prediction of distributed photovoltaic load data and restorative procedure |
| CN110674993A (en) * | 2019-09-26 | 2020-01-10 | 广东电网有限责任公司 | User load short-term prediction method and device |
| CN110782094A (en) * | 2019-10-28 | 2020-02-11 | 国网江苏省电力有限公司苏州供电分公司 | Subentry load prediction method for fine-grained electricity consumption behaviors of residential users |
| CN110956321A (en) * | 2019-11-27 | 2020-04-03 | 深圳市恒泰能源科技有限公司 | Power consumption control method based on neural network, electronic device and storage medium |
| CN111582548A (en) * | 2020-04-14 | 2020-08-25 | 广东卓维网络有限公司 | Power load prediction method based on multivariate user behavior portrait |
| CN111612275A (en) * | 2020-05-29 | 2020-09-01 | 云南电网有限责任公司 | Method and device for predicting load of regional user |
| CN112418485A (en) * | 2020-10-27 | 2021-02-26 | 西安交通大学 | Household load prediction method and system based on load characteristics and power consumption behavior mode |
| CN112734135A (en) * | 2021-01-26 | 2021-04-30 | 吉林大学 | Power load prediction method, intelligent terminal and computer readable storage medium |
| CN113030758A (en) * | 2021-03-17 | 2021-06-25 | 重庆长安新能源汽车科技有限公司 | Aging early warning method and system based on lithium ion battery capacity water jump point, automobile and computer storage medium |
| CN113255973A (en) * | 2021-05-10 | 2021-08-13 | 曙光信息产业(北京)有限公司 | Power load prediction method, power load prediction device, computer equipment and storage medium |
| TWI755986B (en) * | 2020-12-22 | 2022-02-21 | 日商Ｊｆｅ鋼鐵股份有限公司 | Energy Utilization Support Device, Energy Utilization Support Method, and Steel Plant Operation Method |
| CN116979531A (en) * | 2023-09-25 | 2023-10-31 | 山西京能售电有限责任公司 | Novel energy data monitoring method and method for monitoring auxiliary power market |
| CN117220417A (en) * | 2023-11-07 | 2023-12-12 | 国网山西省电力公司信息通信分公司 | Dynamic monitoring method and system for consumer-side electrical load |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2012244897A (en) * | 2011-05-13 | 2012-12-10 | Fujitsu Ltd | Apparatus and method for predicting short-term power load |
| CN103065201A (en) * | 2012-12-19 | 2013-04-24 | 福建省电力有限公司 | Electric load prediction method used for electric power based on factors of temperature and festivals and holidays |
| CN103729689A (en) * | 2013-12-20 | 2014-04-16 | 华南理工大学 | Power grid electric quantity prediction method based on industry classifications and leading industry data |
| CN103793887A (en) * | 2014-02-17 | 2014-05-14 | 华北电力大学 | Short-term electrical load on-line predicting method based on self-adaptation enhancing algorithm |
| US20140358838A1 (en) * | 2013-06-04 | 2014-12-04 | International Business Machines Corporation | Detecting electricity theft via meter tampering using statistical methods |
| CN104200277A (en) * | 2014-08-12 | 2014-12-10 | 南方电网科学研究院有限责任公司 | Medium-and-long-term power load prediction model establishment method |
| CN104200275A (en) * | 2014-06-24 | 2014-12-10 | 国家电网公司 | Power utilization mode classification and control method based on user behavior characteristics |
-
2015
- 2015-02-12 CN CN201510076866.5A patent/CN105989420B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2012244897A (en) * | 2011-05-13 | 2012-12-10 | Fujitsu Ltd | Apparatus and method for predicting short-term power load |
| CN103065201A (en) * | 2012-12-19 | 2013-04-24 | 福建省电力有限公司 | Electric load prediction method used for electric power based on factors of temperature and festivals and holidays |
| US20140358838A1 (en) * | 2013-06-04 | 2014-12-04 | International Business Machines Corporation | Detecting electricity theft via meter tampering using statistical methods |
| CN103729689A (en) * | 2013-12-20 | 2014-04-16 | 华南理工大学 | Power grid electric quantity prediction method based on industry classifications and leading industry data |
| CN103793887A (en) * | 2014-02-17 | 2014-05-14 | 华北电力大学 | Short-term electrical load on-line predicting method based on self-adaptation enhancing algorithm |
| CN104200275A (en) * | 2014-06-24 | 2014-12-10 | 国家电网公司 | Power utilization mode classification and control method based on user behavior characteristics |
| CN104200277A (en) * | 2014-08-12 | 2014-12-10 | 南方电网科学研究院有限责任公司 | Medium-and-long-term power load prediction model establishment method |
Cited By (32)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106570581A (en) * | 2016-10-26 | 2017-04-19 | 东北电力大学 | Attribute association based load prediction system and method in energy Internet environment |
| CN106570581B (en) * | 2016-10-26 | 2019-06-28 | 东北电力大学 | Load prediction system and method under energy internet environment based on Attribute Association |
| CN107146015A (en) * | 2017-05-02 | 2017-09-08 | 联想(北京)有限公司 | Multivariate Time Series Forecasting Methodology and system |
| CN107274025A (en) * | 2017-06-21 | 2017-10-20 | 国网山东省电力公司诸城市供电公司 | A kind of system and method realized with power mode Intelligent Recognition and management |
| CN107274025B (en) * | 2017-06-21 | 2020-09-11 | 国网山东省电力公司诸城市供电公司 | System and method for realizing intelligent identification and management of power consumption mode |
| CN109948820A (en) * | 2017-12-20 | 2019-06-28 | 新智数字科技有限公司 | The optimization method and device of user's energy |
| CN108364120A (en) * | 2018-01-17 | 2018-08-03 | 华北电力大学 | Intelligent residential district demand response cutting load method based on user power utilization irrelevance |
| CN108776939A (en) * | 2018-06-07 | 2018-11-09 | 上海电气分布式能源科技有限公司 | The analysis method and system of user power utilization behavior |
| CN109149644B (en) * | 2018-09-29 | 2020-06-09 | 南京工程学院 | Light-storage integrated online strategy matching and collaborative optimization method based on big data analysis |
| CN109149644A (en) * | 2018-09-29 | 2019-01-04 | 南京工程学院 | A kind of integrated strategy of on-line matching of light storage based on big data analysis and cooperative optimization method |
| CN109376971A (en) * | 2018-12-29 | 2019-02-22 | 北京中电普华信息技术有限公司 | A kind of load curve forecasting method and system towards power consumer |
| CN110097261A (en) * | 2019-04-17 | 2019-08-06 | 三峡大学 | A method of judging user power utilization exception |
| CN110097261B (en) * | 2019-04-17 | 2022-11-18 | 三峡大学 | Method for judging abnormal electricity utilization of user |
| CN110503232A (en) * | 2019-06-28 | 2019-11-26 | 国网浙江省电力有限公司湖州供电公司 | A kind of prediction of distributed photovoltaic load data and restorative procedure |
| CN110309982A (en) * | 2019-07-09 | 2019-10-08 | 南方电网科学研究院有限责任公司 | Power load prediction method, device and equipment based on matrix decomposition |
| CN110674993A (en) * | 2019-09-26 | 2020-01-10 | 广东电网有限责任公司 | User load short-term prediction method and device |
| CN110782094A (en) * | 2019-10-28 | 2020-02-11 | 国网江苏省电力有限公司苏州供电分公司 | Subentry load prediction method for fine-grained electricity consumption behaviors of residential users |
| CN110782094B (en) * | 2019-10-28 | 2022-06-21 | 国网江苏省电力有限公司苏州供电分公司 | Subentry load prediction method for fine-grained electricity consumption behaviors of residential users |
| CN110956321A (en) * | 2019-11-27 | 2020-04-03 | 深圳市恒泰能源科技有限公司 | Power consumption control method based on neural network, electronic device and storage medium |
| CN111582548A (en) * | 2020-04-14 | 2020-08-25 | 广东卓维网络有限公司 | Power load prediction method based on multivariate user behavior portrait |
| CN111612275B (en) * | 2020-05-29 | 2022-04-01 | 云南电网有限责任公司 | Method and device for predicting load of regional user |
| CN111612275A (en) * | 2020-05-29 | 2020-09-01 | 云南电网有限责任公司 | Method and device for predicting load of regional user |
| CN112418485A (en) * | 2020-10-27 | 2021-02-26 | 西安交通大学 | Household load prediction method and system based on load characteristics and power consumption behavior mode |
| TWI755986B (en) * | 2020-12-22 | 2022-02-21 | 日商Ｊｆｅ鋼鐵股份有限公司 | Energy Utilization Support Device, Energy Utilization Support Method, and Steel Plant Operation Method |
| CN112734135A (en) * | 2021-01-26 | 2021-04-30 | 吉林大学 | Power load prediction method, intelligent terminal and computer readable storage medium |
| CN113030758A (en) * | 2021-03-17 | 2021-06-25 | 重庆长安新能源汽车科技有限公司 | Aging early warning method and system based on lithium ion battery capacity water jump point, automobile and computer storage medium |
| CN113030758B (en) * | 2021-03-17 | 2022-05-06 | 重庆长安新能源汽车科技有限公司 | Aging early warning method and system based on lithium ion battery capacity water jump point, automobile and computer storage medium |
| CN113255973A (en) * | 2021-05-10 | 2021-08-13 | 曙光信息产业(北京)有限公司 | Power load prediction method, power load prediction device, computer equipment and storage medium |
| CN116979531A (en) * | 2023-09-25 | 2023-10-31 | 山西京能售电有限责任公司 | Novel energy data monitoring method and method for monitoring auxiliary power market |
| CN116979531B (en) * | 2023-09-25 | 2023-12-12 | 山西京能售电有限责任公司 | Novel energy data monitoring method and method for monitoring auxiliary power market |
| CN117220417A (en) * | 2023-11-07 | 2023-12-12 | 国网山西省电力公司信息通信分公司 | Dynamic monitoring method and system for consumer-side electrical load |
| CN117220417B (en) * | 2023-11-07 | 2024-02-09 | 国网山西省电力公司信息通信分公司 | Dynamic monitoring method and system for consumer-side electrical load |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105989420B (en) | 2020-07-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105989420A (en) | Method of determining user electricity consumption behavior features, method of predicting user electricity consumption load and device | |
| Leva et al. | Analysis and validation of 24 hours ahead neural network forecasting of photovoltaic output power | |
| Foley et al. | Current methods and advances in forecasting of wind power generation | |
| Wang et al. | Research and application of a hybrid forecasting framework based on multi-objective optimization for electrical power system | |
| Rana et al. | Multiple steps ahead solar photovoltaic power forecasting based on univariate machine learning models and data re-sampling | |
| Pfenninger | Dealing with multiple decades of hourly wind and PV time series in energy models: A comparison of methods to reduce time resolution and the planning implications of inter-annual variability | |
| Gomes et al. | Wind speed and wind power forecasting using statistical models: autoregressive moving average (ARMA) and artificial neural networks (ANN) | |
| Wu et al. | A novel hybrid model for short‐term forecasting in PV power generation | |
| Jursa et al. | Short-term wind power forecasting using evolutionary algorithms for the automated specification of artificial intelligence models | |
| Andresen et al. | Validation of Danish wind time series from a new global renewable energy atlas for energy system analysis | |
| Mamlook et al. | A fuzzy inference model for short-term load forecasting | |
| Islam et al. | Energy demand forecasting | |
| López et al. | Application of SOM neural networks to short-term load forecasting: The Spanish electricity market case study | |
| CN105844371A (en) | Electricity customer short-term load demand forecasting method and device | |
| Satish et al. | Effect of temperature on short term load forecasting using an integrated ANN | |
| Suomalainen et al. | Synthetic wind speed scenarios including diurnal effects: Implications for wind power dimensioning | |
| Ahmadi et al. | A fuzzy inference model for short-term load forecasting | |
| Santos et al. | Deep learning and transfer learning techniques applied to short-term load forecasting of data-poor buildings in local energy communities | |
| Gooding et al. | Probability analysis of distributed generation for island scenarios utilizing Carolinas data | |
| Dolara et al. | PV hourly day-ahead power forecasting in a micro grid context | |
| Kwon et al. | Weekly peak load forecasting for 104 weeks using deep learning algorithm | |
| Qureshi et al. | Short-term forecasting of wind power generation using artificial intelligence | |
| CN116526469A (en) | Long-term random dynamic scheduling method for water-wind-solar complementary system | |
| Nano et al. | Load forecasting using multiple linear regression with different calendars | |
| Safiyari et al. | From traditional to modern methods: Comparing and introducing the most powerful model for forecasting the residential natural gas demand |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |