CN108830405A - Real-time electric power load prediction system and method based on multi objective Dynamic Matching - Google Patents
Real-time electric power load prediction system and method based on multi objective Dynamic Matching Download PDFInfo
- Publication number
- CN108830405A CN108830405A CN201810525917.1A CN201810525917A CN108830405A CN 108830405 A CN108830405 A CN 108830405A CN 201810525917 A CN201810525917 A CN 201810525917A CN 108830405 A CN108830405 A CN 108830405A
- Authority
- CN
- China
- Prior art keywords
- load
- data
- index
- curve
- real
- 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
- 238000000034 method Methods 0.000 title claims abstract description 28
- 238000007405 data analysis Methods 0.000 claims abstract description 14
- 239000002245 particle Substances 0.000 claims description 48
- 230000006870 function Effects 0.000 claims description 25
- 238000012549 training Methods 0.000 claims description 10
- 230000010354 integration Effects 0.000 claims description 9
- 238000004458 analytical method Methods 0.000 claims description 8
- 230000005611 electricity Effects 0.000 claims description 8
- 230000005540 biological transmission Effects 0.000 claims description 6
- 240000002853 Nelumbo nucifera Species 0.000 claims description 4
- 235000006508 Nelumbo nucifera Nutrition 0.000 claims description 4
- 235000006510 Nelumbo pentapetala Nutrition 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000010219 correlation analysis Methods 0.000 claims description 4
- 238000013277 forecasting method Methods 0.000 claims description 4
- 238000007726 management method Methods 0.000 claims description 4
- 230000019771 cognition Effects 0.000 claims description 3
- 238000009313 farming Methods 0.000 claims description 3
- 230000013016 learning Effects 0.000 claims description 3
- 238000001556 precipitation Methods 0.000 claims description 3
- 230000009326 social learning Effects 0.000 claims description 3
- 238000005303 weighing Methods 0.000 claims description 3
- 230000003044 adaptive effect Effects 0.000 abstract description 4
- 241001269238 Data Species 0.000 abstract 1
- 238000009412 basement excavation Methods 0.000 abstract 1
- 238000012360 testing method Methods 0.000 description 5
- 239000011159 matrix material Substances 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 238000003786 synthesis reaction 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
- 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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- 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
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Educational Administration (AREA)
- Game Theory and Decision Science (AREA)
- Health & Medical Sciences (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)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The present invention is a kind of real-time electric power load prediction system and method based on multi objective Dynamic Matching, and the system includes:Data acquisition module is connect with data analysis module signal, and data analysis module is connect with multi objective Dynamic Matching module by signal, and multi objective Dynamic Matching module is connect with real-time electric power load prediction module by signal.The method can pass through the multiple datas indexs such as acquisition real-time electric power load data, meteorological data and economic data;The correlation between Power system load data and each factor index is analyzed, adaptive excavation influences the dynamic keyword set of factors of load forecast;And according to the dependence between electric load and key index, establish the adaptive index system of combined influence load forecast, by the A-GPR prediction model of proposition, realize that extract real-time key influence factor set carries out short-term and ultra-short term load forecast.
Description
Technical field
The invention belongs to smart grid Techniques for Prediction of Electric Loads fields, specifically, being a kind of based on multi objective dynamic
Matched real-time electric power load prediction system and method.
Background technique
Load forecast refers to, Operation of Electric Systems characteristic, increase-volume and natural situation are taken into account, history is utilized
Data predict following load.Load prediction is the important function of energy management system, is smart grid system
The basis of economic, the safe and reliable operation of system.The periodic regularity that the load of electric system has itself intrinsic, at the same also by
The influence of factors, such as weather conditions, economic factor.Due to the difference of the part throttle characteristics of each department, for differently
The difference of part throttle characteristics between area, the load prediction work for area all should be in conjunction with local actual conditions, in load spy
Property on the basis of consider the influence factor of load, then suitable method is selected to be predicted, to improve the precision of prediction.Load is pre-
The precision of survey directly affects safety, economy and the power supply quality of electric system, to the investment of power grid, scheduling, layout and
The science of operation has great influence, is the important evidence for carrying out development plan and real-time control.Therefore, how to improve pre-
Survey the emphasis that precision is current load prediction technical research.
Currently used load forecasting method includes traditional prediction method and modern prediction technique two major classes, wherein tradition is pre-
Survey method includes Time Series Method, regression analysis, grey method etc..It is wherein the widest with Time Series Method application
It is general.Time Series Method is to one-dimensional time series data, is according to deduction following a period of time with historical load data
Load data.Grey method is only needed in a small amount of historical data, and therefrom discovery influences the rule of load, and calculation amount is small.?
The load forecasting model set up on the basis of this, while being difficult to cope with the biggish electric load of fluctuation.And historical data is few
Amount and dispersion degree it is larger when, precision of prediction is poor.Modern prediction technique mainly has expert system approach, support vector machines, nerve
Network algorithm etc..Wherein neural network becomes the one of load prediction due to the ability of self-learning capability and processing complex nonlinear
The important method of kind.However, the structure and network parameter of neural network need to determine by subjective experience mostly.Accordingly, it is difficult to protect
The accuracy of prediction result is demonstrate,proved, it is pre- for load that some methods do not account for local meteorologic factor, economic factor in prediction
The influence of survey causes the missing of important information.It, can not be anti-even if considering the influence relationship of single factor and load data
All factors are reflected for the influence degree of load prediction, be easy to cause analysis result error occur, to influence load prediction
Precision.If various influence factors are included in input variable, it is excessive to will cause input variable, aggravates training burden,
Not only precision cannot be improved, reduces the performance of model prediction instead.Therefore, both consider the various factors of influence load prediction,
Compression input variable appropriate again, becoming load prediction must solve the problems, such as.
It is a kind of based on multi objective Dynamic Matching object of the present invention is to propose for the deficiency of existing load forecasting method
Real-time electric power load prediction system, the model of the system have comprehensively considered influence of multiple indexs for load, in processing load
During data, with the difference of the influence factor of load, the variation of time, the weight of each factor can be with the change of time
Change and changes.The present invention fully considered attribute directly or indirectly for the influence of load, make the distribution of last weight more
It is scientific and reasonable, so that key index be selected to establish index system for load prediction, the workload of load prediction is effectively reduced, is mentioned
The accuracy and reliability of high load capacity prediction.An A-GPR real-time electric power load forecasting model is established simultaneously, realizes smart grid
The timeliness of lower power supply and the effective use of new energy.
Summary of the invention
The object of the present invention is to which overcome the deficiencies in the prior art, proposes that one kind is structurally reasonable, functional burdening is predicted accurate
The real-time electric power load prediction system based on multi objective Dynamic Matching of property and high reliablity, and provide scientific and reasonable.
Realize one of the object of the invention the technical solution adopted is that:A kind of real-time electric power based on multi objective Dynamic Matching is negative
Lotus forecasting system, characterized in that it includes:For acquiring load, meteorology, economic data and carrying out Classification Management, system is constructed
The data acquisition module of database;The correlation between influence factor and load is calculated based on MIDM algorithm, is ranked up, is screened
Determinant attribute out constructs the data analysis module of comprehensive index system;Can be according to the random variation of different regions and time, it will
Fixed inertial factor is converted to the inertial factor of dynamic change in model, constructs the multi objective Dynamic Matching mould of A-GPR model
Block;The real-time electric power load that can carry out real-time electric power load prediction according to the comprehensive index system and A-GPR model of building is pre-
Survey module;The data acquisition module is connect with data analysis module signal, data analysis module and multi objective Dynamic Matching
Module by signal connection, multi objective Dynamic Matching module are connect with real-time electric power load prediction module by signal.
Realize the object of the invention two the technical solution adopted is that:A kind of real-time electric power based on multi objective Dynamic Matching is negative
Lotus prediction technique, characterized in that it include in have:
1) the respective function of the meteorological data unit of data acquisition module, load data unit and economic data unit is utilized
It can be carried out data acquisition, temperature, the humidity, precipitation, visibility, wind in area needed for being responsible for acquisition in real time by meteorological data unit
To, wind speed, weather condition data carry out Data Integration, be transferred in system database generation meteorological data table;By load data
Unit is responsible for acquisition industrial electricity load, farming power load, municipal power load, post and telecommunications power load, traffic electricity consumption in real time
Load, household electricity load and commercial power load carry out Data Integration, are transferred in system database and generate load data
Table;Data Integration is carried out by the economic data that economic data unit is responsible in acquisition government's annual economic report in real time, is transferred to
Economic data table is generated in system database;
2) correlation analysis, analysis load number are carried out to each index using MIDM algorithm using the function of data analysis module
According to the degree of correlation between corresponding meteorological data, economic data, so that it is determined that the key index in area needed for influencing, wherein
The step of multi objective dynamical Matching Algorithm is:
(1) each index curve is calculated for the projector distance of load:
Wherein xqiRepresent projection abscissa of the point on influence index curve on load curve, kqiRepresent load curve
The slope of this point, b1(2) abscissa put on influence index curve, b are represented1(1) it represents influence index and removes the vertical seat put upwards
Mark, bqiLoad curve is represented in the intercept of this point, yqiThe point on influence index is represented in the vertical seat of projection on load curve
Mark, q represent influence index 24 hours one day 24 points on curve, and i represents a certain influence index, rqiRepresent influence index pair
In the projector distance of load, xqi2Represent the abscissa of influence index curve subpoint the latter point on load curve, xqi1Generation
The abscissa of the previous point of table influence index curve subpoint on load curve, yqi2Influence index curve is thrown in load curve
The ordinate of shadow point the latter point, yqi1Influence index curve is represented in the ordinate of the previous point of load curve subpoint, Rqi
Load is represented in the fluctuation distance of point-to-point transmission, yt2Represent the ordinate of the latter point on load curve, yt1It represents on load curve
The ordinate of previous point, xt2Represent the abscissa of the latter point on load curve, xt1Represent previous point on load curve
Ordinate, t represent the load 24 hours one day points on curve;
(2) weight of each index is sought:
Wherein wiEach influence factor is indicated for the weighing factor of load, i represents a certain influence index influence index, rqi
Influence factor is represented for the projector distance of load, RtLoad is represented in the fluctuation distance of point-to-point transmission;
(3) weight calculation of aggregate index:
Wherein n represents the number of index, wi,jThe indirect weight of combined index is represented, i, j represent a certain item influence index,
wi' weight that represents aggregate index, give full expression to influence directly or indirectly of each index for load;
3) function of utilizing multi objective Dynamic Matching module, according to the real-time of the load data of different regions and key index
Property correlation establish comprehensive dynamic indicator system;A-GPR model is established using selected load data and comprehensive index system,
Its step is:
(1) initialization population scale N, maximum number of iterations Tmax;
(2)The vector of particle current location;
(3) Fitness, the fitness value of vector x;
(4)The speed of particle, its dimension and vectorIt is identical;
(5)pbest, fitness value corresponding to the desired positions that encounter during each particle flight;
(6)For recording the desired positions encountered during particle flight, pbest its dimension and vectorDimension
It is identical;
(7) the more new formula of particle position:
xi=xi+vi
Wherein, viRepresent the speed of particle, xiThe current location of particle is represented, i represents the current iteration that each particle updates
Number, gbestiRepresent the optimal location of population entirety, pbestiRepresent optimal location individual in population, w represent inertia weight because
Son, φ1Represent individual cognition learning rate, φ2It represents social learning to lead, random number of the rand () between section [0,1], K generation
Table convergence factor, K can be described as:
The inertia weight factor is generally set in standard particle group's algorithm:wt+1=wt=1, since the inertia weight factor is shadow
The variable of current particle speed is rung, biggish value is conducive to global search, and lesser value is conducive to local search, in order to preferably
Search capability is balanced, a kind of A-GPR model is proposed:
Wherein:wtRepresent the inertia weight factor of current iteration number, wmaxFor inertia maximum value, wminFor inertia minimum
Value, t are current iteration number, and T is maximum number of iterations,
So the more new formula of particle position is:
xi=xi+vi
The speed for leading to particle since the inertia weight factor is changing always and position are always in a wide range of and small of population
It is being continuously updated always in range, to balance algorithm in the ability of global search and local search;
(8) according to the position of updated particle, the fitness function E of particle is recalculated, fitness function can describe
For:
Wherein, m is input sample number, yiIndicate the output of process value of current A-GPR model training, yi-1For prior-generation mould
The output valve of type, i represent current number of iterations;
(9) judge whether to meet schedule requirement or whether reach the number of iterations, if meeting termination search;Otherwise it carries out down
Primary search;
(10) information of final particle, the as parameter of A-GPR model are exported;
4) function of utilizing real-time electric power load prediction module, obtains meteorological data, the economic data at moment to be predicted, builds
Vertical comprehensive index system is input in A-GPR model using comprehensive index system and corresponding load data as input vector,
Its load value for exporting the as moment to be predicted.
A kind of the advantages of real-time electric power load prediction system and method for multi objective Dynamic Matching of the invention, is embodied in:
1. due to being connect using data acquisition module with data analysis module signal, data analysis module and multi objective dynamic
The connection of matching module signal, the multi objective that multi objective Dynamic Matching module and real-time electric power load prediction module by signal connect and compose
The real-time electric power load prediction system of Dynamic Matching, it is structurally reasonable, there is data acquisition, data analysis, multi objective dynamic
With in one, can be realized with the functions such as real-time electric power load prediction extract real-time key influence factor set carry out it is in short term and super
Short-term load forecast;
2. due to load variation by weather condition, temperature, humidity, atmospheric pressure, visibility, wind direction, wind speed, GDP shadow
It rings, variation has apparent correlation, and the factor for influencing load has diversity, and method of the invention can be abundant
Consider the multiple indexes of influence load, accurately grasps the situation of change of load, the phase of each index is calculated according to MIDM algorithm
Guan Du, and be ranked up according to the degree of correlation between them, the key index for influencing load is filtered out, the variation characteristic of load is made
It is easier to hold, accuracy is high;
3. the A-GPR method in real-time electric power load prediction module of the invention, make originally fixed model parameter can be with
Adaptive real-time change is carried out according to the input of different indexs, improves the adaptability and robustness of model, is electric power from now on
One completely new developing direction of system electro-load forecast, would be more advantageous in the development of smart grid and energy internet;
4. its methodological science, strong applicability, effect are good.
Detailed description of the invention
Fig. 1 is a kind of real-time electric power load prediction system structure block diagram of multi objective Dynamic Matching;
Fig. 2 is the work flow diagram of Fig. 1.
Specific embodiment
Referring to Figures 1 and 2, a kind of real-time electric power load prediction system based on multi objective Dynamic Matching, including:For adopting
Collection load, meteorology, economic data simultaneously carry out Classification Management, construct the data acquisition module of system database;Based on MIDM algorithm
The correlation between influence factor and load is calculated, is ranked up, filters out determinant attribute, construct the data of comprehensive index system
Analysis module;The inertial factor fixed in model can be converted to dynamic and become according to the random variation of different regions and time
The inertial factor of change constructs the multi objective Dynamic Matching module of A-GPR model;It can be according to the comprehensive index system and A- of building
The real-time electric power load prediction module of GPR model progress real-time electric power load prediction;The data acquisition module and data point
Analyse module by signal connection, data analysis module connect with multi objective Dynamic Matching module by signal, multi objective Dynamic Matching module and
The connection of real-time electric power load prediction module by signal.
A kind of real-time electric power load forecasting method based on multi objective Dynamic Matching, has including in:
1) the respective function of the meteorological data unit of data acquisition module, load data unit and economic data unit is utilized
It can be carried out data acquisition, temperature, the humidity, precipitation, visibility, wind in area needed for being responsible for acquisition in real time by meteorological data unit
To, wind speed, weather condition data carry out Data Integration, be transferred in system database generation meteorological data table;By load data
Unit is responsible for acquisition industrial electricity load, farming power load, municipal power load, post and telecommunications power load, traffic electricity consumption in real time
Load, household electricity load and commercial power load carry out Data Integration, are transferred in system database and generate load data
Table;Data Integration is carried out by the economic data that economic data unit is responsible in acquisition government's annual economic report in real time, is transferred to
Economic data table is generated in system database;
2) correlation analysis, analysis load number are carried out to each index using MIDM algorithm using the function of data analysis module
According to the degree of correlation between corresponding meteorological data, economic data, so that it is determined that the key index in area needed for influencing, wherein
The step of multi objective dynamical Matching Algorithm is:
(1) each index curve is calculated for the projector distance of load:
Wherein xqiRepresent projection abscissa of the point on influence index curve on load curve, kqiRepresent load curve
In the slope of this point, b1(2) abscissa put on influence index curve, b are represented1(1) represent influence index go to put upwards it is vertical
Coordinate, bqiLoad curve is represented in the intercept of this point.yqiThe point represented on influence index is vertical in the projection on load curve
Coordinate, q represent influence index 24 hours one day 24 points on curve, and i represents a certain influence index.rqiRepresent influence index
For the projector distance of load, xqi2Represent the abscissa of influence index curve subpoint the latter point on load curve, xqi1
Represent the abscissa of the previous point of influence index curve subpoint on load curve, yqi2Influence index curve is in load curve
The ordinate of subpoint the latter point, yqi1Influence index curve is represented in the ordinate of the previous point of load curve subpoint.
RqiLoad is represented in the fluctuation distance of point-to-point transmission, yt2Represent the ordinate of the latter point on load curve, yt1Represent load curve
The ordinate of upper previous point, xt2Represent the abscissa of the latter point on load curve, xt1Represent previous point on load curve
Ordinate.T represents the load 24 hours one day points on curve.
(2) weight of index is sought:
Wherein wiInfluence index is indicated for the weighing factor of load, i represents a certain influence index influence index, rqiIt represents
Projector distance of the influence factor for load, RtLoad is represented in the fluctuation distance of point-to-point transmission.
(3) weight calculation of aggregate index:
Wherein n represents the number of index, WijThe indirect weight of combined index is represented, i and j represent some influence index,
wi' weight that represents aggregate index, give full expression to influence directly or indirectly of each index for load;
3) function of utilizing multi objective Dynamic Matching module, according to the real-time of the load data of different regions and key index
Property correlation establish comprehensive dynamic indicator system;A-GPR model is established using selected load data and comprehensive index system,
Its step is:
(1) initialization population scale N, maximum number of iterations Tmax;
(2)The vector of particle current location;
(3) Fitness, the fitness value of vector x;
(4)The speed of particle, its dimension and vectorIt is identical;
(5)pbest, fitness value corresponding to the desired positions that encounter during each particle flight;
(6)For recording the desired positions encountered during particle flight, pbest its dimension and vectorDimension
It is identical;
(7) the more new formula of particle position:
xi=xi+vi
Wherein, viRepresent the speed of particle, xiThe current location of particle is represented, i represents the current iteration that each particle updates
Number, gbestiRepresent the optimal location of population entirety, pbestiRepresent optimal location individual in population, w represent inertia weight because
Son, φ1Represent individual cognition learning rate, φ2It represents social learning to lead, random number of the rand () between section [0,1], K generation
Table convergence factor, K can be described as:
The inertia weight factor is generally set in standard particle group's algorithm:wt+1=wt=1, since the inertia weight factor is shadow
The variable of current particle speed is rung, biggish value is conducive to global search, and lesser value is conducive to local search, in order to preferably
Search capability is balanced, a kind of A-GPR model is proposed:
Wherein:wtRepresent the inertia weight factor of each iteration, wmaxFor inertia maximum value, wminFor inertia minimum value, t is
Current iteration number, T are maximum number of iterations,
So the more new formula of particle position is:
xi=xi+vi
The speed for leading to particle since the inertia weight factor is changing always and position are always in a wide range of and small of population
It is being continuously updated always in range, to balance algorithm in the ability of global search and local search;
(8) according to the position of updated particle, the fitness function E of particle is recalculated, fitness function can describe
For:
Wherein, m is input sample number, yiIndicate the output of process value of current A-GPR model training, yi-1For prior-generation mould
The output valve of type, i represent current number of iterations;
(9) judge whether to meet schedule requirement or whether reach the number of iterations, if meeting termination search;Otherwise it carries out down
Primary search;
(10) information of final particle, the as parameter of A-GPR model are exported;
4) function of utilizing real-time electric power load prediction module, obtains meteorological data, the economic data at moment to be predicted, builds
Vertical comprehensive index system is input in A-GPR model using comprehensive index system and corresponding load data as input vector,
Its load value for exporting the as moment to be predicted.
The following describes the present invention in detail with reference to the accompanying drawings and specific embodiments.
The present invention collects the historical load data of the 2014-2017 of this area using somewhere actual electric network as embodiment,
And corresponding history meteorological data, wherein influence factor data include temperature, humidity, atmospheric pressure, rainfall, wind direction, wind speed,
Visibility, GDP.It is analyzed using multi objective dynamic matching method.
The related coefficient of table 1 load and each influence factor
Correlation analysis is carried out with data in 2016, the degree of correlation of each meteorologic factor is variation at all seasons,
It is maximum in spring weather condition and the degree of correlation of GDP, in the degree of correlation of summer rainfall and humidity maximum, wind speed and temperature in the fall
The degree of correlation of degree is maximum, and temperature and the degree of correlation of GDP are maximum in winter.Selection influence factor is carried out according to the date of pre- observation,
Establish comprehensive index system.
GPR prediction model is:
Wherein:Y ' represents the output of test sample, and y represents the output of training sample, and K (x*, X) is training input variable x
With test input variable X*The rank covariance function matrix of n × 1, K (X, X) be training input variable X n × n rank covariance letter
Matrix number, I are unit rank matrix, δn 2For the hyper parameter of model.
The difficult point for establishing model is the solution of model hyper parameter, and the hyper parameter of model is primarily present in covariance function
And in white noise;Therefore it is solving model hyper parameter, first has to the concrete form for determining covariance function.The association side of Gaussian process
The specific manifestation form of difference function is:
δijFor Kronecher constant, δp 2, l, δn 2For the hyper parameter of model, δp 2For the variance of the kernel function of model, l is spy
Width is levied, i, j represent a certain column of matrix.
For the parameter of Optimized model, the present invention proposes a kind of A-GPR model, and step is:
Step 1:Using variance as fitness, according to hyper parameter constraint condition, population information, scanning frequency like flying are initialized
Degree, current location etc..
Step 2:By hyper parameter speed and location information, the test sample of combined training sample and influence factor is input to
GPR regression model, according to formulaAnd
Rolling forecast is carried out to the test sample of load value, and calculates the fitness of particle.
Step 3:More new particle individual position optimal value and group position optimal value, and whether judge group's adaptive optimal control degree
It meets the requirements.If satisfied, then iteration terminates, 4 are otherwise gone to step.
Step 4:According to formulaAnd xi=xi+viMore
The speed and location information of new particle, go to step 2.
Population scale N=200, greatest iteration number T=150, w are setmax=0.9, wmin=0.4.It is after algorithm
Then the optimal information of model parameter is tested to establish A-GPR load forecasting model trained GPR model.
By establishing A-GPR real-time load prediction model, by the synthesis meteorologic factor of historical load data and prediction time
As improved A-GPR mode input amount, output quantity of the predicted load of prediction time as model chooses this area
On April 20th, 2016 is as prediction day.Using the data in January 1 to April 19 as training sample set, to A-GPR model into
Row training and test, table 2 are the predicted value of the load value that A-GPR model is predicted and GPR model and the error point of true value
Analyse table.
Each time point error analysis table of table 2
3 mean error analytical table of table
Classification | Mean error |
GPR | 5.7% |
A-GPR | 1.4% |
From table 2 and table 3 it can be seen that the real-time load prediction model based on multi objective Dynamic Matching has good prediction
Precision can obtain good prediction result.
The above is specific embodiment of the invention, but protection scope of the present invention is not limited to that, any
Person skilled in the art person can change or replace easily in the range of exposure of the invention, should all cover in guarantor of the invention
Within the scope of shield, therefore, protection scope of the present invention should be all subject to the protection scope in claims.
Claims (2)
1. a kind of real-time electric power load prediction system based on multi objective Dynamic Matching, characterized in that it includes:It is negative for acquiring
Lotus, meteorology, economic data simultaneously carry out Classification Management, construct the data acquisition module of system database;It is calculated based on MIDM algorithm
Correlation between influence factor and load, is ranked up, and filters out determinant attribute, constructs the data analysis of comprehensive index system
Module;The inertial factor fixed in model can be converted to dynamic change according to the random variation of different regions and time
Inertial factor constructs the multi objective Dynamic Matching module of A-GPR model;It can be according to the comprehensive index system and A-GPR of building
The real-time electric power load prediction module of model progress real-time electric power load prediction;The data acquisition module and data analyzes mould
Block signal connection, data analysis module connect with multi objective Dynamic Matching module by signal, multi objective Dynamic Matching module with it is real-time
The connection of load forecast module by signal.
2. a kind of real-time electric power load forecasting method based on multi objective Dynamic Matching, characterized in that it include in have:
1) using the meteorological data unit of data acquisition module, load data unit and economic data unit respective function into
Row data acquisition, by meteorological data unit be responsible in real time acquisition needed for area temperature, humidity, precipitation, visibility, wind direction,
Wind speed, weather condition data carry out Data Integration, are transferred to generation meteorological data table in system database;By load data unit
Be responsible in real time acquisition industrial electricity load, farming power load, municipal power load, post and telecommunications power load, traffic power load,
Household electricity load and commercial power load carry out Data Integration, are transferred in system database and generate loading data sheet;By
The economic data that economic data unit is responsible in acquisition government's annual economic report in real time carries out Data Integration, is transferred to system number
According to generation economic data table in library;
2) using the function of data analysis module using MIDM algorithm to each index carry out correlation analysis, analysis load data and
Degree of correlation between corresponding meteorological data, economic data, so that it is determined that the key index in area needed for influencing, wherein referring to more
Mark dynamical Matching Algorithm the step of be:
(1) each index curve is calculated for the projector distance of load:
Wherein xqiRepresent projection abscissa of the point on influence index curve on load curve, kqiRepresent load curve this point
Slope, b1(2) abscissa put on influence index curve, b are represented1(1) it represents influence index and removes the ordinate put upwards, bqi
Load curve is represented in the intercept of this point, yqiThe point on influence index is represented in the projection ordinate on load curve, q generation
Table influence index 24 hours one day 24 points on curve, i represent a certain influence index, rqiInfluence index is represented for load
Projector distance, xqi2Represent the abscissa of influence index curve subpoint the latter point on load curve, xqi1Representing influences
The abscissa of the previous point of index curve subpoint on load curve, yqi2Influence index curve is after load curve subpoint
The ordinate of one point, yqi1Influence index curve is represented in the ordinate of the previous point of load curve subpoint, RqiIt represents negative
Fluctuation distance of the lotus in point-to-point transmission, yt2Represent the ordinate of the latter point on load curve, yt1It represents previous on load curve
The ordinate of point, xt2Represent the abscissa of the latter point on load curve, xt1Represent the vertical seat of previous point on load curve
Mark, t represent the load 24 hours one day points on curve;
(2) weight of each index is sought:
Wherein wiEach influence factor is indicated for the weighing factor of load, i represents a certain influence index influence index, rqiIt represents
Projector distance of the influence factor for load, RtLoad is represented in the fluctuation distance of point-to-point transmission;
(3) weight calculation of aggregate index:
Wherein n represents the number of index, wi,jThe indirect weight of combined index is represented, i, j represent a certain item influence index, wi' generation
The weight of table aggregate index has given full expression to influence directly or indirectly of each index for load;
3) function of utilizing multi objective Dynamic Matching module, according to the real-time phase of the load data of different regions and key index
Closing property establishes comprehensive dynamic indicator system;A-GPR model is established using selected load data and comprehensive index system, is walked
Suddenly it is:
(1) initialization population scale N, maximum number of iterations Tmax;
(2)The vector of particle current location;
(3) Fitness, the fitness value of vector x;
(4)The speed of particle, its dimension and vectorIt is identical;
(5)pbest, fitness value corresponding to the desired positions that encounter during each particle flight;
(6)For recording the desired positions encountered during particle flight, pbest its dimension and vectorDimension phase
Together;
(7) the more new formula of particle position:
xi=xi+vi
Wherein, viRepresent the speed of particle, xiThe current location of particle is represented, i represents the current iteration number that each particle updates,
gbestiRepresent the optimal location of population entirety, pbestiOptimal location individual in population is represented, w represents the inertia weight factor, φ1
Represent individual cognition learning rate, φ2It represents social learning to lead, random number of the rand () between section [0,1], K represents convergence
The factor, K can be described as:
The inertia weight factor is generally set in standard particle group's algorithm:wt+1=wt=1, since the inertia weight factor is to influence currently
The variable of particle rapidity, biggish value are conducive to global search, and lesser value is conducive to local search, searches to be better balanced
Suo Nengli proposes a kind of A-GPR model:
Wherein:wtRepresent the inertia weight factor of current iteration number, wmaxFor inertia maximum value, wminFor inertia minimum value, t is
Current iteration number, T are maximum number of iterations,
So the more new formula of particle position is:
xi=xi+vi
The speed for leading to particle since the inertia weight factor is changing always and position are always in a wide range of and small range of population
It is inside being continuously updated always, to balance algorithm in the ability of global search and local search;
(8) according to the position of updated particle, the fitness function E of particle is recalculated, fitness function can be described as:
Wherein, m is input sample number, yiIndicate the output of process value of current A-GPR model training, yi-1For the defeated of prior-generation model
It is worth out, i represents current number of iterations;
(9) judge whether to meet schedule requirement or whether reach the number of iterations, if meeting termination search;Otherwise it carries out next time
Search;
(10) information of final particle, the as parameter of A-GPR model are exported;
4) function of utilizing real-time electric power load prediction module, obtains meteorological data, the economic data at moment to be predicted, establishes comprehensive
Index system is closed to be input in A-GPR model using comprehensive index system and corresponding load data as input vector, it is defeated
It is out the load value at moment to be predicted.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810525917.1A CN108830405B (en) | 2018-05-29 | 2018-05-29 | Real-time power load prediction system and method based on multi-index dynamic matching |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810525917.1A CN108830405B (en) | 2018-05-29 | 2018-05-29 | Real-time power load prediction system and method based on multi-index dynamic matching |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108830405A true CN108830405A (en) | 2018-11-16 |
CN108830405B CN108830405B (en) | 2021-11-30 |
Family
ID=64146506
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810525917.1A Active CN108830405B (en) | 2018-05-29 | 2018-05-29 | Real-time power load prediction system and method based on multi-index dynamic matching |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108830405B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111275240A (en) * | 2019-12-27 | 2020-06-12 | 华北电力大学 | Load prediction method based on multi-energy coupling scene |
CN111582619A (en) * | 2020-01-22 | 2020-08-25 | 汕头大学 | Adaptive design method based on correlation and dependency analysis |
CN113283659A (en) * | 2021-06-03 | 2021-08-20 | 上海分未信息科技有限公司 | Power load response task allocation method for virtual power plant |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102694800A (en) * | 2012-05-18 | 2012-09-26 | 华北电力大学 | Gaussian process regression method for predicting network security situation |
CN105701572A (en) * | 2016-01-13 | 2016-06-22 | 国网甘肃省电力公司电力科学研究院 | Photovoltaic short-term output prediction method based on improved Gaussian process regression |
CN106709261A (en) * | 2017-01-10 | 2017-05-24 | 辽宁工程技术大学 | Method for evaluating mine disaster |
CN106779129A (en) * | 2015-11-19 | 2017-05-31 | 华北电力大学(保定) | A kind of Short-Term Load Forecasting Method for considering meteorologic factor |
CN106971240A (en) * | 2017-03-16 | 2017-07-21 | 河海大学 | The short-term load forecasting method that a kind of variables choice is returned with Gaussian process |
CN106997495A (en) * | 2017-04-13 | 2017-08-01 | 云南电网有限责任公司电力科学研究院 | A kind of Methods of electric load forecasting |
-
2018
- 2018-05-29 CN CN201810525917.1A patent/CN108830405B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102694800A (en) * | 2012-05-18 | 2012-09-26 | 华北电力大学 | Gaussian process regression method for predicting network security situation |
CN106779129A (en) * | 2015-11-19 | 2017-05-31 | 华北电力大学(保定) | A kind of Short-Term Load Forecasting Method for considering meteorologic factor |
CN105701572A (en) * | 2016-01-13 | 2016-06-22 | 国网甘肃省电力公司电力科学研究院 | Photovoltaic short-term output prediction method based on improved Gaussian process regression |
CN106709261A (en) * | 2017-01-10 | 2017-05-24 | 辽宁工程技术大学 | Method for evaluating mine disaster |
CN106971240A (en) * | 2017-03-16 | 2017-07-21 | 河海大学 | The short-term load forecasting method that a kind of variables choice is returned with Gaussian process |
CN106997495A (en) * | 2017-04-13 | 2017-08-01 | 云南电网有限责任公司电力科学研究院 | A kind of Methods of electric load forecasting |
Non-Patent Citations (3)
Title |
---|
DIAOKENG7360: "粒子群优化算法简介", 《HTTPS://BLOG.CSDN.NET/DIAOKENG7360/ARTICLE/DETAILS/102222272》 * |
李莉 等: "数据挖掘技术用于负荷与负荷影响因素的相关性分析", 《华北电力大学学报》 * |
梁智 等: "基于变量选择与高斯过程回归的短期负荷预测", 《电力建设》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111275240A (en) * | 2019-12-27 | 2020-06-12 | 华北电力大学 | Load prediction method based on multi-energy coupling scene |
CN111275240B (en) * | 2019-12-27 | 2023-06-09 | 华北电力大学 | Load prediction method based on multi-energy coupling scene |
CN111582619A (en) * | 2020-01-22 | 2020-08-25 | 汕头大学 | Adaptive design method based on correlation and dependency analysis |
CN111582619B (en) * | 2020-01-22 | 2023-09-26 | 汕头大学 | Adaptive design method based on correlation and dependency analysis |
CN113283659A (en) * | 2021-06-03 | 2021-08-20 | 上海分未信息科技有限公司 | Power load response task allocation method for virtual power plant |
CN113283659B (en) * | 2021-06-03 | 2022-11-22 | 上海分未信息科技有限公司 | Power load response task allocation method for virtual power plant |
Also Published As
Publication number | Publication date |
---|---|
CN108830405B (en) | 2021-11-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108022001B (en) | Short-term load probability density prediction method based on PCA (principal component analysis) and quantile regression forest | |
CN106650767B (en) | Flood forecasting method based on cluster analysis and real-time correction | |
CN104156786B (en) | A kind of maximum daily load prediction system of the nonworkdays for considering meteorological multifactor impact | |
CN109164343A (en) | Quantify the Diagnosis Method of Transformer Faults with weighting KNN based on characteristic information | |
CN112288164B (en) | Wind power combined prediction method considering spatial correlation and correcting numerical weather forecast | |
CN104835103A (en) | Mobile network health evaluation method based on neural network and fuzzy comprehensive evaluation | |
CN105117602A (en) | Metering apparatus operation state early warning method | |
CN105373857A (en) | Photovoltaic power station irradiance prediction method | |
CN105701572B (en) | Photovoltaic short-term output prediction method based on improved Gaussian process regression | |
CN108830405A (en) | Real-time electric power load prediction system and method based on multi objective Dynamic Matching | |
CN105225006B (en) | A kind of short-term wind-electricity power nonparametric probability forecasting method | |
CN104408562A (en) | Photovoltaic system generating efficiency comprehensive evaluation method based on BP (back propagation) neural network | |
CN105868853A (en) | Method for predicting short-term wind power combination probability | |
CN108074004A (en) | A kind of GIS-Geographic Information System short-term load forecasting method based on gridding method | |
CN108280998A (en) | Short-time Traffic Flow Forecasting Methods based on historical data dynamic select | |
CN110288196A (en) | Transmission line galloping grading forewarning system method and system | |
CN110322075A (en) | A kind of scenic spot passenger flow forecast method and system based on hybrid optimization RBF neural | |
CN113554466A (en) | Short-term power consumption prediction model construction method, prediction method and device | |
CN106056235A (en) | Power transmission grid efficiency and benefit detection method based on Klee method and matter element extension model | |
CN109165807A (en) | A kind of energy quality comprehensive assessment method based on probabilistic neural network | |
CN110163444A (en) | A kind of water demand prediction method based on GASA-SVR | |
Bao-De et al. | Improved genetic algorithm-based research on optimization of least square support vector machines: an application of load forecasting | |
CN112288157A (en) | Wind power plant power prediction method based on fuzzy clustering and deep reinforcement learning | |
CN117313795A (en) | Intelligent building energy consumption prediction method based on improved DBO-LSTM | |
WO2023245399A1 (en) | Rice production potential simulation method based on land system and climate change coupling |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |