CN113011630A - Method for short-term prediction of space load in zone time of big data power distribution network - Google Patents

Method for short-term prediction of space load in zone time of big data power distribution network Download PDF

Info

Publication number
CN113011630A
CN113011630A CN202110095305.5A CN202110095305A CN113011630A CN 113011630 A CN113011630 A CN 113011630A CN 202110095305 A CN202110095305 A CN 202110095305A CN 113011630 A CN113011630 A CN 113011630A
Authority
CN
China
Prior art keywords
load
block
loads
time
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
Application number
CN202110095305.5A
Other languages
Chinese (zh)
Other versions
CN113011630B (en
Inventor
金正军
王伟福
张伟峰
吴舜裕
申鹂
孙微庭
茅奕晟
李莹莹
高琼
吴震
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hangzhou Power Supply Co of State Grid Zhejiang Electric Power Co Ltd
Original Assignee
Hangzhou Power Supply Co of State Grid Zhejiang Electric Power Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Hangzhou Power Supply Co of State Grid Zhejiang Electric Power Co Ltd filed Critical Hangzhou Power Supply Co of State Grid Zhejiang Electric Power Co Ltd
Priority to CN202110095305.5A priority Critical patent/CN113011630B/en
Publication of CN113011630A publication Critical patent/CN113011630A/en
Application granted granted Critical
Publication of CN113011630B publication Critical patent/CN113011630B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • 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
    • 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
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Business, Economics & Management (AREA)
  • Mathematical Physics (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Analysis (AREA)
  • Computational Mathematics (AREA)
  • Economics (AREA)
  • General Engineering & Computer Science (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • Algebra (AREA)
  • Databases & Information Systems (AREA)
  • Evolutionary Biology (AREA)
  • Software Systems (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Water Supply & Treatment (AREA)
  • Artificial Intelligence (AREA)
  • Primary Health Care (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computing Systems (AREA)
  • Evolutionary Computation (AREA)
  • Public Health (AREA)
  • General Health & Medical Sciences (AREA)
  • Quality & Reliability (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Development Economics (AREA)
  • Probability & Statistics with Applications (AREA)

Abstract

The invention belongs to the technical field of electric power, and particularly relates to a short-term prediction method for space-time load of a large data power distribution network region, which comprises the following steps: acquiring all district load geographical range parameters, urban traffic network topological structures, district load internal electricity utilization type information and block load historical electricity utilization level data in an area; blocking and classifying the platform load to form block loads after the platform load is subdivided; frequency division grading is carried out on the load of the transformer area, and relevance and hysteresis among different frequency characteristic powers of different block loads are calculated; calculating the relevance between the block load and the objective factors; calculating the relevance and time lag coefficient between the block load change trend and the social activity state; training the deep learning model to obtain a block load prediction model; and carrying out model training on the platform load prediction deep learning model to obtain a platform load short-term prediction model, and finally realizing platform load short-term prediction. The invention improves the load prediction accuracy.

Description

Method for short-term prediction of space load in zone time of big data power distribution network
Technical Field
The invention belongs to the technical field of electric power, and particularly relates to a short-term prediction method for space-time load of a large data power distribution network platform area.
Background
The short-term prediction of the space load refers to the load prediction of one day to seven days in the future for the load in a certain area by analyzing the load form and the trend characteristics and considering the correlation characteristics of the load in time and space in a certain geographic range. The station area space load value means that loads in an area are further divided into a plurality of station areas according to a power grid topological structure. The short-term prediction of the space-time load of the transformer area can effectively help power grid operators to make a heliostat operation mode, a load transfer scheme and a distribution network operation state topological structure, and has great significance for improving the power supply safety of a power grid.
The existing platform load prediction generally regards the platform load in an area as an independent load group, and the influence of objective environmental factors on the load of a specific platform is considered, so that intelligent algorithms such as a neural network and an extreme learning machine are adopted for prediction. The load change trend of the actual transformer area has an inseparable incidence relation with external environmental factors, social states, other types of load changes and the like, and different transformer areas have obvious time and space incidence characteristics. Meanwhile, the load of different blocks in the distribution area is not subdivided by the existing distribution area load prediction method, so that correlation of relevant factors is performed by using an artificial intelligence algorithm and the precision is insufficient in the learning process, and the final load prediction accuracy is influenced. In addition, the traditional platform load prediction based on space load is usually load prediction based on classification and subdivision of load types or characteristics, and time and space lag or coupling characteristics among multiple loads are not considered, so that the adaptability to special situations is poor.
Disclosure of Invention
In order to solve the technical problems, the invention provides a short-term prediction method for the space load of a large data power distribution network platform area, which realizes accurate prediction of the platform area load by considering weather, social state and multi-dimensional coupling correlation among different loads, effectively enhances the adaptability of the platform area load prediction method to the prediction of daily social or crowd activity conditions, and improves the load prediction accuracy.
A method for short-term prediction of space load in a cell of a big data power distribution network comprises the following steps:
s1: acquiring all district load geographical range parameters, urban traffic network topological structures, district load internal electricity utilization type information and block load historical electricity utilization level data in an area;
s2: according to the platform area load geographical range parameter, the urban traffic network topological structure, the power utilization type information in the platform area load and the block load historical power utilization level data, the platform area load is divided into blocks and classified to form block loads after the platform area load is subdivided;
s3: frequency division grading is carried out on the load of the transformer area, and relevance and hysteresis among different frequency characteristic powers of different block loads are calculated;
s4: calculating the relevance between the block load and the objective factors;
s5: calculating the relevance and time lag coefficient between the block load change trend and the social activity state;
s6: taking the relevance and the hysteresis among different block loads and different frequency characteristic powers, the relevance among the block loads and objective factors, the relevance among the block load variation trend and the social activity state and a time lag coefficient as input, and training a deep learning model to obtain a block load prediction model;
s7: and taking the block load prediction result as input, performing model training on the platform load prediction deep learning model to obtain a platform load short-term prediction model, and finally realizing platform load short-term prediction.
Preferably, the acquiring geographic range parameters of loads of all the distribution areas in the area, the topological structure of the urban traffic network, the information of the types of power consumption in the loads of the distribution areas, and the historical power consumption level data of the block loads includes:
s11: obtaining platform area load geographical range parameter
Figure BDA0002913974130000021
Wherein m is1Numbering the cells, M1The number of the transformer areas;
s12: according to the topological structure of the urban traffic network, the load of the transformer area is further divided into M2A block load range;
s13: obtaining M2Information of power consumption type in load of individual transformer area
Figure BDA0002913974130000031
Judgment M2Whether the users in the block load belong to the same power utilization type, if so, the original block load division is kept unchanged, and the number M of the block loads3＝M2(ii) a If not, further pair M according to the type of electricity utilization2Dividing the block load into M3A block load;
s14: obtaining M3Historical electricity usage level data for individual block load
Figure BDA0002913974130000032
Figure BDA0002913974130000033
Wherein: t is the Time point, Time isThe total duration.
Preferably, the blocking and classifying the platform loads according to the platform load geographical range parameter, the urban traffic network topology structure, the intra-platform load electricity utilization type information, and the block load historical electricity utilization level data, and forming the block loads after the platform loads are subdivided comprises:
s21: analyzing and calculating to obtain different block load trend change characteristic sequences
Figure BDA0002913974130000034
Figure BDA0002913974130000035
Wherein:
Figure BDA0002913974130000036
is m at3The load trend characteristics of the individual blocks,
Figure BDA0002913974130000037
is a block m3The number of chr _ C characteristic parameters is chr _ C;
s22: with qchrFor input, to M3And carrying out cluster analysis on the block loads to obtain K types of block load sets with similar electricity utilization trend characteristics.
Preferably, the frequency division and classification of the cell loads and the calculation of the correlation and hysteresis between different frequency characteristic powers of different cell loads include:
s31: to M3The block load electricity level data set P ═ { P ═ P1,...,pi,...,pM3The frequency domain load hierarchical decomposition is carried out to obtain a load decomposition matrix set which is divided into F load decomposition sequences
Figure BDA0002913974130000038
S32: taking t' as the length of a time window, and intercepting a slave time point tsStarting to a time point tsBlock i load resolution matrix data ending with + t' -1
Figure BDA0002913974130000039
To obtain
Figure BDA00029139741300000310
S33: will be provided with
Figure BDA00029139741300000311
With other M3-1 block load is respectively subjected to load sequence correlation analysis after time window translation is carried out for delta t time points, and a lag correlation factor vector of the block load i and other block loads j is obtained
Figure BDA00029139741300000312
And lag time coefficient vector
Figure BDA0002913974130000041
S34: step S32 and step S33 are iterated in a circulating mode, the lag correlation factor and the lag time coefficient between all the different block loads in the area are taken as the optimal correlation sequence, the result that the correlation value between all the different block loads is the maximum is taken, and a load lag correlation factor matrix D of the different block loads is formed1rAnd load lag time coefficient matrix D1t：
Figure BDA0002913974130000042
S35: set up D1rAnd D1tThe diagonal value of the matrix is 0, and the matrix D is judged1rAnd D1tWhether the middle block i and the block j belong to the same type of loads with the same power utilization trend characteristics or not, if so, setting the middle block i and the block j respectively
Figure BDA0002913974130000043
If not, then
Figure BDA0002913974130000044
And
Figure BDA0002913974130000045
the value of (a) is not changed.
Preferably, the calculating the correlation between the block load and the objective factor includes:
s41: acquiring forecast regional meteorological data W ═ W1,...,wh,...,wH]And Date type Date is as Date1,...,datedt,...dateDT]Classifying the block loads by date types, wherein DT dates are shared;
s42: respectively carrying out correlation analysis on the meteorological data, the date type data and all block loads to obtain a meteorological incidence matrix D of each block load2Matrix associated with date type D3Wherein:
Figure BDA0002913974130000046
preferably, the calculating the correlation between the block load change trend and the social activity state and the time lag coefficient includes:
s51: calculating load data p of block iiLoad change rate p ″)iObtaining a load change trend matrix P';
s52: obtaining a set of social status data near all block load geographic locations
Figure BDA0002913974130000051
si＝[si_1,...,si_c,...,si_C]，
Figure BDA0002913974130000052
Wherein: s is a set of social status data related to all load blocks, SiIs the time sequence of all the relevant social status data of the block i, C is the serial number of the social status data type, C is the total number of the social status data types,
Figure BDA0002913974130000053
the data value of the load block i, the class c social state at the time point t is obtained;
s53: to be provided witht' is the time window length, and is intercepted from the time point tsStarting to a time point tsThe block i load change trend matrix P 'ending at + t' -1 is obtained
Figure BDA0002913974130000054
S54: taking t' as the time window length, intercepting the vicinity of the block load i from the time point tsStarting to a time point ts+ t' -1 ending social status Window data
Figure BDA0002913974130000055
S55: at a time tsTaking deltat as a time window translation span as a starting point, and circularly taking different tsAfter the value is reached, will
Figure BDA0002913974130000056
Time window data collection with social status
Figure BDA0002913974130000057
Performing correlation analysis to obtain tsTo tsThe lag correlation factor and the lag time coefficient between the block load change trend of the + t' -1 time period and the social state form a social state correlation coupling matrix D4，
Figure BDA0002913974130000058
Wherein:
Figure BDA0002913974130000059
the hysteresis correlation factor of the load change trend of the block i to the c-th social activity state near the block j,
Figure BDA00029139741300000510
a lag time coefficient of the load change trend of the block i to the c-th social activity state near the block j;
s56: establishing a block load node distance association matrix Dis:
Figure BDA00029139741300000511
wherein: lijIs the geographic distance between tile i and tile j, lii＝0。
The technical scheme adopted by the invention has the following beneficial effects: acquiring all district load geographical range parameters, urban traffic network topological structures, district load internal electricity utilization type information and block load historical electricity utilization level data in an area; according to the platform area load geographical range parameter, the urban traffic network topological structure, the power utilization type information in the platform area load and the block load historical power utilization level data, the platform area load is divided into blocks and classified to form block loads after the platform area load is subdivided; frequency division grading is carried out on the load of the transformer area, and relevance and hysteresis among different frequency characteristic powers of different block loads are calculated; calculating the relevance between the block load and the objective factors; calculating the relevance and time lag coefficient between the block load change trend and the social activity state; taking the relevance and the hysteresis among different block loads and different frequency characteristic powers, the relevance among the block loads and objective factors, the relevance among the block load variation trend and the social activity state and a time lag coefficient as input, and training a deep learning model to obtain a block load prediction model; and taking the block load prediction result as input, performing model training on the platform load prediction deep learning model to obtain a platform load short-term prediction model, and finally realizing platform load short-term prediction. The invention can effectively identify the load transfer characteristics among different types and different geographic positions and realize the correlation prediction of power change among multiple types of power utilization blocks. Meanwhile, by considering weather, social states and multi-dimensional coupling correlation among different loads, accurate prediction of the platform area load is achieved, adaptability of the platform area load prediction method to prediction of daily society or crowd activity conditions is effectively enhanced, and load prediction accuracy is improved.
The following detailed description of the present invention will be provided in conjunction with the accompanying drawings.
Drawings
The invention is further described with reference to the accompanying drawings and the detailed description below:
fig. 1 is a schematic flow chart of a short-term prediction method of space load in a zone of a big data power distribution network according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, a method for short-term prediction of space load in a zone of a big data power distribution network includes the following steps:
s1: and acquiring all district load geographical range parameters, urban traffic network topological structures, district load internal electricity utilization type information and block load historical electricity utilization level data in the area.
Obtaining platform area load geographical range parameter
Figure BDA0002913974130000071
Wherein m is1Numbering the cells, M1The number of the distribution areas.
According to the topological structure of the urban traffic network, the load of the transformer area is further divided into M2A block load range.
Obtaining M2The load internal electricity utilization type information of each transformer area is divided into: 4 types of general residents, industry and commerce and others are classified and marked for the load in the step 2
Figure BDA0002913974130000072
Judgment M2Whether the users in the block loads belong to the same power utilization type. If it is notIf so, the original block load division is kept unchanged, and the block load number M is kept unchanged3＝M2(ii) a If not, further pair M according to the type of electricity utilization2Dividing the block load into M3Individual block load. Namely: m3≥M2。
Obtaining M3Data of power consumption level of each block load in near one year
Figure BDA0002913974130000073
Figure BDA0002913974130000074
Wherein: t is the Time point and Time is the total duration. Setting the load sequence as 15 minutes for one point, the load has 4 times 24 times 365 times 35040 Time points in the last year, namely Time 35040.
S2: according to the district load geographic range parameter, the urban traffic network topological structure, the electricity utilization type information in the district load and the block load historical electricity utilization level data, the district load is partitioned and classified to form the block load after the district load is subdivided.
Different block loads are obtained through analysis and calculation: the average load change rate, variance, peak-valley difference, highest power, lowest power, average power, time point of highest power and time point of lowest power are 8 load characteristic parameters to form a load trend change characteristic sequence
Figure BDA0002913974130000081
Wherein:
Figure BDA0002913974130000082
is m at3The load trend characteristics of the individual blocks,
Figure BDA0002913974130000083
is a block m3Chr _ C-th characteristic parameter, chr _ C-8
With qchrFor input, adopting a neighbor propagation clustering algorithm to pair M3And carrying out cluster analysis on the block loads to obtain K types of block load sets with similar electricity utilization trend characteristics.
S3: and (4) carrying out frequency division grading on the load of the cell, and calculating the relevance and the hysteresis between different frequency characteristic powers of different block loads.
Using Fourier frequency domain decomposition method to M3Individual block load electricity level data set
Figure BDA0002913974130000084
Carrying out frequency domain load hierarchical decomposition, dividing the load curve into 3 (F is 3) load decomposition sequences to obtain a load decomposition matrix set
Figure BDA0002913974130000085
Wherein the load decomposition matrix of a single block is pi′。
Figure BDA0002913974130000086
With 2 hours (i.e., t' is 8) as the time window length, the intercept is from the time point tsStarting to a time point ts+ t' -1 ending block i (i ∈ [1, M)3]) Load decomposing matrix data
Figure BDA0002913974130000087
To obtain
Figure BDA0002913974130000088
Figure BDA0002913974130000089
Will be provided with
Figure BDA00029139741300000810
With other M3-1 block load is subjected to load sequence correlation analysis after Time window shift Δ t ═ 2 Time points, totally (Time-t' +1) × (M)3-1) secondary correlation analysis. Obtaining the block load i and other block loads j (0)<j<M3) Vector of lag correlation factors
Figure BDA00029139741300000811
And lag time coefficient vector
Figure BDA00029139741300000812
And circularly iterating the steps to obtain the lag correlation factor and the lag time coefficient between all the two different block loads in the area. Taking the maximum correlation value result of all different block loads as the optimal correlation sequence to form a load lag correlation factor matrix D of different block loads1rAnd load lag time coefficient matrix D1t。
Figure BDA0002913974130000091
Figure BDA0002913974130000092
Set up D1rAnd D1tThe diagonal of the matrix has a value of 0. Judgment matrix D1rAnd D1tWhether the middle blocks i and j belong to the same type of loads (K types in total) with the same power utilization trend characteristics. If yes, respectively setting
Figure BDA0002913974130000093
If not, then
Figure BDA0002913974130000094
And
Figure BDA0002913974130000095
the value of (a) is not changed.
S4: and calculating the relevance between the block load and the objective factors.
Acquiring 4 meteorological data of air temperature, humidity, wind speed and air pressure in a prediction region to form a meteorological data sequence W ═ W1,...,wh,...,wH]And H is 4. The date types are divided into: the working day, weekend, national day, holiday are 3 types, and the Date type Date is Date ═ Date1,...,datedt,...dateDT]The block load is classified by date type, DT is 3.
Respectively carrying out correlation analysis on the meteorological data, the date type data and all block loads to obtain a meteorological incidence matrix D of each block load2Matrix associated with date type D3. Wherein:
Figure BDA0002913974130000096
Figure BDA0002913974130000097
s5: and calculating the relevance and the time lag coefficient between the block load change trend and the social activity state.
Calculating load data p of block iiLoad change rate p ″)iAnd obtaining a load change trend matrix P'. Wherein:
Figure BDA0002913974130000098
Figure BDA0002913974130000101
acquiring 2 social state data sets of load rate and road condition and traffic situation of wireless mobile base stations near geographic positions of loads of all blocks
Figure BDA0002913974130000102
si＝[si_1,si_2,]，
Figure BDA0002913974130000103
Wherein: s is a set of social status data related to all load blocks, SiIs the time sequence of all the relevant social status data of the block i, C is the serial number of the social status data type, C is 2, which is the total number of the social status data types,
Figure BDA0002913974130000104
the data value of the c-th social status of the load block i at the time point t is shown.
Taking t' as 8 as the length of the time window, and intercepting the time point tsStarting to a time point ts+7 ending block i (i ∈ [1, M)3]) The load change trend matrix P' is obtained
Figure BDA0002913974130000105
Taking t' as the time window length, intercepting the vicinity of the block load i from the time point tsStarting to a time point ts+ t' -1 ending social status Window data
Figure BDA0002913974130000106
Figure BDA0002913974130000107
At a time tsTaking deltat as a time window translation span as a starting point, and circularly taking different ts(ts∈[1,Time-7]) After the value is reached, will
Figure BDA0002913974130000108
Time window data collection with social status
Figure BDA0002913974130000109
And (5) performing correlation analysis. To obtain tsTo tsThe lag correlation factor and the lag time coefficient between the block load change trend of the +7 time period and the social state form a social state correlation coupling matrix D4，
Figure BDA00029139741300001010
Wherein:
Figure BDA00029139741300001011
the hysteresis correlation factor of the load change trend of the block i to the c-th social activity state near the block j,
Figure BDA00029139741300001012
the lag time coefficient of the load change trend of the block i to the c-th social activity state near the block j is shown.
Figure BDA0002913974130000111
And establishing a block load node distance association matrix Dis.
Figure BDA0002913974130000112
Wherein: lijIs the geographic distance between tile i and tile j, lii＝0。
S6: and training the deep learning model to obtain a block load prediction model by taking the relevance and the hysteresis among different block loads and different frequency characteristic powers, the relevance among the block loads and objective factors, the relevance among the block load variation trend and the social activity state and a time hysteresis coefficient as input.
Will: (1) a block load matrix P, (2) a load decomposition matrix set P' and (3) a load lag correlation factor matrix D1rAnd (4) a load lag time coefficient matrix D1tMeteorological data W (5) and meteorological correlation matrix D (6)2Date type Date (7) and Date type association matrix D (8)3(9) load change trend matrix P' and (10) social state data set
Figure BDA0002913974130000113
(11) Social status associative coupling matrix D4And (12) learning 12 data sets serving as deep learning training sample data sets in the block load node distance incidence matrix Dis to obtain a distribution network block space-time load short-term prediction model MDqk。
S7: and taking the block load prediction result as input, performing model training on the platform load prediction deep learning model to obtain a platform load short-term prediction model, and finally realizing platform load short-term prediction.
Constructing a load membership matrix between block loads and platform loads
Figure BDA0002913974130000114
Wherein: num is the number of the platform area; innumiInclusion Boolean value, in, for block i to region numnumi0 means that block i does not belong to the station area num, innumi1 indicates that block i belongs to the station area num.
Model to be predicted MDqkOutput block load predicted value and platform load level PtqDistance correlation matrix Dis of block load nodes and load membership matrix RloadPutting the deep learning model into the power distribution network region space-time load short-term prediction model MDtq。
The method improves and obtains all district load geographical range parameters, urban traffic network topological structure, district load electricity type information and block load historical electricity level data; according to the platform area load geographical range parameter, the urban traffic network topological structure, the power utilization type information in the platform area load and the block load historical power utilization level data, the platform area load is divided into blocks and classified to form block loads after the platform area load is subdivided; frequency division grading is carried out on the load of the transformer area, and relevance and hysteresis among different frequency characteristic powers of different block loads are calculated; calculating the relevance between the block load and the objective factors; calculating the relevance and time lag coefficient between the block load change trend and the social activity state; taking the relevance and the hysteresis among different block loads and different frequency characteristic powers, the relevance among the block loads and objective factors, the relevance among the block load variation trend and the social activity state and a time lag coefficient as input, and training a deep learning model to obtain a block load prediction model; and taking the block load prediction result as input, performing model training on the platform load prediction deep learning model to obtain a platform load short-term prediction model, and finally realizing platform load short-term prediction. The invention can effectively identify the load transfer characteristics among different types and different geographic positions and realize the correlation prediction of power change among multiple types of power utilization blocks. Meanwhile, by considering weather, social states and multi-dimensional coupling correlation among different loads, accurate prediction of the platform area load is achieved, adaptability of the platform area load prediction method to prediction of daily society or crowd activity conditions is effectively enhanced, and load prediction accuracy is improved.
While the invention has been described with reference to specific embodiments, it will be understood by those skilled in the art that the invention is not limited thereto, and may be embodied in other forms without departing from the spirit or essential characteristics thereof. Any modification which does not depart from the functional and structural principles of the present invention is intended to be included within the scope of the claims.

Claims (6)

1. A method for short-term prediction of space load in a cell of a big data power distribution network is characterized by comprising the following steps:
s1: acquiring all district load geographical range parameters, urban traffic network topological structures, district load internal electricity utilization type information and block load historical electricity utilization level data in an area;
s2: according to the platform area load geographical range parameter, the urban traffic network topological structure, the power utilization type information in the platform area load and the block load historical power utilization level data, the platform area load is divided into blocks and classified to form block loads after the platform area load is subdivided;
s3: frequency division grading is carried out on the load of the transformer area, and relevance and hysteresis among different frequency characteristic powers of different block loads are calculated;
s4: calculating the relevance between the block load and the objective factors;
s5: calculating the relevance and time lag coefficient between the block load change trend and the social activity state;
s6: taking the relevance and the hysteresis among different block loads and different frequency characteristic powers, the relevance among the block loads and objective factors, the relevance among the block load variation trend and the social activity state and a time lag coefficient as input, and training a deep learning model to obtain a block load prediction model;
s7: and taking the block load prediction result as input, performing model training on the platform load prediction deep learning model to obtain a platform load short-term prediction model, and finally realizing platform load short-term prediction.
2. The method for short-term prediction of the space load of the district of the big data power distribution network according to claim 1, wherein the obtaining of the geographical range parameters of the load of all districts, the topological structure of the urban traffic network, the information of the type of the power consumption in the district load, and the historical power consumption level data of the block load comprises:
s11: obtaining platform area load geographical range parameter
Figure FDA0002913974120000011
Wherein m is1Numbering the cells, M1The number of the transformer areas;
s12: according to the topological structure of the urban traffic network, the load of the transformer area is further divided into M2A block load range;
s13: obtaining M2Information of power consumption type in load of individual transformer area
Figure FDA0002913974120000012
Judgment M2Whether the users in the block load belong to the same power utilization type, if so, the original block load division is kept unchanged, and the number M of the block loads3＝M2(ii) a If not, further pair M according to the type of electricity utilization2Dividing the block load into M3A block load;
s14: obtaining M3Historical electricity usage level data for individual block load
Figure FDA0002913974120000021
Figure FDA0002913974120000022
Wherein: t is the Time point and Time is the total duration.
3. The method for short-term prediction of the platform area air load of the big data power distribution network according to claim 1, wherein the step of partitioning and classifying the platform area load according to the platform area load geographic range parameter, the urban traffic network topology structure, the platform area load internal electricity type information and the block load historical electricity level data to form the block load after the platform area load is subdivided comprises the following steps:
s21: analyzing and calculating to obtain different block load trend change characteristic sequences
Figure FDA0002913974120000023
Figure FDA0002913974120000024
Wherein:
Figure FDA0002913974120000025
is m at3The load trend characteristics of the individual blocks,
Figure FDA0002913974120000026
is a block m3The number of chr _ C characteristic parameters is chr _ C;
s22: with qchrFor input, to M3And carrying out cluster analysis on the block loads to obtain K types of block load sets with similar electricity utilization trend characteristics.
4. The method of claim 1, wherein the frequency division classification of the cell load and the calculation of the correlation and hysteresis between different frequency characteristic powers of different cell loads comprise:
s31: to M3Individual block load electricity level data set
Figure FDA00029139741200000212
Carrying out frequency domain load grading decomposition to obtain load divided into F load decomposition sequencesSet of decomposition matrices
Figure FDA0002913974120000027
S32: taking t' as the length of a time window, and intercepting a slave time point tsStarting to a time point tsBlock i load resolution matrix data ending with + t' -1
Figure FDA0002913974120000028
To obtain
Figure FDA0002913974120000029
S33: will be provided with
Figure FDA00029139741200000210
With other M3-1 block load is respectively subjected to load sequence correlation analysis after time window translation is carried out for delta t time points, and a lag correlation factor vector of the block load i and other block loads j is obtained
Figure FDA00029139741200000211
And lag time coefficient vector
Figure FDA0002913974120000031
S34: step S32 and step S33 are iterated in a circulating mode, the lag correlation factor and the lag time coefficient between all the different block loads in the area are taken as the optimal correlation sequence, the result that the correlation value between all the different block loads is the maximum is taken, and a load lag correlation factor matrix D of the different block loads is formed1rAnd load lag time coefficient matrix D1t：
Figure FDA0002913974120000032
Figure FDA0002913974120000033
S35: set up D1rAnd D1tThe diagonal value of the matrix is 0, and the matrix D is judged1rAnd D1tWhether the middle block i and the block j belong to the same type of loads with the same power utilization trend characteristics or not, if so, setting the middle block i and the block j respectively
Figure FDA0002913974120000034
If not, then
Figure FDA0002913974120000035
And
Figure FDA0002913974120000036
the value of (a) is not changed.
5. The method according to claim 1, wherein the calculating the correlation between the block load and the objective factor comprises:
s41: acquiring forecast regional meteorological data W ═ W1,...,wh,...,wH]And Date type Date is as Date1,...,datedt,...dateDT]Classifying the block loads by date types, wherein DT dates are shared;
s42: respectively carrying out correlation analysis on the meteorological data, the date type data and all block loads to obtain a meteorological incidence matrix D of each block load2Matrix associated with date type D3Wherein:
Figure FDA0002913974120000037
6. the big data power distribution network bay area space load short-term prediction method as claimed in claim 1, wherein said calculating the correlation between block load variation trend and social activity state and time lag coefficient comprises:
s51: calculation areaLoad data p of block iiLoad change rate p ″)iObtaining a load change trend matrix P';
s52: obtaining a set of social status data near all block load geographic locations
Figure FDA0002913974120000041
si＝[si_1,...,si_c,...,si_C]，
Figure FDA0002913974120000042
Wherein: s is a set of social status data related to all load blocks, SiIs the time sequence of all the relevant social status data of the block i, C is the serial number of the social status data type, C is the total number of the social status data types,
Figure FDA0002913974120000043
the data value of the load block i, the class c social state at the time point t is obtained;
s53: taking t' as the length of a time window, and intercepting a slave time point tsStarting to a time point tsThe block i load change trend matrix P 'ending at + t' -1 is obtained
Figure FDA0002913974120000044
S54: taking t' as the time window length, intercepting the vicinity of the block load i from the time point tsStarting to a time point ts+ t' -1 ending social status Window data
Figure FDA0002913974120000045
S55: at a time tsTaking deltat as a time window translation span as a starting point, and circularly taking different tsAfter the value is reached, will
Figure FDA0002913974120000046
Time window data collection with social status
Figure FDA0002913974120000047
Performing correlation analysis to obtain tsTo tsThe lag correlation factor and the lag time coefficient between the block load change trend of the + t' -1 time period and the social state form a social state correlation coupling matrix D4，
Figure FDA0002913974120000048
Wherein:
Figure FDA0002913974120000049
the hysteresis correlation factor of the load change trend of the block i to the c-th social activity state near the block j,
Figure FDA00029139741200000410
a lag time coefficient of the load change trend of the block i to the c-th social activity state near the block j;
s56: establishing a block load node distance association matrix Dis:
Figure FDA00029139741200000411
wherein: lijIs the geographic distance between tile i and tile j, lii＝0。
CN202110095305.5A 2021-01-25 2021-01-25 Short-term prediction method for space-time load of big data distribution network area Active CN113011630B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110095305.5A CN113011630B (en) 2021-01-25 2021-01-25 Short-term prediction method for space-time load of big data distribution network area

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110095305.5A CN113011630B (en) 2021-01-25 2021-01-25 Short-term prediction method for space-time load of big data distribution network area

Publications (2)

Publication Number Publication Date
CN113011630A true CN113011630A (en) 2021-06-22
CN113011630B CN113011630B (en) 2024-01-23

Family

ID=76385373

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110095305.5A Active CN113011630B (en) 2021-01-25 2021-01-25 Short-term prediction method for space-time load of big data distribution network area

Country Status (1)

Country Link
CN (1) CN113011630B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113592185A (en) * 2021-08-05 2021-11-02 四川大学 Power load prediction method based on Transformer
CN114707772A (en) * 2022-06-06 2022-07-05 山东大学 Power load prediction method and system based on multi-feature decomposition and fusion
CN117113159A (en) * 2023-10-23 2023-11-24 国网山西省电力公司营销服务中心 Deep learning-based power consumer side load classification method and system

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2009089594A (en) * 2007-09-28 2009-04-23 Kankoku Denryoku Kosha Temporal-spatial load analysis system of power facility utilizing inspection data and calculation method of load
JP2010204974A (en) * 2009-03-04 2010-09-16 Oki Electric Ind Co Ltd Time series data prediction device
CN105844371A (en) * 2016-05-19 2016-08-10 北京中电普华信息技术有限公司 Electricity customer short-term load demand forecasting method and device
JP2018063574A (en) * 2016-10-13 2018-04-19 アズビル株式会社 Trend predicting device and load predicting device
US20180351355A1 (en) * 2016-11-02 2018-12-06 China Electric Power Research Institute Company Limited Method for identifying pattern of load cycle
CN109376972A (en) * 2019-01-01 2019-02-22 中南大学 A kind of wisdom Power Network Short-Term Electric Load Forecasting method based on block cluster
CN110009145A (en) * 2019-03-29 2019-07-12 山大地纬软件股份有限公司 Methods of electric load forecasting and device based on time-frequency conversion feature extraction and autoregression trend prediction technology
CN110298501A (en) * 2019-06-21 2019-10-01 河海大学常州校区 Electric load prediction technique based on long Memory Neural Networks in short-term
CN110674999A (en) * 2019-10-08 2020-01-10 国网河南省电力公司电力科学研究院 Cell load prediction method based on improved clustering and long-short term memory deep learning
US20230096258A1 (en) * 2019-06-21 2023-03-30 Siemens Aktiengesellschaft Power load data prediction method and device, and storage medium

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2009089594A (en) * 2007-09-28 2009-04-23 Kankoku Denryoku Kosha Temporal-spatial load analysis system of power facility utilizing inspection data and calculation method of load
JP2010204974A (en) * 2009-03-04 2010-09-16 Oki Electric Ind Co Ltd Time series data prediction device
CN105844371A (en) * 2016-05-19 2016-08-10 北京中电普华信息技术有限公司 Electricity customer short-term load demand forecasting method and device
JP2018063574A (en) * 2016-10-13 2018-04-19 アズビル株式会社 Trend predicting device and load predicting device
US20180351355A1 (en) * 2016-11-02 2018-12-06 China Electric Power Research Institute Company Limited Method for identifying pattern of load cycle
CN109376972A (en) * 2019-01-01 2019-02-22 中南大学 A kind of wisdom Power Network Short-Term Electric Load Forecasting method based on block cluster
CN110009145A (en) * 2019-03-29 2019-07-12 山大地纬软件股份有限公司 Methods of electric load forecasting and device based on time-frequency conversion feature extraction and autoregression trend prediction technology
CN110298501A (en) * 2019-06-21 2019-10-01 河海大学常州校区 Electric load prediction technique based on long Memory Neural Networks in short-term
US20230096258A1 (en) * 2019-06-21 2023-03-30 Siemens Aktiengesellschaft Power load data prediction method and device, and storage medium
CN110674999A (en) * 2019-10-08 2020-01-10 国网河南省电力公司电力科学研究院 Cell load prediction method based on improved clustering and long-short term memory deep learning

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
马晓东等: "基于区域划分的分布式短期负荷预测方案设计", 自动化与仪器仪表, no. 12 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113592185A (en) * 2021-08-05 2021-11-02 四川大学 Power load prediction method based on Transformer
CN113592185B (en) * 2021-08-05 2022-10-14 四川大学 Power load prediction method based on Transformer
CN114707772A (en) * 2022-06-06 2022-07-05 山东大学 Power load prediction method and system based on multi-feature decomposition and fusion
CN117113159A (en) * 2023-10-23 2023-11-24 国网山西省电力公司营销服务中心 Deep learning-based power consumer side load classification method and system

Also Published As

Publication number Publication date
CN113011630B (en) 2024-01-23

Similar Documents

Publication Publication Date Title
CN113011630A (en) Method for short-term prediction of space load in zone time of big data power distribution network
Hu et al. Federated region-learning: An edge computing based framework for urban environment sensing
CN110909912B (en) Park electric power system net load combination prediction method based on self-adaptive error feedback
CN107800140B (en) Large user power supply access decision method considering load characteristics
Li et al. Federated learning-based short-term building energy consumption prediction method for solving the data silos problem
CN108205717A (en) A kind of photovoltaic generation power Multiple Time Scales Forecasting Methodology
CN105590174A (en) Enterprise power consumption load prediction method based on K-means clustering RBF neural network
CN109978284B (en) Photovoltaic power generation power time-sharing prediction method based on hybrid neural network model
CN106022538B (en) Photovoltaic power generation prediction method based on K-means clustering improved generalized weather
CN105069519A (en) Intelligent power grid park terminal user energy demand condition dynamic prediction system and method
CN111461921B (en) Load modeling typical user database updating method based on machine learning
Zhao et al. Spatio-temporal analysis and forecasting of distributed PV systems diffusion: A case study of Shanghai using a data-driven approach
CN106600063B (en) Space load prediction method based on fuzzy rough set theory and spatio-temporal information
CN110533243A (en) A kind of building energy consumption prediction technique based on fractal theory
CN113516294B (en) Space-time load prediction method suitable for power grid planning based on city planning
CN110472770A (en) A kind of photovoltaic power generation power prediction method
CN110309978B (en) Neural network photovoltaic power prediction model and method based on secondary dynamic adjustment
CN109359671B (en) Classification intelligent extraction method for hydropower station reservoir dispatching rules
CN116485582A (en) Heat supply optimization regulation and control method and device based on deep learning
Salgado et al. Hybrid fuzzy clustering neural networks to wind power generation forecasting
Li et al. A gray rbf model improved by genetic algorithm for electrical power forecasting
Zhou et al. A data-driven approach to forecasting the distribution of distributed photovoltaic systems
CN105652658A (en) Photovoltaic electric field data analysis and real-time environmental monitoring-based active power control system
Zhou et al. The analysis and prediction of power consumption in industry based on machine learning
Li et al. A prediction method of charging station capacity based on deep learning

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