CN109376907B - High-voltage distribution network transformer substation load prediction method suitable for integrated planning of transmission and distribution network - Google Patents

High-voltage distribution network transformer substation load prediction method suitable for integrated planning of transmission and distribution network Download PDF

Info

Publication number
CN109376907B
CN109376907B CN201811115905.8A CN201811115905A CN109376907B CN 109376907 B CN109376907 B CN 109376907B CN 201811115905 A CN201811115905 A CN 201811115905A CN 109376907 B CN109376907 B CN 109376907B
Authority
CN
China
Prior art keywords
load
grid
distribution network
layer
calculation
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.)
Active
Application number
CN201811115905.8A
Other languages
Chinese (zh)
Other versions
CN109376907A (en
Inventor
王曦冉
来聪
何英静
李帆
沈舒仪
章敏捷
徐旸
谷纪亭
郁丹
蔡优悠
陈旭阳
牛威
周海波
施进平
但扬清
王婷婷
何东
冯伟
常安
李青
翁华
吴君
唐人
周林
刘林萍
吕韵
张代红
李春
胡哲晟
王思远
孙擎宇
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.)
State Grid Corp of China SGCC
Economic and Technological Research Institute of State Grid Zhejiang Electric Power Co Ltd
Zhejiang Huayun Electric Power Engineering Design Consulting Co
Original Assignee
State Grid Corp of China SGCC
Economic and Technological Research Institute of State Grid Zhejiang Electric Power Co Ltd
Zhejiang Huayun Electric Power Engineering Design Consulting Co
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 State Grid Corp of China SGCC, Economic and Technological Research Institute of State Grid Zhejiang Electric Power Co Ltd, Zhejiang Huayun Electric Power Engineering Design Consulting Co filed Critical State Grid Corp of China SGCC
Priority to CN201811115905.8A priority Critical patent/CN109376907B/en
Publication of CN109376907A publication Critical patent/CN109376907A/en
Application granted granted Critical
Publication of CN109376907B publication Critical patent/CN109376907B/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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas 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

Abstract

The invention discloses a high-voltage distribution network transformer substation load prediction method suitable for integrated planning of a transmission network and a distribution network. At present, the prediction mode of the distribution network is difficult to adapt to the development condition of the whole city, and the load prediction of the transmission and distribution network is difficult to be performed through one method. The technical scheme of the invention comprises the following steps: the method is characterized in that the method is combined with the integrated planning demand characteristics of the transmission and distribution network, the high-voltage distribution network load prediction method combining the Elman neural network model and the space load prediction is adopted, the medium-long term load of the high-voltage distribution network is predicted by combining the historical load and the space prediction with the high-voltage distribution network substation supply area as a prediction unit, and a supply area load calculation model based on the grid development degree is introduced. The method integrates the advantages of perspective maximum load density data in space planning, and dynamically introduces measured data of the transformer substation in the past years into a calculation sample to correct the grid development degree; and after the grid development degree is obtained through prediction, the load condition of the high-voltage distribution network supply area is obtained through recalculation in consideration of the division change of the power transformation supply area in the next year.

Description

High-voltage distribution network transformer substation load prediction method suitable for integrated planning of transmission and distribution network
Technical Field
The invention relates to the field of load prediction of a transformer substation supply area, in particular to a high-voltage distribution network transformer substation load prediction method suitable for integrated planning of a transmission network and a distribution network.
Background
At present, methods for predicting power consumption mainly fall into two categories: a conventional prediction method and an intelligent prediction method.
The conventional prediction method mainly comprises the following steps: elastic coefficient method, regression analysis method, time series prediction method, yield value and consumption method and derivation method thereof. The intelligent prediction method mainly comprises a neural network, fuzzy logic, an expert system and the like. The methods all have strong pertinence, the load is difficult to reasonably divide by adopting a prediction mode suitable for the whole region, the development condition of the whole city is difficult to adapt by adopting a prediction mode of a power distribution network, and the load prediction of a transmission and distribution network is difficult to be penetrated by one method.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a prediction method which can meet the integrated planning requirements of a transmission and distribution network and aims at the supply area load of a transformer substation in a high-voltage distribution network.
The technical scheme adopted by the invention is as follows: a high-voltage distribution network transformer substation load prediction method suitable for integrated planning of a transmission network and a distribution network comprises the following steps:
The method is characterized in that the integrated planning demand characteristics of the transmission and distribution network are combined, a high-voltage distribution network load prediction method combining an Elman neural network model and space load prediction is adopted, a high-voltage distribution network substation supply area is taken as a prediction unit, medium and long-term loads of the high-voltage distribution network are predicted by combining historical loads and space prediction, and a supply area load calculation model based on grid development degree is introduced.
The invention introduces the concept of grid development degree, breaks through the traditional mode of adopting fixed load and grid corresponding, links up the historical load trend of the transformer substation supply area and the condition of refining grid load, and meets the requirement of transformer substation load prediction of a regional power distribution network.
As a supplement to the above technical solution, the specific contents of the supply area load calculation model based on the grid development degree are as follows:
the high-voltage distribution network is effectively linked with the load distribution of a higher-level voltage class through an operation mode, and is adapted to the load distribution prediction of a lower-level distribution network through space geographical division of a station supply area; is provided withThe ground city is divided into m grids according to municipal planning, and the total area is SmThe total number of k 110 KV substations in the range is equal to e in the supply area of the 110 KV substation in the ith year by taking the supply area of the 110 KV substation as an analysis areaiA vector set of grid block areas is S i＝{s(1i),s(2i),…,s(ei) }, there are
Figure BDA0001810563440000011
When calculating the load of grid block, only recording 10 KV and below large users, the maximum load density vector set is Pi max＝{pmax(1i),pmax(2i),…,pmax(ei) H, the vector set of the maximum volume rate inside the grid is betaimax＝{βmax(1i),βmax(2i),…,βmax(ei) }, introducing an area development degree vector gammai＝{γ(1i),γ(2i),…,γ(ei) Denotes the degree of load development within the grid, where γ (e)i)∈[0,1]Introduction of land utilization of gridi＝{θ(1i),θ(2i),…,θ(ei) Denotes the ratio of grid volume rate to target construction scale over a period of time, where θ (e)i)∈[0,1]；
The voltage class of 35 KV and above in the grid for directly supplying large user load can not be obtained by multiplying load density, and has Y value by means of declaration data systemi＝{y(1i),y(2i),…,y(ei) 0 for non-industrial value, and Δ Y for newly added report amount in the current yeariThe method comprises the following steps:
Figure BDA0001810563440000021
Figure BDA0001810563440000022
in the formula uiThe coincidence rate; wiIs the 110 thousand in the ith yearGridding comprehensive loads in the photovoltaic supply area; because the general land utilization construction progress has a certain positive relation with the load development progress, in order to accelerate the calculation convergence, the | theta (e) is seti)-γ(ei)|∈[0,0.2]；Pi,βiThe actual load density vector and the actual volume fraction vector of the ith year are respectively.
The invention describes the proportion of the annual grid to the maximum economic-load grid planning value by adopting the development degree, and fits the influence of the economic growth condition on the grid area load under the actual area development condition by limiting and predicting the development degree characteristics.
As a supplement to the above technical solution, the transformer substation supply area load prediction method trains a history sample by using an improved Elman network model:
and adopting the following formulas to calculate and correct the weight for the historical gridding load data:
Figure BDA0001810563440000023
Figure BDA0001810563440000024
Figure BDA0001810563440000031
for the S-shaped curve correction sample, the following method is adopted to calculate the correction weight:
Figure BDA0001810563440000032
Figure BDA0001810563440000033
Figure BDA0001810563440000034
Figure BDA0001810563440000035
in the above formulas, k represents a calculation serial number, and i and j represent the ith row and the jth column of the matrix respectively;
Figure BDA0001810563440000036
respectively the k and k +1 times of calculation of the connection weight from the input layer to the hidden layer,
Figure BDA0001810563440000037
the correction weight from the input layer to the hidden layer;
Figure BDA0001810563440000038
respectively calculating the connection weight from the k-th time to the hidden layer and the k + 1-th time,
Figure BDA0001810563440000039
the correction weight from the receiving layer to the hidden layer;
Figure BDA00018105634400000310
respectively the k and k +1 times of calculation of the connection weight from the hidden layer to the output layer,
Figure BDA00018105634400000311
the correction weight from the hidden layer to the output layer; etah、ηc、ηoutLearning rate factors of a hidden layer, a supporting layer and an output layer respectively;
Figure BDA00018105634400000312
the actual output value is calculated for the kth sample,
Figure BDA00018105634400000313
obtaining values for the kth calculation of the ith sample, i.e. sample value, EpA, b, c, d are all error objective functionsA sensitivity control coefficient adjusted according to the search sensitivity; xh、
Figure BDA00018105634400000314
Respectively representing an output value of a hidden layer, an output value of an ith sample hidden layer and an output value of a jth sample hidden layer; x c、
Figure BDA00018105634400000315
The output value of the receiving layer, the output value of the ith sample receiving layer and the output value of the jth sample receiving layer are respectively;
Figure BDA00018105634400000316
obtaining a value for the (k-1) th calculation of the jth sample;
the calculated correction weight value
Figure BDA00018105634400000317
To the connection weight
Figure BDA00018105634400000318
And updating to obtain the connection weight of the prediction calculation.
As a supplement to the above technical solution, the computation logic of the Elman neural network model is as follows:
for the neural network model, a vector is input
Figure BDA0001810563440000041
Output vectors for n-dimensional vectors, hidden layer and support layer
Figure BDA0001810563440000042
And
Figure BDA0001810563440000043
is a vector with n +1 dimension, and the output is a single value; considering the above, WhThe dimension of the connection weight from the input layer to the hidden layer is (n +1) x n, WcThe dimension of the connection weight from the bearer layer to the hidden layer is (n +1) × (n +1), WoutFor implicit layer to output layer connectionsThe value and the dimension are n +1, and the operation formula of the Elman neural network model is as follows:
Xh(k)＝f[Wh·Xc(k)+Wc·Sin(k-1)]，
Xc(k)＝Xh(k-1)，
Sout(k)＝g(Wout·Xh(k))，
wherein k represents a calculation sequence number f [ W ]h·Xc(k)+Wc·Sin(k-1)]For the hidden layer element stimulus function, g (W)out·Xh(k) Is the excitation function of the output layer element; during sample training, the error objective function is as follows:
Figure BDA0001810563440000044
Figure BDA0001810563440000045
the actual output value is calculated for the kth sample,
Figure BDA0001810563440000046
obtained for the kth calculation for the ith sample.
As a supplement to the above technical solution, while training is performed by using a history sample, an "S" type curve learning sample is used as a correction sample;
History sample:
the gridding decomposition and calculation are carried out on the whole county and city area in the past i years, the area is divided into m partitions, the vector set of the grid block area is S ═ { S (1), S (2), …, S (m) }, and for the grid S (m), the following are included:
βim＝si(m)/s(m)，
Figure BDA0001810563440000047
wherein beta isimIs the volume fraction of the m grid in the ith year,
Figure BDA0001810563440000048
planning the maximum volume fraction, s, for the m-gridi(m) is the development area of the ith year of the mth grid; thetaimM grid land availability in the ith year;
calculating the i-th grid development degree gammai(m) having:
Figure BDA0001810563440000049
and simultaneously, the supply area of the 110 kV transformer substation covering the mth grid is as follows:
Figure BDA0001810563440000051
γ'i＝{γ'(1i),γ'(2i),…,γ'(ei)}，
Figure BDA0001810563440000052
if gamma'i(m)-γi(m)|>ζ，
γi(m)＝|γ'i(m)+γi(m)|/2，
Wherein, wimIs the ith year of the gate load of m grids, w'imIs the predicted value of the gate load of the i-th year from the year on the m-grid, yimIs the load of the large user in the grid, gamma'iProviding an intra-area grid development degree vector set, W 'for a 110 kV substation supply area including an m-th grid in year i'iIs the load actual measurement data of the 110 kilovolt transformer substation in the ith year, gamma'i(m) the m-grid development degree obtained through the i-th year load actual measurement calculation, and zeta is a difference threshold, and the development degree is checked and corrected;
Figure BDA0001810563440000053
for the load of the m grid in future saturation periods, Δ yiIs newly added in the ith yearLarge user load of, Δ Yi＝{Δy1,...ΔynIs a newly added large user load vector set, u iThe simultaneous rate;
learning samples by using an S-shaped curve:
considering the development conditions of economy and power consumption, planning 2040 years as power consumption saturation years, and taking 1991 as a statistic initial year, wherein 50 years of power consumption load experiences an S-shaped curve to reach a saturation level;
adding an S-shaped curve into a sample for co-training, promoting the fitting degree of development degree sample training, and for an auxiliary sample, providing a fitting curve:
Figure BDA0001810563440000054
this is described in conjunction with the regional grid development as:
Figure BDA0001810563440000055
wherein i is year, cmIs the land property of the m-grid, cmC for living, business, municipal, financial, industrial, entertainmentmAnd (4, 5, 6) correcting the curve according to different land characteristics, and jointly learning the learning sample by using the historical actual gridding numerical value and the fitting curve.
As a supplement to the above technical solution, the input and output values are calculated for the degree of development and the degree of land development:
the single input is carried out by adopting historical data and land properties of the last five years, and in the ith year, the mth grid input is as follows:
xm(i)＝{cm,sm,γi-4,γi-3,γi-2,γi-1,γi,θi,εi-4,εi-3,εi-2,εi-1,εi}；
wherein x ism(i) Input quantity of m grid in i year, cm,smAre respectively asThe land property and the area of the m grids are calculated by the previous section to obtain the development degree condition gamma of the grids for 5 years (including the current year) i-4～γiObtaining economic growth rate epsilon of local market within five years through statistical datai-4～εi；
Output ym(i)＝(γi+1,θi+1)；
Training the elman neural network model by taking historical data and a fitting curve as learning samples to obtain connection weights of the function carrying layer, the input layer to the middle layer and the middle layer to the output layer
Figure BDA0001810563440000061
Substituting into a prediction calculation; each prediction adopts the aforementioned elman neural network model to generate a prediction value gamma of the development degree and the land availability in each iterationi+1,θi+1And the calculation is carried into the next calculation, the iteration is carried out for 3 times in total, the gridding load development condition in the later 3 years is predicted, the three-year development degree and the land utilization degree of the mth grid are obtained, and the three-year development degree and the land utilization degree are recorded as { gamma (m) (m is m)i+1),γ(mi+2),γ(mi+3),θ(mi+1),θ(mi+2),θ(mi+3) }; the method comprises the following steps of carrying out prediction calculation on all m grids in the city to obtain a 3-year development degree prediction value of each grid in the city, and dividing the grids according to the annual power supply range of a 110 KV substation, wherein the prediction value is expressed as: si+b＝{s(1i+b),s(2i+b),…,s(ei+b) B belongs to {1,2,3}, and a forecast value W of the supply area load is calculatedi+b：
Figure BDA0001810563440000062
The invention adopts a development degree concept to link the space-time grid load density with the maximum planned grid load density, extracts continuous trend variables (development degree) in the load development process, forms a data-deductible historical database, takes an S-shaped curve as an auxiliary sample, and adopts an improved Eman neural network to carry out the prediction calculation of the grid development degree; and after the predicted value is obtained, calculating the load predicted value condition in the supply area under the future grid development degree and the future supply area range by taking the 110 kilovolt supply area as a statistical range.
The method carries out modeling design on the characteristics of fully adapting to grid solidification and power transformation supply area change, integrates the advantages of perspective maximum load density data in space planning, and dynamically introduces the measured data of the transformer substation in the past year into a calculation sample to correct the grid development degree; and after the grid development degree is obtained through prediction, the load condition of the power supply area of the high-voltage distribution network is obtained through recalculation in consideration of the division change of the power transformation power supply area in the next year.
Drawings
FIG. 1 is a structural form diagram of an improved Elman neural network model in the embodiment of the invention;
FIG. 2 is a flow chart of a conventional power distribution network gridding load prediction method;
FIG. 3 is a graph illustrating the development of the development process according to the embodiment of the present invention;
FIG. 4 is a flowchart of a load prediction method according to an embodiment of the present invention.
Detailed Description
The invention is further described with reference to the drawings and the detailed description.
Elman neural network model
The method comprises an input layer, a hidden layer, a carrying layer and an output layer, wherein the hidden layer adopts a nonlinear excitation function, the carrying layer obtains the nth output of the hidden layer, and the feedback acts on the (n + 1) th calculation of the hidden layer. In combination with the requirement of the invention, the vector dimension of the network input layer is n, the dimension of the implicit hierarchical carrying layer is considered to be n +1, and the number of elements of the output layer is set to be 1 in a targeted manner in consideration of calculation precision and algorithm logic.
For the neural network, a vector is input
Figure BDA0001810563440000071
Output vectors for n-dimensional vectors, hidden layer and support layer
Figure BDA0001810563440000072
And
Figure BDA0001810563440000073
is a vector of dimension n +1, and the output is a single value. Considering the above, WhThe dimension of the connection weight from the input layer to the hidden layer is (n +1) x n, WcThe dimension of the connection weight from the bearer layer to the hidden layer is (n +1) × (n +1), WoutThe dimension is n +1 for the connection weight from the hidden layer to the output layer. The operation of the neural network is as follows:
Xh(k)＝f[Wh·Xc(k)+Wc·Sin(k-1)]，
Xc(k)＝Xh(k-1)，
Sout(k)＝g(Wout·Xh(k))，
wherein k represents a calculation sequence number f [ W ]h·Xc(k)+Wc·Sin(k-1)]For the hidden layer element stimulus function, g (W)out·Xh(k) Excitation function f (k) for output layer elements is the excitation function for hidden layer elements. Error objective function E in sample trainingpThe following were used:
Figure BDA0001810563440000074
Figure BDA0001810563440000075
the actual output value is calculated for the kth sample,
Figure BDA0001810563440000076
the value (sample value) is obtained for the kth calculation for the ith sample.
2. Gridding analysis method
The classification and partition prediction method is a space load prediction method based on a classification and partition principle, is also called a load density index method in many documents, and is characterized in that different power consumption properties of a planned area are classified according to a detailed urban control planning map (hereinafter referred to as a control planning map), and then the planned area is partitioned, namely the planned area is classified and partitioned according to the control planning map. On the basis of completing the classification subareas, collecting load historical data of each subarea, and selecting different power utilization properties and load density indexes of different subareas by combining load density indexes of developed cities or areas at home and abroad, so that the load distribution condition and the saturated load of a planned area are predicted. The conventional gridding load prediction method is shown in fig. 2.
3. Selection of particle size
The load grid granularity is mainly divided into three types, namely large, medium and small, and is suitable for analysis and calculation without calibers. Wherein, the large particle number takes an administrative district as a large district, or takes public facilities such as a main road, a national road, a green belt and the like or natural terrains such as a river and the like as boundaries; the medium particle size is generally 3-5km2The land property, street, greening, river channel and the like are generally used as boundaries; the small granularity is based on a city gauge control diagram, and the area of each block is generally 1km2The following. The invention is mainly designed for high-voltage distribution networks, and the detailed development condition in a supply area needs to be considered, so that the granularity is small.
Because the city control and regulation graph has detailed planning for each functional area, planners can easily determine the load types of different subareas. The load change of each cell is less influenced by the power supply range change of power equipment or a transformer substation. Because the power load property of each cell cannot change along with the change of the power supply range of the power equipment and the transformer substation, the historical load data of each cell still is one of the reliable bases for analyzing the load development rule.
Examples
The embodiment provides a high-voltage distribution network transformer substation load prediction method suitable for integrated planning of a transmission network and a distribution network, which comprises the following steps: the method is characterized in that the method is combined with the integrated planning demand characteristics of the transmission and distribution network, the high-voltage distribution network load prediction method combining the Elman neural network model and the space load prediction is adopted, the medium-long term load of the high-voltage distribution network is predicted by combining the historical load and the space prediction with the high-voltage distribution network substation supply area as a prediction unit, and a supply area load calculation model based on the grid development degree is introduced. The specific flow of the high-voltage distribution network transformer substation load prediction method is detailed in fig. 4.
In the integrated planning process, the supply area of the transformer substation, the district area and the county area are important basic level areas, and under the condition that the layered and zoning of the power grid are basically completed, the supply area of a high-voltage network (such as 110 kilovolts) station is generally positioned in the district area and the county area, and the influence of the trans-district power supply load on the whole after the power distribution network frame is basically straightened is small. Traditional province, city and grid load prediction is carried out in a space and administrative division range, and the prediction is difficult to effectively connect with balance, calculation and the like of a power grid and sites required by planning.
The specific contents of the supply area load calculation model based on the grid development degree are as follows:
the high-voltage distribution network is effectively linked with the load distribution of a higher-level voltage class through an operation mode, and is adapted to the load distribution prediction of a lower-level distribution network through space geographical division of a station supply area; the local city is divided into m grids according to municipal planning, and the total area is SmThe total number of k 110 KV substations in the range is equal to e in the supply area of the 110 KV substation in the ith year by taking the supply area of the 110 KV substation as an analysis areaiA vector set of grid block areas is Si＝{s(1i),s(2i),…,s(ei) }, there are
Figure BDA0001810563440000081
Considering that large users with voltage class of more than 35 kV are directly connected into a transformer substation with voltage class of more than 220 kV, only large users with voltage class of 10 kV or less are recorded when grid block load is calculated (the large users with voltage class of more than 35 kV are independently considered in subsequent calculation), and the maximum load density vector set is P i max＝{pmax(1i),pmax(2i),…,pmax(ei) H, the vector set of the maximum volume rate inside the grid is betai max＝{βmax(1i),βmax(2i),…,βmax(ei) }, introducing an area development degree vector gammai＝{γ(1i),γ(2i),…,γ(ei) Denotes the degree of load development within the grid, where γ (e)i)∈[0,1]Introduction of land utilization of gridi＝{θ(1i),θ(2i),…,θ(ei) Generation(s) }Expressing the ratio of grid volume fraction to target construction size over a period of time, where θ (e)i)∈[0,1]；
The voltage class of 35 KV and above in the grid can not be obtained by multiplying load density, and Y is calculated by means of the declaration data systemi＝{y(1i),y(2i),…,y(ei) } (non-industrial value is 0), and the newly-increased report volume in the current year is delta YiThe method comprises the following steps:
Figure BDA0001810563440000091
Figure BDA0001810563440000092
in the formula uiThe load is the same time rate (all grids do not reach the maximum load at the same time, and the same time rate represents the proportion of the average grid load to the maximum load per se when the overall load is maximum); wiSupplying the 110 KV power supply indoor gridding comprehensive load for the ith year; because the general land utilization construction progress has a certain positive relation with the load development progress, in order to accelerate the calculation convergence, the | theta (e) is seti)-γ(ei)|∈[0,0.2]；Pi,βiThe actual load density vector and the actual volume fraction vector of the ith year are respectively.
The invention describes the proportion of the annual grid to the maximum economic-load grid planning value by adopting the development degree, and fits the influence of the economic growth condition on the grid area load under the actual area development condition by limiting and predicting the development degree characteristics.
The transformer substation supply area load prediction method is characterized in that an improved Elman network model is adopted to train historical samples:
and adopting the following formulas to calculate and correct the weight for the historical gridding load data:
Figure BDA0001810563440000093
Figure BDA0001810563440000101
Figure BDA0001810563440000102
for the S-shaped curve correction sample, the following method is adopted to calculate the correction weight:
Figure BDA0001810563440000103
Figure BDA0001810563440000104
Figure BDA0001810563440000105
Figure BDA0001810563440000106
in the above formulas, k represents a calculation serial number, and i and j represent the ith row and the jth column of the matrix respectively;
Figure BDA0001810563440000107
respectively the k and k +1 times of calculation of the connection weight from the input layer to the hidden layer,
Figure BDA0001810563440000108
the correction weight from the input layer to the hidden layer;
Figure BDA0001810563440000109
respectively calculating the connection weight from the k-th time to the hidden layer and the k + 1-th time,
Figure BDA00018105634400001010
the correction weight from the receiving layer to the hidden layer;
Figure BDA00018105634400001011
respectively the k and k +1 times of calculation of the connection weight from the hidden layer to the output layer,
Figure BDA00018105634400001012
the correction weight from the hidden layer to the output layer; etah、ηc、ηoutLearning rate factors of a hidden layer, a supporting layer and an output layer respectively;
Figure BDA00018105634400001013
the actual output value is calculated for the kth sample,
Figure BDA00018105634400001014
obtaining values for the kth calculation of the ith sample, i.e. sample value, EpA, b, c and d are sensitivity control coefficients which are error objective functions and are adjusted according to the search sensitivity; xh、
Figure BDA00018105634400001015
Respectively representing an output value of a hidden layer, an output value of an ith sample hidden layer and an output value of a jth sample hidden layer; x c、
Figure BDA0001810563440000111
The output value of the receiving layer, the output value of the ith sample receiving layer and the output value of the jth sample receiving layer are respectively;
Figure BDA0001810563440000112
obtaining a value for the (k-1) th calculation of the jth sample;
the calculated correction weight value
Figure BDA0001810563440000113
To the connection weight
Figure BDA0001810563440000114
And updating to obtain the connection weight of the prediction calculation.
The computation logic of the Elman neural network model is as follows:
for the neural network model, a vector is input
Figure BDA0001810563440000115
Output vectors for n-dimensional vectors, hidden layer and support layer
Figure BDA0001810563440000116
And
Figure BDA0001810563440000117
is a vector with n +1 dimension, and the output is a single value; considering the above, WhThe dimension of the connection weight from the input layer to the hidden layer is (n +1) x n, WcThe dimension of the connection weight from the bearer layer to the hidden layer is (n +1) × (n +1), WoutFor the connection weight from the hidden layer to the output layer, the dimensionality is n +1, and the operation formula of the Elman neural network model is as follows:
Xh(k)＝f[Wh·Xc(k)+Wc·Sin(k-1)]，
Xc(k)＝Xh(k-1)，
Sout(k)＝g(Wout·Xh(k))，
wherein k represents a calculation sequence number f [ W ]h·Xc(k)+Wc·Sin(k-1)]For the hidden layer element stimulus function, g (W)out·Xh(k) Is the excitation function of the output layer element; during sample training, the error objective function is as follows:
Figure BDA0001810563440000118
Figure BDA0001810563440000119
the actual output value is calculated for the kth sample,
Figure BDA00018105634400001110
obtained for the kth calculation for the ith sample.
When a historical sample is adopted for training, an S-shaped curve learning sample is adopted as a correction sample;
History sample:
the gridding decomposition and calculation are carried out on the whole county and city area in the past i years, the area is divided into m partitions, the vector set of the grid block area is S ═ S (1), S (2), …, S (m) }, and for a grid S (m), the following are provided:
βim＝si(m)/s(m)，
Figure BDA00018105634400001111
wherein beta isimIs the volume fraction of the m grid in the ith year,
Figure BDA00018105634400001112
planning the maximum volume fraction, s, for the m-gridi(m) is the development area of the mth grid in year i; thetaimM grid land availability in the ith year;
calculating the i-th grid development degree gammai(m) having:
Figure BDA0001810563440000121
and simultaneously, the supply area of the 110 kV transformer substation covering the mth grid is as follows:
Figure BDA0001810563440000122
γ'i＝{γ'(1i),γ'(2i),…,γ'(ei)}，
Figure BDA0001810563440000123
if gamma'i(m)-γi(m)|>ζ，
γi(m)＝|γ'i(m)+γi(m)|/2，
Wherein, wimIs the ith year of the gate load of m grids, w'imIs the predicted value of the gate load of the i-th year from the year on the m-grid, yimIs the load of the large user in the grid, gamma'iProviding an intra-area grid development degree vector set, W 'for a 110 kV substation supply area including an m-th grid in year i'iIs the load actual measurement data of the 110 kilovolt transformer substation in the ith year, gamma'i(m) the m-grid development degree obtained through the i-th year load actual measurement calculation, and zeta is a difference threshold, and the development degree is checked and corrected;
Figure BDA0001810563440000124
for the load of the m grid in future saturation periods, Δ yiThe load of a newly increased large user in the ith year, delta Yi＝{Δy1,...ΔynIs a newly added large user load vector set, u iThe grid load is the same time rate (all grids do not reach the maximum load at the same time, and the same time rate represents the proportion of the average grid load to the maximum load when the whole load is maximum).
Learning samples by using an S-shaped curve:
considering the economic and power consumption development conditions, planning to be power utilization saturation years in 2040 years, taking 1991 as a statistic initial year, and during 50 years, the power utilization load is subjected to an S-shaped curve to reach a saturation level, as shown in FIG. 3.
Adding the S-shaped curve into the sample for co-training, promoting the fitting degree of development degree sample training, and for the auxiliary sample, having a fitting curve:
Figure BDA0001810563440000125
this is described in conjunction with the regional grid development as:
Figure BDA0001810563440000131
wherein i is year, cmIs the land property of m grids, cmC for living, business, municipal, financial, industrial, entertainmentmAnd (4, 5, 6) correcting the curve according to different land characteristics, and jointly learning the learning sample by using the historical actual gridding numerical value and the fitting curve.
For cmWhen {1,2,3,4,5,6}, i ∈ [0,50 ], respectively](step size is 0.2) into fm(i) And calculating to obtain a set array F { i, Fm(i),cmThe S-shaped learning sample refers to an array F { i, F }m(i),cm}. For example when i ═ 10.2, c mIn case of 3, f is calculatedm(i) 0.085, then {10.2,0.085,3} is an element in the array.
Calculating the input and output values aiming at the development degree and the land development degree:
the single input is carried out by adopting historical data and land properties of the last five years, and in the ith year, the mth grid input is as follows:
xm(i)＝{cm,sm,γi-4,γi-3,γi-2,γi-1,γi,θi,εi-4,εi-3,εi-2,εi-1,εi}；
wherein x ism(i) Input quantity of m grid in i year, cm,smRespectively the land property and the area of the m grids, and the development degree condition gamma of the grids for 5 years (including the current year) is obtained through the calculation of the previous sectioni-4～γiObtaining economic growth rate epsilon of local market within five years through statistical datai-4～εi；
Output ym(i)＝(γi+1,θi+1)；
Training the elman neural network model by using historical data and a fitting curve as learning samples to obtain the connection of a function bearing layer, an input layer to a middle layer and the connection of the middle layer to an output layerWeighted value
Figure BDA0001810563440000132
Carrying out a prediction calculation; each prediction adopts the aforementioned elman neural network model to generate a prediction value gamma of the development degree and the land availability in each iterationi+1,θi+1And the calculation is carried into the next calculation, the iteration is carried out for 3 times in total, the gridding load development condition in the later 3 years is predicted, the three-year development degree and the land utilization degree of the mth grid are obtained, and the three-year development degree and the land utilization degree are recorded as { gamma (m) (m is m)i+1),γ(mi+2),γ(mi+3),θ(mi+1),θ(mi+2),θ(mi+3) }; the method comprises the following steps of carrying out prediction calculation on all m grids in the city to obtain a 3-year development degree prediction value of each grid in the city, and dividing the grids according to the annual power supply range of a 110 KV substation, wherein the prediction value is expressed as: s i+b＝{s(1i+b),s(2i+b),…,s(ei+b) B belongs to {1,2,3}, and a forecast value W of the supply area load is calculatedi+b：
Figure BDA0001810563440000133
Application example
The related calculation and implementation steps are as follows:
1. and carrying out gridding segmentation on the city and land range, dividing the city and land range into m grids, and carrying out grid segmentation and formulation according to city planning.
2. And calling historical data of the power grid in the city range, and inquiring the load of the transformer substation to which the corresponding grid belongs in the ith year and the data of the grid gateway of m.
3. And calculating development degree historical data of each grid in the ith year by adopting a supply area load calculation model of the grid development degree.
4. And calculating the historical development degrees and the land utilization degrees of all grids and all years to finish the acquisition of historical samples.
5. And (3) performing property classification on all grids in the region, and fitting and correcting the sample by adopting an S-shaped function.
6. And leading the historical samples and the corrected samples into an Elman neural network according to an input and output format for learning to obtain a connection weight of the grid development degree.
7. And importing the historical data into an Elman neural network, and calculating by adopting the weight to obtain a development degree predicted value in the (i + 1) th year.
8. And (5) carrying out iterative calculation by adopting a new predicted value to obtain the grid development degree and the land utilization degree in the last 3 years.
9. And importing the range information of the substation supply area required to be calculated, and calculating by adopting a supply area load calculation model of the grid development degree to obtain a load prediction value of the substation supply area in the last 3 years.

Claims (5)

1. A high-voltage distribution network transformer substation load prediction method suitable for integrated planning of a transmission network and a distribution network is characterized by comprising the following steps:
the method is characterized in that the method is combined with the integrated planning demand characteristics of the transmission and distribution network, the high-voltage distribution network load prediction method combining an Elman neural network model and space load prediction is adopted, the medium-long-term load of the high-voltage distribution network is predicted by combining historical load and space prediction with the supply area of a high-voltage distribution network transformer substation as a prediction unit, and a supply area load calculation model based on grid development degree is introduced;
the specific contents of the supply area load calculation model based on the grid development degree are as follows:
the high-voltage distribution network is effectively linked with the load distribution of a higher-level voltage class through an operation mode, and is adapted to the load distribution prediction of a lower-level distribution network through space geographical division of a station supply area; the local city is divided into m grids according to municipal planning, and the total area is SmZ 110 KV substations are in total in the range, the supply area of the 110 KV substation is taken as an analysis area, and the supply area of the ith year transformer substation is in total eiA vector set of grid block areas is Si＝{s(1i),s(2i),…,s(ei) }, there are
Figure FDA0003534691740000011
When calculating the load of grid block, only recording the large users of 10 KV and below, and the maximum load density vector set is P i max＝{pmax(1i),pmax(2i),…,pmax(ei) }, the vector set of the maximum volume rate inside the grid is
Figure FDA0003534691740000012
Introducing a region development degree vector gammai＝{γ(1i),γ(2i),…,γ(ei) Denotes the degree of load development within the grid, where γ (e)i)∈[0,1]Introduction of land availability of grid thetai＝{θ(1i),θ(2i),…,θ(ei) Represents the ratio of the grid volume fraction to the target construction scale over a period of time, where θ (e)i)∈[0,1]；
The voltage class of 35 KV and above in the grid can not be obtained by multiplying load density, and Y is calculated by means of the declaration data systemi＝{y(1i),y(2i),…,y(ei) The non-industrial value is 0, and the newly added large user load vector group is delta YiThe method comprises the following steps:
Figure FDA0003534691740000013
Figure FDA0003534691740000014
Figure FDA0003534691740000015
in the formula uiThe coincidence rate; wiSupplying the 110 KV power supply indoor gridding comprehensive load for the ith year; to accelerate the convergence of the calculation, | θ (e) is seti)-γ(ei)|∈[0,0.2]；Pi,βiThe actual load density vector and the actual volume fraction vector of the ith year are respectively.
2. The method for predicting the load of the high-voltage distribution network substation according to claim 1, wherein a historical sample is trained by adopting an improved Elman neural network model;
and adopting the following formulas to calculate and correct the weight for the historical gridding load data:
Figure FDA0003534691740000021
Figure FDA0003534691740000022
Figure FDA0003534691740000023
for the S-shaped curve learning sample, the correction weight is calculated by adopting the following method:
Figure FDA0003534691740000024
Figure FDA0003534691740000025
Figure FDA0003534691740000026
Figure FDA0003534691740000027
in the above formulas, k represents a calculation serial number, and i and j represent the ith row and the jth column of the matrix respectively;
Figure FDA0003534691740000028
respectively input layer to implicit The connection weight value obtained by k and k +1 times of calculation of the layer,
Figure FDA0003534691740000029
the correction weight from the input layer to the hidden layer;
Figure FDA00035346917400000210
respectively calculating the connection weight from the k-th time to the hidden layer and the k + 1-th time,
Figure FDA00035346917400000211
the correction weight from the receiving layer to the hidden layer;
Figure FDA0003534691740000031
respectively the k and k +1 times of calculation of the connection weight from the hidden layer to the output layer,
Figure FDA0003534691740000032
the correction weight from the hidden layer to the output layer; etah、ηc、ηoutLearning rate factors of a hidden layer, a supporting layer and an output layer respectively;
Figure FDA0003534691740000033
the actual output value is calculated for the kth sample,
Figure FDA0003534691740000034
obtaining values for the kth calculation of the ith sample, i.e. sample value, EpA, b, c and d are sensitivity control coefficients which are error objective functions and are adjusted according to the search sensitivity; xh、
Figure FDA0003534691740000035
Respectively representing an output value of a hidden layer, an output value of an ith sample hidden layer and an output value of a jth sample hidden layer; xc、
Figure FDA0003534691740000036
The output value of the receiving layer, the output value of the ith sample receiving layer and the output value of the jth sample receiving layer are respectively;
Figure FDA0003534691740000037
obtaining a value for the (k-1) th calculation of the jth sample;
the calculated correction weight value
Figure FDA0003534691740000038
To the connection weight
Figure FDA0003534691740000039
And updating to obtain the connection weight of the prediction calculation.
3. The method for predicting the load of the high-voltage distribution network substation according to claim 2, wherein the computation logic of the Elman neural network model is as follows:
For the neural network model, a vector is input
Figure FDA00035346917400000310
Output vectors for n-dimensional vectors, hidden layer and support layer
Figure FDA00035346917400000311
And
Figure FDA00035346917400000312
is a vector with n +1 dimension, and the output is a single value; considering the above, WhThe dimension of the connection weight from the input layer to the hidden layer is (n +1) x n, WcThe dimension of the connection weight from the bearer layer to the hidden layer is (n +1) × (n +1), WoutFor the connection weight from the hidden layer to the output layer, the dimensionality is n +1, and the operation formula of the Elman neural network model is as follows:
Xh(k)＝f[Wh·Xc(k)+Wc·Sin(k-1)]，
Xc(k)＝Xh(k-1)，
Sout(k)＝g(Wout·Xh(k))，
wherein k represents a calculation sequence number f [ W ]h·Xc(k)+Wc·Sin(k-1)]For the hidden layer element stimulus function, g (W)out·Xh(k) Is the excitation function of the output layer element; during sample training, the error objective function is as follows:
Figure FDA00035346917400000313
Figure FDA00035346917400000314
the actual output value is calculated for the kth sample,
Figure FDA00035346917400000315
obtained for the kth calculation for the ith sample.
4. The method for predicting the load of the high-voltage distribution network substation according to claim 2, wherein an S-shaped curve learning sample is adopted as a correction sample while training is carried out by adopting a historical sample;
history sample:
the gridding decomposition and calculation are carried out on the whole county and city area in the past i years, the area is divided into m partitions, the vector set of the grid block area is S ═ { S (1), S (2), …, S (m) }, and for the grid S (m), the following are included:
βim＝si(m)/s(m)，
Figure FDA0003534691740000041
Wherein beta isimIs the volume fraction of the m grid in the ith year,
Figure FDA0003534691740000042
planning the maximum volume fraction, s, for the m-gridi(m) is the development area of the ith year of the mth grid; theta.theta.imM grid land availability in the ith year;
calculating the i-th grid development degree gammai(m) having:
Figure FDA0003534691740000043
and simultaneously, the supply area of the 110 kV transformer substation covering the mth grid is as follows:
Figure FDA0003534691740000044
γi'＝{γ'(1i),γ'(2i),…,γ'(ei)}，
Figure FDA0003534691740000045
if gammai'(m)-γi(m)|＞ζ，
γi(m)＝|γi'(m)+γi(m)|/2，
Wherein, wimIs the ith year of the gate load of m grids, w'imIs the predicted value of the gate load of the i-th year from the year on the m-grid, yimIs the load of the large user, gamma, in the gridi' supply district for 110 KV substation containing m-th grid supply district in i year grid development degree vector set, Wi' load measured data, gamma, of the 110 KV substation in the ith yeari' (m) is m grid development degree obtained by the actual measurement calculation of the load in the ith year, and zeta is a difference threshold, and the development degree is checked and corrected;
Figure FDA0003534691740000046
for the load of the m grid in future saturation periods, Δ yiFor new big users in the ith yearLoad capacity, Δ Yi＝{Δy1,...ΔynIs a newly added large user load vector set, uiThe coincidence rate;
learning samples of the "S" type curve:
considering the development conditions of economy and power consumption, planning to be power consumption saturation years in 2040 years, and taking 1991 as a statistic initial year, wherein the power consumption load reaches a saturation level through an S-shaped curve in 50 years;
Adding an S-shaped curve into a sample for co-training, promoting the fitting degree of development degree sample training, and for an auxiliary sample, providing a fitting curve:
Figure FDA0003534691740000051
this is described in conjunction with the area grid development degree as:
Figure FDA0003534691740000052
wherein i is year, cmIs the land property of the m-grid, cmC for { residential, commercial, municipal, financial, industrial, entertainment }mCorrecting the curve according to different land characteristics, wherein the curve is 1,2,3,4,5 and 6; and the learning samples adopt historical actual gridding numerical values and fitting curves to learn together.
5. The method for predicting the load of the high-voltage distribution network substation according to claim 4, wherein the input and output values are calculated according to the degree of development and the degree of land development:
the single input is carried out by adopting historical data and land properties of the last five years, and in the ith year, the mth grid input is as follows:
xm(i)＝{cm,sm,γi-4,γi-3,γi-2,γi-1,γi,θi,εi-4,εi-3,εi-2,εi-1,εi}；
wherein x ism(i) Input quantity of m grid in i year, cm,smRespectively the land property and the area of the m grids, and the development degree condition gamma of the grids for 5 years is obtained through the calculation of the previous sectioni-4～γiObtaining economic growth rate epsilon of local market within five years through statistical datai-4～εi；
Output ym(i)＝(γi+1,θi+1)；
Training the elman neural network model by taking historical data and a fitting curve as learning samples to obtain connection weights of the function carrying layer, the input layer to the middle layer and the middle layer to the output layer
Figure FDA0003534691740000053
Substituting into a prediction calculation; each prediction adopts the elman neural network model to generate a predicted value gamma of the development degree and the land availability in each iterationi+1,θi+1And the calculation is carried into the next calculation, the iteration is carried out for 3 times in total, the gridding load development condition in the later 3 years is predicted, the three-year development degree and the land utilization degree of the mth grid are obtained, and the three-year development degree and the land utilization degree are recorded as { gamma (m) (m is m)i+1),γ(mi+2),γ(mi+3),θ(mi+1),θ(mi+2),θ(mi+3) }; the method comprises the following steps of carrying out prediction calculation on all m grids in the city to obtain a 3-year development degree prediction value of each grid in the city, and dividing the grids according to the annual power supply range of a 110 KV substation, wherein the prediction value is expressed as: si+g＝{s(1i+g),s(2i+g),…,s(ei+g) And (e) determining g belongs to {1,2,3}, and calculating a supply area load predicted value Wi+g：
Figure FDA0003534691740000054
CN201811115905.8A 2018-09-25 2018-09-25 High-voltage distribution network transformer substation load prediction method suitable for integrated planning of transmission and distribution network Active CN109376907B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811115905.8A CN109376907B (en) 2018-09-25 2018-09-25 High-voltage distribution network transformer substation load prediction method suitable for integrated planning of transmission and distribution network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811115905.8A CN109376907B (en) 2018-09-25 2018-09-25 High-voltage distribution network transformer substation load prediction method suitable for integrated planning of transmission and distribution network

Publications (2)

Publication Number Publication Date
CN109376907A CN109376907A (en) 2019-02-22
CN109376907B true CN109376907B (en) 2022-06-28

Family

ID=65401765

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811115905.8A Active CN109376907B (en) 2018-09-25 2018-09-25 High-voltage distribution network transformer substation load prediction method suitable for integrated planning of transmission and distribution network

Country Status (1)

Country Link
CN (1) CN109376907B (en)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109902868A (en) * 2019-02-25 2019-06-18 国网河南省电力公司电力科学研究院 A kind of large user's industry expansion aided analysis method and device based on part throttle characteristics
CN110009136A (en) * 2019-03-12 2019-07-12 国网江西省电力有限公司电力科学研究院 A kind of load forecasting method of distribution transformer and distribution line
CN110765562B (en) * 2019-10-08 2023-09-01 杭州阳斯信息技术有限公司 Automatic target net rack generation method based on multi-source information and multi-constraint
CN111401638B (en) * 2020-03-17 2024-02-02 国网上海市电力公司 Spatial load prediction method based on extreme learning machine and load density index method
CN113723719B (en) * 2021-11-03 2022-02-11 国网北京市电力公司 Distribution transformation-based near-intermediate load prediction method, device, equipment and medium

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104599055A (en) * 2015-01-04 2015-05-06 湘潭大学 Power load demand management method for iron and steel enterprises
CN104636822A (en) * 2015-01-21 2015-05-20 广州市香港科大霍英东研究院 Residential load prediction method of elman-based neural network
CN105303259A (en) * 2015-11-06 2016-02-03 国网上海市电力公司 Gridding concept based power distribution network planning management method
CN107067104A (en) * 2017-03-30 2017-08-18 昆明能讯科技有限责任公司 A kind of power supply grid load forecasting method based on GIS-Geographic Information System
CN107169589A (en) * 2017-04-12 2017-09-15 佛山电力设计院有限公司 A kind of low pressure grid based on KNN and roulette algorithm becomes more meticulous load forecasting method

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104599055A (en) * 2015-01-04 2015-05-06 湘潭大学 Power load demand management method for iron and steel enterprises
CN104636822A (en) * 2015-01-21 2015-05-20 广州市香港科大霍英东研究院 Residential load prediction method of elman-based neural network
CN105303259A (en) * 2015-11-06 2016-02-03 国网上海市电力公司 Gridding concept based power distribution network planning management method
CN107067104A (en) * 2017-03-30 2017-08-18 昆明能讯科技有限责任公司 A kind of power supply grid load forecasting method based on GIS-Geographic Information System
CN107169589A (en) * 2017-04-12 2017-09-15 佛山电力设计院有限公司 A kind of low pressure grid based on KNN and roulette algorithm becomes more meticulous load forecasting method

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Power Load Forecasting based on improved Elman Neural Network;Xi Han等;《2018 ICEI》;20180525;第1-8页 *
基于多级聚类和改进Elman神经网络的空间负荷预测方法;傅荣荣;《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑》;20180315(第3期);摘要，第1.1、4.3.3节 *

Also Published As

Publication number Publication date
CN109376907A (en) 2019-02-22

Similar Documents

Publication Publication Date Title
CN109376907B (en) High-voltage distribution network transformer substation load prediction method suitable for integrated planning of transmission and distribution network
CN108846517B (en) Integration method for predicating quantile probabilistic short-term power load
CN110705743B (en) New energy consumption electric quantity prediction method based on long-term and short-term memory neural network
CN110474339B (en) Power grid reactive power control method based on deep power generation load prediction
CN112149879B (en) New energy medium-and-long-term electric quantity prediction method considering macroscopic volatility classification
CN101480143B (en) Method for predicating single yield of crops in irrigated area
CN107292534A (en) The yardstick competition evaluation method and device of urban power distribution network long term dynamics investment
CN104835103A (en) Mobile network health evaluation method based on neural network and fuzzy comprehensive evaluation
CN112101684B (en) Plug-in hybrid electric vehicle real-time energy management method and system
CN113554466B (en) Short-term electricity consumption prediction model construction method, prediction method and device
Szeto et al. Strategies for road network design over time: robustness under uncertainty
CN110490409B (en) DNN-based low-voltage transformer area line loss rate benchmarking value setting method
CN102063563A (en) Network-province-local integrative bus load predicting method
CN104050547A (en) Non-linear optimization decision-making method of planning schemes for oilfield development
CN105846425A (en) Economic dispatching method based on general wind power forecasting error model
CN105225006A (en) A kind of short-term wind-electricity power nonparametric probability forecasting method
CN111461921B (en) Load modeling typical user database updating method based on machine learning
CN101739614A (en) Hierarchy-combined prediction method for communication service
CN113537569B (en) Short-term bus passenger flow prediction method and system based on weight stacking decision tree
Li et al. An Integrated Artificial Neural Network-based Precipitation Revision Model.
CN107818678B (en) Real-time online modification method and device for power information acquisition system
Huang et al. Hybrid neural network models for hydrologic time series forecasting based on genetic algorithm
CN112098869B (en) Self-adaptive electric vehicle SOC estimation method based on big data
CN113139341A (en) Electric quantity demand prediction method and system based on federal integrated learning
CN114839586B (en) Low-voltage station metering device misalignment calculation method based on EM algorithm

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