CN103295081A - Electrical power system load prediction method based on back propagation (BP) neural network - Google Patents

Electrical power system load prediction method based on back propagation (BP) neural network Download PDF

Info

Publication number
CN103295081A
CN103295081A CN2013102733708A CN201310273370A CN103295081A CN 103295081 A CN103295081 A CN 103295081A CN 2013102733708 A CN2013102733708 A CN 2013102733708A CN 201310273370 A CN201310273370 A CN 201310273370A CN 103295081 A CN103295081 A CN 103295081A
Authority
CN
China
Prior art keywords
neural network
load
input
output
training
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN2013102733708A
Other languages
Chinese (zh)
Inventor
杨明莉
刘三明
王致杰
张卫
丁国栋
李义新
高叶军
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.)
Shanghai Dianji University
Original Assignee
Shanghai Dianji University
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 Shanghai Dianji University filed Critical Shanghai Dianji University
Priority to CN2013102733708A priority Critical patent/CN103295081A/en
Publication of CN103295081A publication Critical patent/CN103295081A/en
Pending legal-status Critical Current

Links

Abstract

An electric power system load prediction method based on a back propagation (BP) neural network comprises the following steps of (1) confirming input-output vectors according to specified electric power load prediction; (2) establishing a BP neural network model according to the input-output vectors; (3) training the BP neural network; (4) inputting a test sample to test the trained BP neural network, judging whether an error of an output predicted value and an actual value is smaller than a set threshold value or not and if so, carrying out a step (5); and (5) obtaining actually required load prediction according to the predicted value. The implicit presentation of internal relation of predictive factors is achieved through automatic learning of the BP neural network model and weight distribution of the BP neural network, the accuracy of the electric power system load prediction is high, and safe and stable operation and economical efficiency of an electric power system can be guaranteed effectively.

Description

Power system load Forecasting Methodology based on the BP neural network
Technical field
The present invention relates to the power system load electric powder prediction, particularly relate to a kind of power system load Forecasting Methodology based on the BP neural network.
Background technology
The power system load prediction is the important component part of Power System Planning, also is the basis of Economical Operation of Power Systems, and it is all of crucial importance to Power System Planning and operation.By load prediction accurately, can arrange the unit start and stop economically, reduce spinning reserve capacity, reasonably arrange turnaround plan, reduce cost of electricity-generating, increase economic efficiency.
Load prediction according to the time of prediction can be divided into for a long time, mid-term and short-term load forecasting.Wherein in short-term load forecasting, all load predictions (following 7 days), daily load prediction (prediction in following 24 hours) and a few hours prediction in advance are most important to the real time execution scheduling of electric system.Because to predicting that scheduling will be foundation with the result of load prediction future constantly, the accuracy of load prediction results will directly influence the result of scheduling, thereby safe and stable operation and the economy of electric system are brought material impact.
Load prediction all has great significance to electric power system control, operation and plan.The power system load variation is subjected to many-sided the influence.On the one hand, load variations exists the random fluctuation that the position uncertain factor causes; On the other hand, have the regularity that the cycle changes again, this makes that also load curve has similarity.Simultaneously, owing to be subjected to the influence of special circumstances such as weather, festivals or holidays, make load variations difference occur again.
Therefore, need a kind of mode that can improve the load prediction precision of electric system, to guarantee safe and stable operation and the economy of electric system.
Summary of the invention
The objective of the invention is to, a kind of power system load Forecasting Methodology based on the BP neural network is provided, automatic study by the BP neural network model, with weights distribution realization recessive express of the inner link between predictive factors by neural BP neural network, realization can effectively guarantee safe and stable operation and the economy of electric system to the accurate prediction of power system load.
For achieving the above object, the present invention has adopted following technical scheme.
A kind of power system load Forecasting Methodology based on the BP neural network may further comprise the steps: (1) is according to specifying load forecast to determine the input and output vector; (2) according to input and output vector structure BP neural network model; (3) the BP neural network is carried out network training; (4) the BP neural network of input test sample after to training carried out network test, whether judges the predicted value of output and the error between the actual value less than setting threshold, if execution in step (5) then; (5) obtain the load prediction of actual needs according to described predicted value.
The advantage that the present invention is based on the power system load Forecasting Methodology of BP neural network is: the BP neural network model has utilization approaches any continuous function with arbitrary accuracy character, automatic study by model, with weights distribution realization recessive express of the inner link between predictive factors by neural BP neural network, to power system load prediction accuracy height, can effectively guarantee safe and stable operation and the economy of electric system.
Description of drawings
Fig. 1 is the process flow diagram that the present invention is based on the power system load Forecasting Methodology of BP neural network.
Embodiment
Below in conjunction with accompanying drawing the power system load Forecasting Methodology that the present invention is based on the BP neural network being elaborated, but should be pointed out that embodiments of the present invention and embodiment are the preferred versions for task of explanation, is not limitation of the scope of the invention.
Referring to Fig. 1, next the process flow diagram of the power system load Forecasting Methodology based on the BP neural network of the present invention elaborates to the described step of this method.
S11: according to specifying load forecast to determine the input and output vector.
All regard 7 day every day in a week as one type, allow this potential relation of seven types of neural network learning, and inform the load type of the electric system of prediction that day.
Step S11 can adopt, and step realizes:
(11) every setting-up time electric load is measured a day the previous day in prediction, obtained corresponding measurement data as the electric load input variable.For example, in the previous day of predicting day, every 2 hours electric load is carried out 1 time and measure, namely obtain one day 12 groups of data.Owing to can not undergo mutation between the adjacent point of load value curve, therefore back one value constantly is inevitable relevant with the value of eve, unless special circumstances such as major accident occur.
(12) obtain the Meteorological Characteristics data on same day load forecast day as the Meteorological Characteristics input variable.Because electric load is also relevant with environmental factor, such as the highest and the lowest temperature etc.Therefore, also need to obtain by means such as weather forecasts the highest temperature and the lowest temperature and the weather characteristics value (fine day, cloudy day or rainy day) of prediction day.Here use numeral weather characteristics value: 0 expression is fine, 0.5 expression cloudy day, 1 expression rainy day.Therefore here with the Meteorological Characteristics data on the same day load forecast day input variable as the BP neural network, namely the input variable of BP neural network is exactly the vector of one 15 dimension.
(13) determine that prediction day same day required load value number is as output vector.Object vector is prediction 12 load values on day same day, in namely one day every the electric load of an integral point, so output vector is the vector of one 12 dimension.
(14) input/output variable of determining is carried out normalized.After obtaining input/output variable, it is carried out normalized, process data into the data between [0,1], adopt following formula to carry out the normalized of inputoutput data here.
V i - min V i max V i - min V i
In the following formula, V iRepresent in one day the burden with power value that measures every two hours integral points, minV iRepresent the minimum value of the electric system burden with power data of surveying in a day, maxV iRepresent the maximal value of the electric system burden with power data of surveying in a day.
S12: according to input and output vector structure BP neural network model.
Step S12 can adopt, and step realizes:
(21) determine the neuronal quantity of input layer according to the amount of element of input vector.For example input vector has 15 elements, so the neuron of input layer has 15.
(22) determine BP neural network middle layer neuronal quantity according to the neuronal quantity of input layer.For example, the neuron of input layer has 15, and according to the Kolmogorv theorem as can be known, BP neural network middle layer neuron can be got 31.The neuron transport function in BP neural network middle layer adopts S type tan tansig.
(23) determine the neuronal quantity of output layer according to the amount of element of output vector, wherein, output layer neuron transport function adopts S type logarithmic function.For example, output vector has 12, so the neuron in the output layer is made as 12.Output layer neuron transport function adopts S type logarithmic function logsig.
S13: the BP neural network is carried out network training.
Set training parameter according to the neuronal quantity of structure BP neural network model the BP neural network is carried out network training, just can be used for the practical application of load forecast behind the BP neural metwork training.Described training parameter comprises: frequency of training, training objective and pace of learning.When BP neural network structure more complicated, when neuron number is many, can suitably increase frequency of training and learning rate.
S14: the BP neural network of input test sample after to training carried out network test, whether judges the predicted value of output and the error between the actual value less than setting threshold, if execution in step S15 then; Otherwise need re-construct the BP neural network model.
The BP neural network that trains need be tested just can judge whether can drop into actual use.After with training sample the BP neural network being carried out network training, continue to BP neural network input test sample, output is predicted value.Predicted value is compared with actual electric power load value, can obtain error between the two, if error, illustrates then that the BP neural network model of setting up satisfies the requirement of using less than setting threshold, threshold value can be made as 0.3.
S15: the load prediction of obtaining actual needs according to described predicted value.
The predicted value of step S14 output and the error between the actual value less than setting threshold after, the predicted value of BP neural network output can be reduced to the load prediction of actual needs by following formula.
T=out*(max(V i)-min(V i))+min(V i)
Next provide the embodiment of technique scheme.Here with the integral point burden with power value in 10 days July in 2004 to July 20 in certain short of electricity city, south, and the Meteorological Characteristics state vector on July 11st, 2004 to July 21 is as the training sample of network, the electric load in prediction July 21.
One, according to specifying load forecast to determine the input and output vector
In the previous day of predicting day, every 2 hours electric load is carried out 1 time and measure, namely obtain one day 12 groups of data.Obtain the prediction weather characteristics value of day (fine day, cloudy day or rainy day) by means such as weather forecasts.Here use numeral weather characteristics value: 0 expression is fine, 0.5 expression cloudy day, 1 expression rainy day.Therefore the input variable of BP neural network is exactly the vector of one 15 dimension.Object vector is prediction 12 load values on day same day, in namely one day every the electric load of an integral point, so output vector is the vector of one 12 dimension.
The input/output variable of determining is carried out normalized, and the data after the normalization are as shown in table 1 below.
The sample date Electric load Meteorological Characteristics
2012-7-10 0.24520.14660.13140.22430.55230.66420.70150.69810.68210.69450.75490.8215 ?
2012-7-11 0.22170.15810.14080.23040.51340.53120.68190.71250.72650.68470.78260.8325 0.24150.30270
2012-7-12 0.25250.16270.15070.24060.55020.56360.70510.73520.74590.70150.80640.8156 0.23850.31250
2012-7-13 0.20160.11050.12430.19870.50210.52320.68190.69520.70150.68250.78250.7895 0.22160.27011
2012-7-14 0.21150.12010.13120.20190.55320.57360.70290.70320.71890.70190.79650.8025 0.23520.25060.5
20127-15 0.23350.13220.15340.22140.56230.58270.71980.72760.73590.75060.80920.8221 0.25420.31250
2012-7-16 0.23680.14320.16530.22050.58230.59710.71360.71290.72630.71530.80910.8217 0.26010.31980
2012-7-17 0.23420.13680.16020.21310.57260.58220.71010.70980.71270.71210.79950.8126 0.25790.30990
2012-7-18 0.21130.12120.13050.18190.49520.53120.68860.68980.69990.73230.77210.7956 0.23010.28670.5
2012-7-19 0.20050.11210.12070.16050.45560.50220.65530.66730.67980.70230.75210.7756 0.22340.27991
2012-7-20 0.21230.12570.13430.20790.55790.57160.70590.71450.72050.74010.80190.8136 0.23140.29770
2012-7-21 0.21190.12150.16210.21610.61710.61590.71550.72010.72430.72980.81790.8229 0.23170.29360
Data after table 1. normalization
Two, according to input and output vector structure BP neural network model
Because input vector has 15 elements here, so the neuron of input layer has 15, according to the Kolmogorv theorem as can be known, network middle layer neuron can be got 31, output vector has 12, so the neuron in the output layer is made as 12, the neuron transport function in network middle layer adopts S type tan tansig, and output layer neuron transport function adopts S type logarithmic function logsig.
Three, the BP neural network is carried out network training
Just can be used for the practical application of load forecast behind the network training.Consider the network structure more complicated, neuron number is many, needs suitably to increase frequency of training and learning rate.Being set as follows shown in the table 2 of training parameter.
Frequency of training Training objective Learning rate
1000 0.01 0.1
Table 2. training parameter
Four, the input test sample carries out network test to the BP neural network after training.
The network that trains need be tested just can judge whether can drop into actual use.Here utilize the electric load on July 20th, 2012 and 21 days Meteorological Characteristics data to predict 21 days power load, can meet the demands with the check predicated error.Utilize emulation function sim to come the output of computational grid, operation result is: out=0.21200.12090.16090.21500.61780.61610.71311.00000.72 380.72770.82360.8234.
Predicted value is compared and can be found with the actual electric load on July 21st, 2012, and predicted value and the error between the actual value of network are very little, and maximum error also has only 0.28, satisfy the requirement of using.
Five, obtain the load prediction of actual needs according to described predicted value
The predicted value of network output is reduced to the predicted value T of actual needs by following formula.
T=out*(max(V i)-min(V i))+min(V i)
The above only is preferred implementation of the present invention; should be pointed out that for those skilled in the art, under the prerequisite that does not break away from the principle of the invention; can also make some improvements and modifications, these improvements and modifications also should be considered as protection scope of the present invention.

Claims (6)

1. the power system load Forecasting Methodology based on the BP neural network is characterized in that, may further comprise the steps:
(1) according to specifying load forecast to determine the input and output vector;
(2) according to input and output vector structure BP neural network model;
(3) the BP neural network is carried out network training;
(4) the BP neural network of input test sample after to training carried out network test, whether judges the predicted value of output and the error between the actual value less than setting threshold, if execution in step (5) then;
(5) obtain the load prediction of actual needs according to described predicted value.
2. the power system load Forecasting Methodology based on the BP neural network according to claim 1 is characterized in that, step (1) further comprises:
(11) every setting-up time electric load is measured a day the previous day in prediction, obtained corresponding measurement data as the electric load input variable;
(12) obtain the Meteorological Characteristics data on same day load forecast day as the Meteorological Characteristics input variable;
(13) determine that prediction day same day required load value number is as output vector;
(14) input/output variable of determining is carried out normalized.
3. the power system load Forecasting Methodology based on the BP neural network according to claim 2 is characterized in that, the normalization formula that adopts in the step (14) is:
4. the power system load Forecasting Methodology based on the BP neural network according to claim 1 is characterized in that, step (2) further comprises:
(21) determine the neuronal quantity of input layer according to the amount of element of input vector;
(22) determine BP neural network middle layer neuronal quantity according to the neuronal quantity of input layer and output layer, wherein, the neuron transport function in BP neural network middle layer adopts S type tan;
(23) determine the neuronal quantity of output layer according to the amount of element of output vector, wherein, output layer neuron transport function adopts S type logarithmic function.
5. the power system load Forecasting Methodology based on the BP neural network according to claim 1, it is characterized in that, step (3) further comprises: set training parameter according to the neuronal quantity of structure BP neural network model the BP neural network is carried out network training, wherein said training parameter comprises: frequency of training, training objective and pace of learning.
6. the power system load Forecasting Methodology based on the BP neural network according to claim 1 is characterized in that, step (5) further comprises: adopt formula:
T=out*(max(Vi)-min(Vi))+min(Vi)
The predicted value of BP neural network output is reduced to the load prediction of actual needs.
CN2013102733708A 2013-07-02 2013-07-02 Electrical power system load prediction method based on back propagation (BP) neural network Pending CN103295081A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2013102733708A CN103295081A (en) 2013-07-02 2013-07-02 Electrical power system load prediction method based on back propagation (BP) neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2013102733708A CN103295081A (en) 2013-07-02 2013-07-02 Electrical power system load prediction method based on back propagation (BP) neural network

Publications (1)

Publication Number Publication Date
CN103295081A true CN103295081A (en) 2013-09-11

Family

ID=49095901

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2013102733708A Pending CN103295081A (en) 2013-07-02 2013-07-02 Electrical power system load prediction method based on back propagation (BP) neural network

Country Status (1)

Country Link
CN (1) CN103295081A (en)

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103581188A (en) * 2013-11-05 2014-02-12 中国科学院计算技术研究所 Network security situation forecasting method and system
CN103593718A (en) * 2013-11-26 2014-02-19 国家电网公司 Load combination forecasting method and device
CN104484700A (en) * 2014-11-07 2015-04-01 西安文理学院 Load prediction model input variable optimizing method based on BP (back propagation) network model
CN106096761A (en) * 2016-06-01 2016-11-09 新奥泛能网络科技股份有限公司 A kind of building load Forecasting Methodology based on neutral net and device
CN106096726A (en) * 2016-05-31 2016-11-09 华北电力大学 A kind of non-intrusion type load monitoring method and device
CN106600050A (en) * 2016-12-10 2017-04-26 国网辽宁省电力有限公司锦州供电公司 BP neural network-based ultra-short load prediction method
CN107274012A (en) * 2017-06-06 2017-10-20 上海电机学院 Short-term wind power forecast method based on cloud evolution particle cluster algorithm
CN107423839A (en) * 2017-04-17 2017-12-01 湘潭大学 A kind of method of the intelligent building microgrid load prediction based on deep learning
CN107844859A (en) * 2017-10-31 2018-03-27 深圳达实智能股份有限公司 Large medical equipment energy consumption Forecasting Methodology and terminal device based on artificial intelligence
CN109034500A (en) * 2018-09-04 2018-12-18 湘潭大学 A kind of mid-term electric load forecasting method of multiple timings collaboration
CN109034495A (en) * 2018-08-30 2018-12-18 珠海吉瓦科技有限公司 Electric Load Prediction System based on edge calculations
CN109473985A (en) * 2019-01-16 2019-03-15 江苏圣通电力新能源科技有限公司 One kind being based on BP neural network smart grid distribution method
CN110084717A (en) * 2019-03-25 2019-08-02 华电电力科学研究院有限公司 A kind of Utility Boiler moisture content of coal calculation method based on BP neural network

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103093285A (en) * 2013-01-22 2013-05-08 清华大学 Short-term load forecast method based on artificial neural network

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103093285A (en) * 2013-01-22 2013-05-08 清华大学 Short-term load forecast method based on artificial neural network

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
SHEN JIA等: "The Application of BP Neural Networks based on Chaotic Analysis algorithm in Short-term power Load Forecasting", 《PROCEEDINGS OF 2006 INTERNATIONAL SYMPOSIUM ON DISTRIBUTED COMPUTING AND APPLICATIONS TO BUSINESS, ENGINEERING AND SCIENCE》, 1 October 2006 (2006-10-01), pages 685 - 687 *
朱雨晨等: "基于BP神经网络的电力系统短期负荷预测", 《黑龙江电力》, vol. 34, no. 06, 15 December 2012 (2012-12-15), pages 439 - 445 *
熊永胜: "基于BP神经网络的电力系统短期负荷预测", 《成都大学学报(自然科学版)》, vol. 31, no. 02, 30 June 2012 (2012-06-30), pages 167 - 169 *

Cited By (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103581188A (en) * 2013-11-05 2014-02-12 中国科学院计算技术研究所 Network security situation forecasting method and system
CN103581188B (en) * 2013-11-05 2016-08-03 中国科学院计算技术研究所 A kind of network security situation prediction method and system
CN103593718A (en) * 2013-11-26 2014-02-19 国家电网公司 Load combination forecasting method and device
CN104484700A (en) * 2014-11-07 2015-04-01 西安文理学院 Load prediction model input variable optimizing method based on BP (back propagation) network model
CN106096726A (en) * 2016-05-31 2016-11-09 华北电力大学 A kind of non-intrusion type load monitoring method and device
CN106096726B (en) * 2016-05-31 2018-12-18 华北电力大学 A kind of non-intrusion type load monitoring method and device
CN106096761A (en) * 2016-06-01 2016-11-09 新奥泛能网络科技股份有限公司 A kind of building load Forecasting Methodology based on neutral net and device
CN106600050A (en) * 2016-12-10 2017-04-26 国网辽宁省电力有限公司锦州供电公司 BP neural network-based ultra-short load prediction method
CN107423839A (en) * 2017-04-17 2017-12-01 湘潭大学 A kind of method of the intelligent building microgrid load prediction based on deep learning
CN107274012B (en) * 2017-06-06 2020-07-28 上海电机学院 Short-term wind power prediction method based on cloud evolution particle swarm algorithm
CN107274012A (en) * 2017-06-06 2017-10-20 上海电机学院 Short-term wind power forecast method based on cloud evolution particle cluster algorithm
CN107844859B (en) * 2017-10-31 2020-08-07 深圳达实智能股份有限公司 Large medical equipment energy consumption prediction method based on artificial intelligence and terminal equipment
CN107844859A (en) * 2017-10-31 2018-03-27 深圳达实智能股份有限公司 Large medical equipment energy consumption Forecasting Methodology and terminal device based on artificial intelligence
CN109034495A (en) * 2018-08-30 2018-12-18 珠海吉瓦科技有限公司 Electric Load Prediction System based on edge calculations
CN109034500A (en) * 2018-09-04 2018-12-18 湘潭大学 A kind of mid-term electric load forecasting method of multiple timings collaboration
CN109473985A (en) * 2019-01-16 2019-03-15 江苏圣通电力新能源科技有限公司 One kind being based on BP neural network smart grid distribution method
CN110084717A (en) * 2019-03-25 2019-08-02 华电电力科学研究院有限公司 A kind of Utility Boiler moisture content of coal calculation method based on BP neural network

Similar Documents

Publication Publication Date Title
CN103295081A (en) Electrical power system load prediction method based on back propagation (BP) neural network
Ma et al. Scenario generation of wind power based on statistical uncertainty and variability
CN103218675B (en) A kind of based on the short-term load forecasting method of cluster and moving window
CN102622473B (en) Optimization design method for step stress accelerated degradation test based on Bayesian theory
CN103279804B (en) The Forecasting Methodology of super short-period wind power
CN105631578A (en) Risk assessment-orientated modeling method of power transmission and transformation equipment failure probability model
CN105654207A (en) Wind power prediction method based on wind speed information and wind direction information
CN104318332A (en) Power load predicting method and device
CN103488869A (en) Wind power generation short-term load forecast method of least squares support vector machine
CN103544544A (en) Energy consumption forecasting method and device
CN103123665A (en) Short-term power load forecasting method based on fuzzy clustering similar day
CN103424654A (en) Method for assessing voltage sag sensitivity of sensitive equipment
CN102509027B (en) Wind powder combined predication method based on cross entropy theory
CN102938562B (en) Prediction method of total wind electricity power in area
CN104951834A (en) LSSVM (least squares support vector machine) wind speed forecasting method based on integration of GA (genetic algorithm) and PSO (particle swarm optimization)
Breker et al. Capacity of low-voltage grids for distributed generation: classification by means of stochastic simulations
CN105303268A (en) Wind power generation output power prediction method based on similarity theory
CN105488335A (en) Lyapunov exponent based power system load prediction method and apparatus
Yang et al. Photovoltaic power forecasting with a rough set combination method
CN105305426B (en) Mapreduce two-step short-period load prediction method based on deviation control mechanism
CN109657881A (en) A kind of neural network photovoltaic power generation prediction technique and system suitable for small sample
Yan et al. A robust probabilistic wind power forecasting method considering wind scenarios
Motepe et al. South African distribution networks load forecasting using ANFIS
Sharma Energy forecasting based on predictive data mining techniques in smart energy grids
Xiyun et al. Wind power probability interval prediction based on Bootstrap quantile regression method

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C12 Rejection of a patent application after its publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20130911