CN104504465A  Power generation fuel supply prediction method  Google Patents
Power generation fuel supply prediction method Download PDFInfo
 Publication number
 CN104504465A CN104504465A CN201410778580.7A CN201410778580A CN104504465A CN 104504465 A CN104504465 A CN 104504465A CN 201410778580 A CN201410778580 A CN 201410778580A CN 104504465 A CN104504465 A CN 104504465A
 Authority
 CN
 China
 Prior art keywords
 model
 amp
 autocorrelogram
 seasonal
 power generation
 Prior art date
Links
 230000001932 seasonal Effects 0.000 claims abstract description 25
 238000000034 methods Methods 0.000 claims description 11
 230000000694 effects Effects 0.000 description 3
 238000005516 engineering processes Methods 0.000 description 3
 238000004364 calculation methods Methods 0.000 description 1
Classifications

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06Q—DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTING PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTING PURPOSES, NOT OTHERWISE PROVIDED FOR
 G06Q10/00—Administration; Management
 G06Q10/04—Forecasting or optimisation, e.g. linear programming, "travelling salesman problem" or "cutting stock problem"

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06Q—DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTING PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTING 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
Abstract
Description
Technical field
The present invention relates to a kind of fuel used to generate electricity supply Forecasting Methodology, belong to thermal power generating technology field.
Background technology
The supply of fuel used to generate electricity is subject to the impact of the Various Complex factors such as energy policy, Supply and demand trend, resource distribution, supply price, communications and transportation, Market Games, lacks rationally effective supply Forecasting Methodology and technological means for a long time.But, along with the enforcement that factory's net is separated, the grasp wretched insufficiency that grid company and power dispatching station are supplied fuel used to generate electricity, can not meet the requirement of electric power supply work, in the period of especially hemiplegia in water, power supply shortage, the prediction of fuel used to generate electricity supply is carried out and is sent out a management of power use to alleviation imbalance between power supply and demand, in order and play a part very important.
By the different in kind of Forecasting Methodology, prediction can be divided into qualitative forecasting and quantitative forecast.Conventional Qualitative Forecast Methods has subjective probability method, the poll projected method, Delphi method, analogy method, Study on Relative Factors method etc.Quantivative approach can be divided into causal method and time series analysis method etc. again, and causal method is also structural relation analytic approach.It is the reason by analyzing change, finds out the contact method between cause and effect, sets up forecast model, and predicts following development tendency and possibility level accordingly.Time series analysis method is also history extension method.It is based on the time series data of history, uses certain mathematical method to find data movement rule and stretches out, the development tendency that prediction is following.
The supply situation of fuel used to generate electricity is subject to the impact of Various Complex factor, and particularly along with the variation in season, fuel used to generate electricity supply there will be larger change.This sequential containing seasonal move, mathematically its development law of matching carry out predicting to be quite complicated.If but we can manage to isolate secular trend from sequential, and find out the rule of seasonal move, the two is combined prediction, problem just can be made to be simplified, also can reach the requirement of precision of prediction.
Summary of the invention
The object of the invention is to overcome deficiency of the prior art, provide a kind of fuel used to generate electricity to supply Forecasting Methodology, consider seasonal effect factor, can Accurate Prediction fuel used to generate electricity.
For achieving the above object, the technical solution adopted in the present invention is: a kind of fuel used to generate electricity supply Forecasting Methodology, comprises the steps:
Step one: tranquilization process: the nonstationary of raw data is converted to stationary time series;
Step 2: Model Identification and determine rank: according to autocorrelogram and partial autocorrelation figure, set up forecast model, Confirming model parameter:
The forecast model set up is:
Its exponent number (p, d, q) × (P, D, Q) _{s}represent, wherein:
θ(B)＝1θ _{1}Bθ _{2}B ^{2}…θ _{q}B ^{q}；
U(B ^{S})＝1u _{1}B ^{S}u _{2}B ^{2S}…u _{P}B ^{PS}；
V(B ^{S})＝1v _{1}B ^{S}v _{2}B ^{2S}…v _{Q}B ^{QS}；
S represents the observation number in a Seasonal Cycle, represent the correlationship of different cycles point in same period, represent the correlationship on the same period point of different cycles;
Step 3: model parameter estimation: by autocorrelogram determination parameter q and Q, determines p and P by partial autocorrelation figure, in conjunction with AIC and BIC criterion, finally determines p, q, P, Q; According to the stationarity determination parameter d of data; Periodicity according to data determines parameter D;
Step 4: Model suitability is checked: carry out residual error independence test or test for heteroscedasticity, amendment forecast model, until residual sequence is white noise sequence, all useful informations are extracted.
The concrete operation step of described tranquilization process is: carry out logtransformation, first order difference and seasonal difference computing successively to the nonstationary of raw data, be converted to stationary time series.
The formula of described residual error independence test is as follows:
Wherein: x represents random number, n represents degree of freedom, and m represents exponent number, r _{k}represent word related coefficient, ε _{t}represent residual sequence.
Compared with prior art, the beneficial effect that the present invention reaches is: the fuel used to generate electricity supply Forecasting Methodology based on Multiplicative Seasonality Model of proposition, forecast model adopts the convolution of random seaconal model and ARIMA model, under the prerequisite considering historical data and influence factor, better reflect the Seasonal of fuel used to generate electricity supply, improve fuel used to generate electricity precision of prediction.
Accompanying drawing explanation
Fig. 1 is operational flowchart of the present invention.
Fig. 2 is the time series chart of south electric network fuel used to generate electricity raw data.
Fig. 3 is the autocorrelogram of south electric network fuel used to generate electricity raw data.
Fig. 4 is the partial autocorrelation figure of south electric network fuel used to generate electricity raw data
Fig. 5 is the time series chart after the tranquilization process corresponding with Fig. 2.
Fig. 6 is the autocorrelogram after the tranquilization process corresponding with Fig. 3.
Fig. 7 is the partial autocorrelation figure after the tranquilization process corresponding with Fig. 4.
Fig. 8 is the residual error autocorrelogram of south electric network forecast model when carrying out residual test.
Fig. 9 is the residual error partial autocorrelation figure of south electric network forecast model when carrying out residual test.
Figure 10 is south electric network the whole network fuel used to generate electricity supply predicted value and actual value comparison diagram.
Embodiment
The basic ideas of seasonal move Forecasting Methodology are: first find and describe the whole mathematical model of sequential overall development trend and the tendency equation of separation trend; Next finds out the impact of seasonal move on forecasting object, is namely separated seasonal effect; Finally tendency equation and seasonal effect factor are merged, obtain the forecast model that can describe time series overall development rule, and for prediction.
Introduce random seaconal model, ARIMA model below respectively:
1, random seaconal model: be to the same period point of different cycles in seasonal random sequence between a kind of matching of correlationship.
AR (1): can be reduced to:
MA(1)：
The ARMA expressionform of seasonal form model is:
U(B ^{S})W _{t}＝V(B ^{S})e _{t}
Here,
2, ARIMA model:
If { X _{t}for zeromean stationary time series (t is time parameter t=1,2 ...), if and meet following condition:
(1) with θ (B) without common factor, wherein b is delay operator, BX _{t}=X _{t1}, B ε _{t}=ε _{t1}, B ^{j}x _{t}=X _{tj};
(3) { ε _{t}it is white noise sequence;
(4)E(X _{t}ε _{s})＝0,t＜s。
Then deserving to be called the model stated is autoregressive movingaverage model, is designated as ARMA (p, q).Wherein p is called Autoregressive, and q is called running mean exponent number, real coefficient be called autoregressive coefficient, θ _{1}, θ _{2}..., θ _{q}be called running mean coefficient.
Arma modeling race is the most important model race of one of Stationary Time Series, but for there being seasonal grade for Nonstationary timeseries, this model is just no longer practical, and the thought of ARIMA modeling is exactly by difference by sequential tranquilization, then adopts arma modeling that differentiated sequential is discussed.ARIMA (p, d, the q) model form of general Out of season is:
Multiplicative Seasonality Model is the convolution of random seaconal model and ARIMA model, its exponent number (p, d, q) × (P, D, Q) _{s}, form is:
Wherein:
θ(B)＝1θ _{1}Bθ _{2}B ^{2}…θ _{q}B ^{q}
U(B ^{S})＝1u _{1}B ^{S}u _{2}B ^{2S}…u _{P}B ^{PS}
V(B ^{S})＝1v _{1}B ^{S}v _{2}B ^{2S}…v _{Q}B ^{QS}
The value of s is the number observed in a Seasonal Cycle, here represent the correlationship of different cycles point in same period, then describe the correlationship on the same period point of different cycles, the two combines the effect just simultaneously featuring two factors.
Below in conjunction with accompanying drawing, the invention will be further described.
As shown in Figure 1, fuel used to generate electricity supply Forecasting Methodology, comprises the steps:
Step one: tranquilization process: the nonstationary of raw data is converted to stationary time series, be specially: logtransformation, first order difference and seasonal difference computing are carried out successively to the nonstationary of raw data, is converted to stationary time series.
Step 2: Model Identification and determine rank: according to autocorrelogram and partial autocorrelation figure, set up forecast model, Confirming model parameter:
The forecast model set up is:
Its exponent number (p, d, q) × (P, D, Q) _{s}represent, wherein:
θ(B)＝1θ _{1}Bθ _{2}B ^{2}…θ _{q}B ^{q}；
U(B ^{S})＝1u _{1}B ^{S}u _{2}B ^{2S}…u _{P}B ^{PS}；
V(B ^{S})＝1v _{1}B ^{S}v _{2}B ^{2S}…v _{Q}B ^{QS}；
S represents the observation number in a Seasonal Cycle, represent the correlationship of different cycles point in same period, represent the correlationship on the same period point of different cycles.
Step 3: model parameter estimation: by autocorrelogram determination parameter q and Q, determines p and P by partial autocorrelation figure, in conjunction with AIC and BIC criterion, finally determines p, q, P, Q; According to the stationarity determination parameter d of data; Periodicity according to data determines parameter D;
Step 4: Model suitability is checked: carry out residual error independence test or test for heteroscedasticity, amendment forecast model, until residual sequence is white noise sequence, all useful informations are extracted.
The concrete operation step of Model suitability inspection is as follows:
Step 401: residual error independence test: inspection formula is as follows:
Wherein: x represents random number, n represents degree of freedom, and m represents exponent number, r _{k}represent word related coefficient, ε _{t}represent residual sequence.
White noise verification method is: whether the adjoint p value of observing q statistic increases, if reduce to show residual sequence ε _{t}for white noise sequence; Otherwise, represent residual sequence ε _{t}not white noise sequence, also there is useful information in residual sequence and be not extracted, need to revise forecast model further.
Be described in further detail the present invention below in conjunction with specific embodiment, following examples only for technical scheme of the present invention is clearly described, and can not limit the scope of the invention with this.
Now be predicted as embodiment with southern the whole network grid generation fuel duty, concrete operation step is as follows:
Step one: tranquilization process:
As shown in Figure 2, Figure 3, Figure 4, the time series chart of south electric network fuel used to generate electricity raw data, autocorrelogram and partial autocorrelation figure is respectively.If mean value and the variance of a sequence are always constant, then it is claimed to be stable.Need jiggly time series to be converted into stationary sequence before series model in estimated time.Assess a seasonal effect in time series stationarity by data plot and autocorrelation function and its figure.If data plot presents linear or secondary trend, then time series is jiggly.If autocorrelation function drops to 0 after the several value of first few, then sequence is stable.If after front several value, autocorrelation function does not drop to 0, but successively reduces, then sequence is not steady.As can be seen from Fig. 2 to Fig. 4 all, former sequence chart is obviously not steady, there is fluctuation tendency.As shown in Figures 5 to 7, be the figure corresponding after tranquilization process of Fig. 2 to Fig. 3 respectively, as can be seen from Fig. 5 to Fig. 7, the trend after data logtransformation after first order difference and seasonal difference in sequence obtains elimination.
Step 2: Model Identification and determine rank:
Composition graphs 5 to Fig. 7, attempts setting up ARIMA (p, d, q) (P, D, Q) model, has carried out first order difference and single order seasonal difference, therefore d=D=1 above to data after taking the logarithm.
Step 3: model parameter estimation:
By autocorrelogram and partial autocorrelation figure and correlation criterion, after tentative calculation relatively, Selection parameter is chosen as ARIMA (0,1,1) (0,1,0) ^{12}.
Step 4: Model suitability is checked:
The adaptive test of model, i.e. the residual sequence ε of model _{t}independence test.If residual sequence is not white noise sequence, illustrates that the information also had in residual sequence is not extracted, need improve further master mould.Here we use the Chisquare Test of residual sequence, and formula is:
From Fig. 8 residual error autocorrelogram and Fig. 9 partial autocorrelation figure: residual error is white noise, and all information is extracted.Model suitability is upchecked.
By the identification of raw data tranquilization process, model and determine the steps such as rank, model parameter estimation, Model suitability inspection, south electric network the whole network fuel used to generate electricity predicts that final institute established model is:
(1B ^{12})(1B)lnx _{t}＝ε _{t}0.415ε _{t1}
South electric network the whole network fuel used to generate electricity supply predicted value and actual value comparison diagram are as shown in Figure 10, the fuel used to generate electricity supply Forecasting Methodology based on Multiplicative Seasonality Model that the present invention proposes, model adopts the convolution of random seaconal model and ARIMA model, under the prerequisite considering historical data and influence factor, better reflect the Seasonal of fuel used to generate electricity supply.By predicting south electric network the whole network fuel used to generate electricity supply, by the contrast with actual value, proving that this Forecasting Methodology predicts the outcome comparatively accurate, there is certain feasibility.
The above is only the preferred embodiment of the present invention; it should be pointed out that for those skilled in the art, under the prerequisite not departing from the technology of the present invention principle; can also make some improvement and distortion, these improve and distortion also should be considered as protection scope of the present invention.
Claims (3)
Priority Applications (1)
Application Number  Priority Date  Filing Date  Title 

CN201410778580.7A CN104504465A (en)  20141216  20141216  Power generation fuel supply prediction method 
Applications Claiming Priority (1)
Application Number  Priority Date  Filing Date  Title 

CN201410778580.7A CN104504465A (en)  20141216  20141216  Power generation fuel supply prediction method 
Publications (1)
Publication Number  Publication Date 

CN104504465A true CN104504465A (en)  20150408 
Family
ID=52945860
Family Applications (1)
Application Number  Title  Priority Date  Filing Date 

CN201410778580.7A CN104504465A (en)  20141216  20141216  Power generation fuel supply prediction method 
Country Status (1)
Country  Link 

CN (1)  CN104504465A (en) 
Cited By (3)
Publication number  Priority date  Publication date  Assignee  Title 

CN105680464A (en) *  20160225  20160615  浙江大学  Dispatching method considering battery loss for peak clipping and valley filling of battery energy storage system 
CN106126483A (en) *  20160621  20161116  湖北天明气和网络科技有限公司  A kind of method and device of weather forecasting 
CN106447746A (en) *  20160926  20170222  贵州电网有限责任公司输电运行检修分公司  Air temperature spacetime prediction distribution diagram drawing method 
Citations (5)
Publication number  Priority date  Publication date  Assignee  Title 

CN101771758A (en) *  20081231  20100707  北京亿阳信通软件研究院有限公司  Dynamic determine method for normal fluctuation range of performance index value and device thereof 
JP2010224832A (en) *  20090323  20101007  Itochu TechnoSolutions Corp  Electric power market price prediction method 
CN101894309A (en) *  20091105  20101124  南京医科大学  Epidemic situation predicting and early warning method of infectious diseases 
CN102869033A (en) *  20120921  20130109  西南交通大学  ARMAARCH model family based prediction method for GPRS (general packet radio service) data services of GSM (global system for mobile communications) communication system 
CN103745094A (en) *  20131225  20140423  河海大学  Interactive simulationprediction method and interactive simulationprediction system for hydrologic series of tidal reaches 

2014
 20141216 CN CN201410778580.7A patent/CN104504465A/en not_active Application Discontinuation
Patent Citations (5)
Publication number  Priority date  Publication date  Assignee  Title 

CN101771758A (en) *  20081231  20100707  北京亿阳信通软件研究院有限公司  Dynamic determine method for normal fluctuation range of performance index value and device thereof 
JP2010224832A (en) *  20090323  20101007  Itochu TechnoSolutions Corp  Electric power market price prediction method 
CN101894309A (en) *  20091105  20101124  南京医科大学  Epidemic situation predicting and early warning method of infectious diseases 
CN102869033A (en) *  20120921  20130109  西南交通大学  ARMAARCH model family based prediction method for GPRS (general packet radio service) data services of GSM (global system for mobile communications) communication system 
CN103745094A (en) *  20131225  20140423  河海大学  Interactive simulationprediction method and interactive simulationprediction system for hydrologic series of tidal reaches 
NonPatent Citations (3)
Title 

张璇: "ARIMA乘积季节模型在全社会用电量预测中的应用", 《现代经济信息》 * 
汤岩 等: "基于季节ARIMA模型的电力系统负荷短期预测", 《数学的实践与认识》 * 
陈艳华 等: "ARIMA乘积季节模型在大坝位移监测中的应用", 《测绘地理信息》 * 
Cited By (4)
Publication number  Priority date  Publication date  Assignee  Title 

CN105680464A (en) *  20160225  20160615  浙江大学  Dispatching method considering battery loss for peak clipping and valley filling of battery energy storage system 
CN105680464B (en) *  20160225  20190524  浙江大学  A kind of peak load shifting battery energy storage system dispatching method considering battery loss 
CN106126483A (en) *  20160621  20161116  湖北天明气和网络科技有限公司  A kind of method and device of weather forecasting 
CN106447746A (en) *  20160926  20170222  贵州电网有限责任公司输电运行检修分公司  Air temperature spacetime prediction distribution diagram drawing method 
Similar Documents
Publication  Publication Date  Title 

Wei et al.  Remaining useful life prediction and state of health diagnosis for lithiumion batteries using particle filter and support vector regression  
Laptev et al.  Timeseries extreme event forecasting with neural networks at uber  
Wan et al.  Probabilistic forecasting of wind power generation using extreme learning machine  
Sevlian et al.  Short term electricity load forecasting on varying levels of aggregation  
CN102208028B (en)  Fault predicting and diagnosing method suitable for dynamic complex system  
AlRashidi et al.  Long term electric load forecasting based on particle swarm optimization  
Riffonneau et al.  Optimal power flow management for grid connected PV systems with batteries  
US20150317589A1 (en)  Forecasting system using machine learning and ensemble methods  
Alamaniotis et al.  Evolutionary multiobjective optimization of kernelbased veryshortterm load forecasting  
US20150302313A1 (en)  Method of predicating ultrashortterm wind power based on selflearning composite data source  
KR101012863B1 (en)  Load forecasting analysis system for generation of customer baseline load  
CN101620045B (en)  Method for evaluating reliability of stepping stress quickened degradation experiment based on time sequence  
Guo et al.  A research on a comprehensive adaptive grey prediction model CAGM (1, N)  
Mahsin  Modeling rainfall in Dhaka division of Bangladesh using time series analysis  
CN101520652B (en)  Method for evaluating service reliability of numerical control equipment  
Li et al.  A new reliability prediction model in manufacturing systems  
CN103793854A (en)  Multiple combination optimization overhead transmission line operation risk informatization assessment method  
CN103530347B (en)  A kind of Internet resources method for evaluating quality based on big data mining and system  
Hossen et al.  Shortterm load forecasting using deep neural networks (DNN)  
CN102542155B (en)  Particle filter residual life forecasting method based on accelerated degradation data  
Zhang et al.  Remaining useful life prediction for lithiumion batteries based on exponential model and particle filter  
CN101894221A (en)  Method for predicting service life of product by accelerated degradation testing based on degenerate distribution nonstationary time series analysis  
Li et al.  A mutated particle filter technique for system state estimation and battery life prediction  
CN103745119A (en)  Oilimmersed transformer fault diagnosis method based on fault probability distribution model  
Lotfalipour et al.  Prediction of CO2 emissions in Iran using grey and ARIMA models 
Legal Events
Date  Code  Title  Description 

C06  Publication  
PB01  Publication  
C10  Entry into substantive examination  
SE01  Entry into force of request for substantive examination  
C41  Transfer of patent application or patent right or utility model  
TA01  Transfer of patent application right 
Effective date of registration: 20161207 Address after: 510623 Guangdong city of Guangzhou province Tianhe District Pearl River Metro Chinese Sui Road No. 6 Applicant after: China Southern Power Grid Co., Ltd. Applicant after: Nanjing Automation Co., Ltd., China Electronics Corp. Address before: 210009 Gulou District, Jiangsu, Nanjing new model road, No. 38 Applicant before: Nanjing Automation Co., Ltd., China Electronics Corp. 

RJ01  Rejection of invention patent application after publication  
RJ01  Rejection of invention patent application after publication 
Application publication date: 20150408 