CN107862476A - A kind of metering table demand computational methods based on data analysis - Google Patents
A kind of metering table demand computational methods based on data analysis Download PDFInfo
- Publication number
- CN107862476A CN107862476A CN201711268203.9A CN201711268203A CN107862476A CN 107862476 A CN107862476 A CN 107862476A CN 201711268203 A CN201711268203 A CN 201711268203A CN 107862476 A CN107862476 A CN 107862476A
- Authority
- CN
- China
- Prior art keywords
- calculated
- model
- metering
- metering table
- chronomere
- 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
Links
- 238000000205 computational method Methods 0.000 title claims abstract description 16
- 238000007405 data analysis Methods 0.000 title claims abstract description 14
- 238000009499 grossing Methods 0.000 claims abstract description 70
- 238000004364 calculation method Methods 0.000 claims abstract description 6
- 238000005259 measurement Methods 0.000 claims description 19
- 230000001932 seasonal effect Effects 0.000 claims description 8
- 238000009434 installation Methods 0.000 claims description 4
- 238000011156 evaluation Methods 0.000 claims 1
- 230000008676 import Effects 0.000 claims 1
- 238000004519 manufacturing process Methods 0.000 abstract description 4
- 230000000694 effects Effects 0.000 description 16
- 238000000034 method Methods 0.000 description 12
- 238000004422 calculation algorithm Methods 0.000 description 8
- 241001269238 Data Species 0.000 description 6
- 238000012360 testing method Methods 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 4
- 230000007423 decrease Effects 0.000 description 3
- 238000007726 management method Methods 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 238000012300 Sequence Analysis Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013278 delphi method Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000013213 extrapolation Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000001373 regressive effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000013179 statistical model Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06315—Needs-based resource requirements planning or analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- Public Health (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- General Health & Medical Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of metering table demand computational methods based on data analysis, comprise the following steps:Step A:Data acquisition is carried out with arranging;Step B:Model creation is carried out, exponential smoothing model, multi-parameter seaconal model and gray model is respectively created;Step C:Historical data is directed respectively into the exponential smoothing model created in step B, multi-parameter seaconal model and gray model, plan table amount is calculated respectively;Step D:Choose Optimal calculation model of the minimum model of difference as metering table type to be calculated;Step E:Using the Optimal calculation model of the metering table type to be calculated obtained in step D, it is calculated with table demand.The present invention can be derived that the metering table demand of accurate period to be calculated, enable the enterprise to carry out scheduling of production according to the metering table demand of the accurate period to be calculated calculated.
Description
Technical field
Table demand computational methods, more particularly to a kind of meter based on data analysis are used the present invention relates to electric-power metering industry
Amount uses table demand computational methods.
Background technology
At present, after netting provincial electric-power metering center and setting up, mainly completed with the collection of table demand by provincial measurement centre.
Refer to provincial measurement centre with the collection of table demand and require that company of various regions cities and counties reports and submits plan table number in our unit's certain production cycle
The process of amount.With buying of the collection quality of table demand to annual goods and materials, calibrating and the smooth development, the provincial metering that dispense work
The quantity in stock and inventory structures at center, the quick response of demand order, the reasonability of Asset Allocation have significant impact.
At present in terms of measurement instrument demand management, most municipalities directly under the Central Government, net provincial company are matched somebody with somebody according to monthly progress measurement instrument
That send demand carries report, and next month measurement instrument demand is reported and submitted to provincial measurement centre from district electric company or electric company of districts and cities,
It is determined that measurement instrument is monthly dispensed by measurement centre after dispatching demand.Such a demand collection method lacks to measurement instrument demand
Analyse in depth, demand carry report it is more extensive, the subjective consciousness of demand management personnel is larger to the interference for reporting and submitting result accuracy, and
And that electric energy meter exceeds the time limit be present.Meanwhile part municipality directly under the Central Government and net provincial company power information collection construction project connect substantially
Draw to an end, the measurement instrument demand of Engineering will significantly decline, and the demand of measurement instrument is mainly derived from industry and expands new clothes, repairing
The naturality demands such as maintenance, have been further exacerbated by the difficulty of artificial prediction.
The content of the invention
, can be according to collection it is an object of the invention to provide a kind of metering table demand computational methods based on data analysis
Obtained metering table demand history data, metering table requirement forecasting analysis model under different mode is established by selection, counted
The metering table demand for drawing the accurate period to be calculated is calculated, is enabled the enterprise to accurate according to what is calculated
The metering table demand of period to be calculated carries out scheduling of production.
The present invention uses following technical proposals:
A kind of metering table demand computational methods based on data analysis, comprise the following steps:
Step A:Data acquisition is carried out with arranging, obtains metering table type to be calculated before chronomere to be calculated
Several chronomeres in historical data;
Step B:Model creation is carried out, exponential smoothing model, multi-parameter seaconal model and gray model is respectively created;
Step C:For metering table type to be calculated, the time nearest apart from chronomere to be calculated will be removed
Historical data outside unit, the exponential smoothing model created in step B, multi-parameter seaconal model and gray model are directed respectively into,
Metering table type to be calculated is calculated respectively when apart from chronomere to be calculated nearest one using three kinds of models
Between unit plan with table amount;
Step D:By the metering table type to be calculated being calculated in one apart from chronomere to be calculated recently
Chronomere plans with table amount, with metering table type to be calculated in historical data nearest apart from chronomere to be calculated
The actual of a chronomere be compared with table amount, choose the minimum model of difference as metering table type to be calculated
Optimal calculation model;
Step E:Using the Optimal calculation model of the metering table type to be calculated obtained in step D, distance will be included
The historical data in several chronomeres including a nearest chronomere of chronomere to be calculated, is imported to be calculated
The Optimal calculation model of metering table type, obtain metering table type to be calculated and use table demand in chronomere to be calculated
Amount.
Described metering table type to be calculated is divided into single-phase electric energy meter, three-phase electricity according to device class and specification of equipment
Can table, transformer and acquisition terminal, metering table type to be calculated according to equipment purposes be divided into new clothes table, replacing table and
Failure table.
In described step A, gone through by data pump installation according to imposing a condition to extract from sales service Database Systems
History data.
Described chronomere is divided into annual and monthly according to measurement period.
Described exponential smoothing model is:
Si=α xi+(1-α)Si-1;
Wherein, SiIt is preceding i issues according to the smooth value of the i-th issue evidence, Si-1It is preceding i-1 issues according to the i-th -1 issue evidence
Smooth value, xiFor the actual observed value of the i-th phase, i is natural number, and α is horizontal smoothing factor, and α span is [0,1].
Described multi-parameter seaconal model is:
Wherein, tiAnd ti-1Respectively using preceding i phases and preceding i-1 issues according to the smooth value to trend increment, piFor the i-th season phase
The exponential smoothing value of item is saved, k is seasonal periodicity length, and α, β and γ distinguish the smoothing factor of horizontal item, trend term and season item,
Horizontal smoothing factor, trend smoothing factor and seasonal exponential smooth are referred to as, α, β and γ span are [0,1],For the predicted value of i+h phases, pi-k+hFor the exponential smoothing value of the season item of the i-th-k+h phases, h is smooth issue backward.
Described gray model is:
Wherein, for a to develop grey number, u is grey actuating quantity.
The present invention is analyzed according to metering table demand history data by establishing measurement instrument requirement forecasting under different mode
Model, reversely verified using historical data and select optimal models, then be calculated by accurate historical data it is more accurate
Period to be calculated metering table demand, realize measurement instrument in the recent period and medium-long term demand the management and control that becomes more meticulous, make enterprise
Industry can carry out scheduling of production according to the metering table demand of the accurate period to be calculated calculated, ultimately form
The buying of measurement instrument and calibrating plan aid decision strategy, support provincial measurement centre measurement instruments at different levels directly with point and intelligence
The straight of all switch cabinets matches somebody with somebody pattern.
Brief description of the drawings
Fig. 1 is exponential smoothing model fitted figure in single-phase meter year new clothes table requirement forecasting in the present invention;
Fig. 2 shows for 2016 annual data predicted values in single-phase meter year new clothes table requirement forecasting in the present invention and forecast interval
It is intended to;
Fig. 3 in the present invention single-phase meter year new clothes with 2012-2015 annual datas in table requirement forecasting establish gray model
Fitting to historical data and to prediction effect figure in 2016;
Fig. 4 is exponential smoothing model fitted figure in the annual replacing table requirement forecasting of single-phase meter in the present invention;
Fig. 5 shows for 2016 annual data predicted values in single-phase meter year replacing table requirement forecasting in the present invention and forecast interval
It is intended to;
Fig. 6 establishes gray model for single-phase meter year replacing in the present invention with 2012-2015 annual datas in table requirement forecasting
Fitting to historical data and to prediction effect figure in 2016;
Fig. 7 is exponential smoothing model fitted figure in single-phase meter year failure table requirement forecasting in the present invention;
Fig. 8 shows for 2016 annual data predicted values in single-phase meter year failure table requirement forecasting in the present invention and forecast interval
It is intended to;
Fig. 9 in the present invention single-phase meter year failure with 2012-2015 annual datas in table requirement forecasting establish gray model
Fitting to historical data and to prediction effect figure in 2016;
Figure 10 is single-phase meter year new clothes table requirement forecasting displaying figure in the present invention;
Figure 11 is the annual replacing table requirement forecasting displaying figure of single-phase meter in the present invention;
Figure 12 is single-phase meter year failure table requirement forecasting displaying figure in the present invention;
Figure 13 is single-phase meter year total amount table requirement forecasting displaying figure in the present invention.
Embodiment
The present invention is made with detailed description below in conjunction with drawings and examples:
Metering table demand computational methods of the present invention based on data analysis, comprise the following steps:
Step A:Data acquisition is carried out with arranging, obtains metering table type to be calculated before chronomere to be calculated
Several chronomeres in historical data;
In step A, historical data is extracted from the actual, historical data that power marketing operation system shifts to an earlier date typing, the present invention
In, it can be extracted by data pump installation according to imposing a condition from sales service Database Systems.Wherein, wait to count in the present invention
The metering table type of calculation is divided into single-phase electric energy meter, three-phase electric energy meter, transformer and collection according to device class and specification of equipment
Terminal, metering table type to be calculated are divided into new clothes with table, replacing with table and failure table according to equipment purposes.Chronomere
Be divided into according to measurement period it is annual and monthly, i.e., be finally calculated correspond to year table demand and it is monthly use table demand.
Year, table demand can distinguish tabulating equipment purposes per year according to device class and specification of equipment during data preparation;It is monthly to be needed with table
Ask and distinguish monthly tabulating equipment purposes according to device class and specification of equipment.
Step B:Model creation is carried out, exponential smoothing model, multi-parameter seaconal model and gray model is respectively created;
Method currently used for prediction has a lot, is generally divided into qualitative forecasting and quantitative forecast.The method master of qualitative forecasting
There are Delphi method, subjective probability method, Scenario Prediction method;The method of quantitative forecast mainly has Regression Forecast, time series point
Solution, time series exponential smoothing, stationary time series predicted method etc..Exponential smoothing algorithm is due to simple in construction, general effect
The advantages that good, is widely used in the fields such as business, environmental science.It is logical too small amount of, incomplete for grey forecasting model
Information, founding mathematical models and a kind of method made prediction.Exponential smoothing, multi-parameter seaconal model and grey forecasting model
Required modeling information is few, and computing is convenient, and modeling accuracy is high, suffers from being widely applied in various prediction fields, is processing sample
The effective tool of this forecasting problem, suitable for table requirement forecasting.
The distinguishing feature of exponential smoothing algorithm is, (or actual to other predictions to newest observed value with the weight of maximum
Value) weight to successively decrease, so predicted value can reflect newest information, the information of and can reflecting history data, so that in advance
Result is surveyed more to tally with the actual situation.
Exponential smoothing algorithm belongs to non-statistical model, available for the content using the time as sequence analysis of deterministic type, refers to
The target of number exponential smoothing is to distinguish master data pattern and random fluctuation using " smooth " historical data.This is equivalent in history
Maximum value or minimum value is eliminated in data to obtain " smooth value " of the time series, i.e. the predicted value to future.Index
Smoothing prediction model is applied to the prediction of leveling style data, and in of the invention, described exponential smoothing model is single exponential smoothing,
It is as follows:
Si=α xi+(1-α)Si-1;
Wherein, SiIt is preceding i issues according to the smooth value of the i-th issue evidence, Si-1It is preceding i-1 issues according to the i-th -1 issue evidence
Smooth value, xiFor the actual observed value of the i-th phase, i is natural number, and α is horizontal smoothing factor, and α span is [0,1], right
In the smoothing factor α speed for acting as controlling flexible strategy to decline, α is closer to 1, then the weight of recent observation is bigger;Conversely,
α is closer to 0, then the weight of history observation is bigger.To optimize certain fit standard, α actual value is typically by computer
Selection, common fit standard is the residual sum of squares (RSS) between actual value and predicted value.The preceding i issues occurred above and below
According to the data in i chronomere before being.
In time series, it would be desirable to which based on the time series, currently existing data predict it in walking afterwards
Gesture, multi-parameter seasonality algorithm (third index flatness, Holt-Winters) can be very good to carry out the prediction of time series.
The algorithm can be predicted containing trend and seasonal time series pair simultaneously, be to be based on single exponential smoothing and secondary finger
Number smoothing algorithm.
Single exponential smoothing algorithm has been described above, and Secondary Exponential Smoothing Method remains the information of trend, makes
The trend of data before the time series that must be predicted can include.Double smoothing is by adding a new variable t come table
Trend after showing smoothly, calculation formula are as follows:
Si=α xi+(1-α)(Si-1+ti-1)
ti=β (Si-Si-1)+(1-β)ti-1
Wherein xiFor the actual observed value of the i-th phase, SiAnd Si-1Respectively preceding i phases and preceding i-1 issues according to the i-th phase and i-th-
The smooth value of 1 phase, tiAnd ti-1Respectively using preceding i phases and preceding i-1 issues according to the smooth value to trend increment, smoothing factor α and
β distinguishes the exponential smoothing of controlled level item and trend term, is referred to as horizontal smoothing factor and trend smoothing factor, in addition α and β
Span be [0,1],For i+h predicted value, h is the issue of prediction extrapolation.
And Three-exponential Smoothing remains seasonal information on the basis of double smoothing so that it can be predicted
With seasonable time series.Three-exponential Smoothing with the addition of a new parameter p come the trend after representing smooth, the present invention
In, multi-parameter seaconal model is:
Wherein, tiAnd ti-1Respectively using preceding i phases and preceding i-1 issues according to the smooth value to trend increment, piFor the i-th season phase
The exponential smoothing value of item is saved, k is seasonal periodicity length, and for monthly data k=12, γ is the smoothing factor of season item, is referred to as
Seasonal exponential smooth, γ span is [0,1],For the predicted value of i+h phases, wherein pi-k+hFor the season of the i-th-k+h phases
The exponential smoothing value of item is saved, h is smooth issue backward.
In addition, for S, t, the selection of p initial values is not especially big for the overall influence of algorithm, and common value is S0
=x0, t0=x1-x0, p0=0.And for smoothing factor α, β and γ of horizontal component, trend part and season part, its parameter
Value means that more greatly the weight of recent observation is bigger.
In the present invention, gray model is:
Wherein, for a to develop grey number, u is grey actuating quantity.
Gray model (GM) is exactly to lead to too small amount of, imperfect information, establishes grey differential prediction model, and things is sent out
Exhibition rule makes the long-term description (the prediction science branch theoretical in fuzzy prediction field, method is more perfect) of ambiguity.GM tables
Show the grey differential equation model of gray theory.GM (1,1) is the grey differential equation model of one variable of single order.GM (1,1) is predicted
Model is a kind of the most frequently used gray dynamic Prediction model, and its modeling principle is:
Provided with one group of original series:x(0)=(x(0)(1),x(0)(2),....,x(0)(n))
Original series are made with monovalence Accumulating generation, obtains x(1)=(x(1)(1),x(1)(2),....,x(1)(n))
Wherein:
Remake x(1)Single order average generation, obtain
X=(x (2), x (3) ... .x (n))
Wherein:(the x of x (k)=- 1/2(1)(k-1)+x(1)(k)) k=1,2,3 ... .., n constitute Grey Simulation, can established
Gray model, the general expression of GM (1,1) model are:
Wherein a represents the developing state of behavior sequence estimate, u is grey actuating quantity, and reflection is several to develop grey number
According to the relation of change.
This differential equation is solved to obtain:
Parameter a in formula, μ can be tried to achieve by least square method:
Wherein:
Reduce to obtain x by regressive(0)Forecast model be
GM (1,1) model generally use after-test residue checking is tested to model progress, and it is mainly to residual distribution
Statistical nature test, basis for estimation is variance ratio C and small error possibility p, variance ratio C=S2/S1Wherein absolute error sequence
The standard deviation of rowWith original series standard deviationSmall error possibility be p=P (| Δ(0)(k)-mean
(Δ(0)) | < 0.6745S1)。
Posterior difference examination differentiates as shown in table 1 with reference to table:
Table 1
P | C | Model accuracy |
>0.95 | <0.35 | It is good |
>0.80 | <0.50 | It is qualified |
>0.70 | <0.65 | Scrape through |
<0.70 | >0.65 | It is unqualified |
Step C:For metering table type to be calculated, the time nearest apart from chronomere to be calculated will be removed
Historical data outside unit, the exponential smoothing model created in step B, multi-parameter seaconal model and gray model are directed respectively into,
Metering table type to be calculated is calculated respectively when apart from chronomere to be calculated nearest one using three kinds of models
Between unit plan with table amount;
Step D:By the metering table type to be calculated being calculated in one apart from chronomere to be calculated recently
Chronomere plans with table amount, with metering table type to be calculated in historical data nearest apart from chronomere to be calculated
The actual of a chronomere be compared with table amount, choose the minimum model of difference as metering table type to be calculated
Optimal calculation model;
Step E:Using the Optimal calculation model of the metering table type to be calculated obtained in step D, distance will be included
The historical data in several chronomeres including a nearest chronomere of chronomere to be calculated, is imported to be calculated
The Optimal calculation model of metering table type, obtain metering table type to be calculated and use table demand in chronomere to be calculated
Amount.
Below in conjunction with specific embodiment, respectively to calculate the new clothes table requirement forecasting of single-phase meter year, single-phase meter year more
Exemplified by using table requirement forecasting and the failure table requirement forecasting of single-phase meter year instead, metering table of the present invention is expanded on further
Demand computational methods.
Step A:Data acquisition is carried out first with arranging, the different year installation statistics with table type of statistics single-phase meter,
As shown in table 2.In the present embodiment, metering table type to be calculated is respectively new clothes table, replacing table and failure table, always
It is used to refer to table amount, chronomere is year.
Table 2
Step B:Model creation is carried out, exponential smoothing model, multi-parameter seaconal model and gray model is respectively created;
Step C:For new clothes table, replacing table and failure these three types of table, by the historical data outside 2016
The historical data of i.e. 2012 to 2015, exponential smoothing model, multi-parameter seaconal model and gray model are directed respectively into, calculated
New clothes table, replacing table and these three types of failure table are obtained in 2016 to plan with table amount;
Step D:By the plan of the new clothes table being calculated, replacing table and failure table these three types in 2016
With table amount, respectively compared with these three in historical data are with actual with table amount of the table type in 2016, difference is chosen most
Optimal calculation model of the small model as metering table type to be calculated;
1. single-phase meter year new clothes table requirement forecasting:
(1) exponential smoothing model
It is as shown in Figure 1 to the fit solution of historical data with 2012-2015 annual data onset index smoothing models:
α estimate is 6.610696e-05 in exponential smoothing model, very close 0, illustrate that the sequence is more steady, go through
History data fluctuate near some value.
By model to 2016 annual data predicted values and forecast interval as shown in Fig. 2 and table 3.
Table 3
Time | Observed value | Prediction | Relative error | Lo 80 | Hi 80 |
2016 | 793714 | 791479 | 0.28% | 712788 | 870170 |
From the point of view of prediction result, relative error only has 0.28%, 80% forecast interval be in 712788 and 870170 it
Between, illustrate that prediction effect is fine.
(2) gray model
Fitting of the gray model to historical data is established and to prediction effect such as Fig. 3 in 2016 with 2012-2015 annual datas
It is shown;
Model result exports:
GM (1,1) estimates of parameters:
Development coefficient-a=0.06290011;Grey actuating quantity u=686536.9;
X (0) analogue value:
791475;759971.4;809309.1;861849.8;917801.5;
Residual sum of squares (RSS)=2218081580;
Average relative error=3.131143%;
Relative accuracy=96.86886%;
Posteriority difference ratio test:
C value=0.3578659;
C values belong to [0.35,0.5], and GM (1,1) model prediction accuracy grade is:It is qualified.
Table 4
Time | Observed value | Predicted value | Relative error |
2016 | 793714 | 917801 | 15.63% |
Can be seen that posteriority difference ratio examines C values from model data result is 0.36, is belonged between [0.35,0.5], model is pre-
Accuracy class is surveyed to be qualified, is 15.63% to Relative Error in 2016.Compared with exponential smoothing model, effect is relatively
Difference, it is proposed that using exponential smoothing model.
2. single-phase meter year, which is changed, uses table requirement forecasting:
(1) exponential smoothing model
It is as shown in Figure 4 to the fit solution of historical data with 2012~2015 annual data onset index smoothing prediction models:
α estimate is 0.9999339, very close 1 in exponential smoothing model, illustrates the recent observation of the sequence pre-
Weight in survey is very high.
By model to 2016 annual data predicted values and forecast interval as shown in Fig. 5 and table 5:
Table 5
Time | Observed value | Prediction | Relative error | Lo 80 | Hi 80 |
2016 | 2643331 | 2518100 | 4.74% | 1800077 | 3236123 |
From the point of view of prediction result, relative error 4.74%, 80% forecast interval be in 1800077 and 3236123 it
Between, although Relative Error is smaller, because data stochastic volatility is larger, cause forecast interval scope accurately poor.
(2) gray model
Fitting of the gray model to historical data is established and to prediction effect such as Fig. 6 in 2016 with 2012-2015 annual datas
It is shown.
Model result exports
GM (1,1) estimates of parameters:
Development coefficient-a=0.3400266;Grey actuating quantity u=972942.3;
X (0) analogue value:
459995;1345103;1889850;2655211;3730531;
Residual sum of squares (RSS)=307211581386;
Average relative error=17.93661%;
Relative accuracy=82.06339%;
Posteriority difference ratio test:
C value=0.223116
C values<0.35, GM (1,1) precision of prediction grade is:It is good
Table 6
Time | Observed value | Prediction | Relative error |
2016 | 2643331 | 3730531 | 41.13% |
It can be seen that posteriority difference than examining C values as 0.22, less than 0.35, model prediction accuracy grade from model data result
Preferably, it is 41.13% to Relative Error in 2016, although models fitting effect is preferable, prediction effect is poor, explanation
The model is not suitable for the prediction of the data, it is proposed that using exponential smoothing model.
3. the failure table requirement forecasting of single-phase meter year,
(1) exponential smoothing model
It is as shown in Figure 7 to the fit solution of historical data with 2012~2015 annual data onset index smoothing prediction models:
α estimate is about 0.95 in exponential smoothing model, and very close 1, illustrate that the recent observation of the sequence is being predicted
In weight it is very high.
By model to 2016 annual data predicted values and forecast interval as shown in Fig. 8 and table 7 below:
Table 7
Time | Observed value | Prediction | Relative error | Lo 80 | Hi 80 |
2016 | 68747 | 71994 | 4.72% | 68050 | 75938 |
From the point of view of prediction result, relative error 4.72%, 80% forecast interval is between 68050 and 75938, is said
Bright forecast result of model is fine.
(2) gray model
Fitting of the gray model to historical data is established and to prediction effect such as Fig. 9 in 2016 with 2012-2015 annual datas
It is shown.
Model result exports:
GM (1,1) estimates of parameters:
Development coefficient-a=-0.002805841;Grey actuating quantity u=72672.24;
X (0) analogue value:
67147;72382.24;72179.43;71977.19;71775.52;
Residual sum of squares (RSS)=458.1893;
Average relative error=0.01616526%;
Relative accuracy=99.98383%;
Posteriority difference ratio test:
C value=0.003258309;
C values<0.35, GM (1,1) precision of prediction grade is:It is good
Table 8
Time | Observed value | Prediction | Relative error |
2016 | 68747 | 71776 | 4.41% |
It can be seen that posteriority difference than examining C values as 0.003, less than 0.35, model prediction accuracy etc. from model data result
Level is 4.41% to Relative Error in 2016, prediction effect is good preferably.
The prediction effect of correlation index's smoothing model and gray model, both of which have preferable prediction effect, and both of which can
To use.
From the point of view of exponential smoothing model and gray model are to single-phase meter requirement forecasting effect, exponential smoothing model prediction effect
Relative error is relatively low, and relative accuracy is higher, and precision of prediction illustrates that exponential smoothing model is adapted to single-phase meter year all more than 95%
Requirement forecasting.
Step E:Utilize new clothes table, the optimal meter of replacing table and failure table these three types obtained in step D
Model is calculated, the history in several chronomeres including a chronomere nearest apart from chronomere to be calculated will be included
Data, the data of i.e. 2012 to 2016 years in the present embodiment, import in Optimal calculation model, obtain metering table class to be calculated
Type is in the use table demand that chronomere to be calculated is 2017.
(1) single-phase meter year new clothes forecast analysis, as shown in Figure 10:
Based on 2012~2016 data single-phase meter new clothes annual predictions
Note:Predicted value 1 is predicted with multi-parameter seaconal model;
Predicted value 2 is predicted with gray model.
(2) single-phase meter year, which is changed, uses table requirement forecasting, as shown in figure 11:
Annual prediction is changed based on 2012~2016 data single-phase meters
Time | Observed value | Predicted value 1 | Predicted value 2 |
2012 | 459995 | 2499707 | 1112239 |
2013 | 1039253 | 1115760 | 1407269 |
2014 | 2331292 | 1676896 | 1780558 |
2015 | 2518112 | 3123941 | 2252864 |
2016 | 2643331 | 3167259 | 2850452 |
2017 | 3168376 | 3606556 |
Note:Predicted value 1 is predicted with multi-parameter seaconal model;
Predicted value 2 is predicted with gray model.
3. the failure table requirement forecasting of single-phase meter year, as shown in figure 12:
Based on 2012~2016 data single-phase meter failure annual predictions
Note:Predicted value 1 is predicted with multi-parameter seaconal model;
Predicted value 2 is predicted with gray model.
4. single-phase meter year " total amount " forecast analysis, as shown in figure 13:
Based on 2012~2016 data single-phase meter " total amount " annual predictions
Note:Predicted value 1 is predicted with multi-parameter seaconal model;
Predicted value 2 is predicted with gray model.
Claims (7)
1. a kind of metering table demand computational methods based on data analysis, it is characterised in that comprise the following steps:
Step A:Data acquisition is carried out with arranging, if obtaining metering table type to be calculated before chronomere to be calculated
Historical data in Gan Ge chronomeres;
Step B:Model creation is carried out, exponential smoothing model, multi-parameter seaconal model and gray model is respectively created;
Step C:For metering table type to be calculated, a chronomere nearest apart from chronomere to be calculated will be removed
Outer historical data, the exponential smoothing model created in step B, multi-parameter seaconal model and gray model are directed respectively into, utilized
Metering table type to be calculated is calculated in a time list nearest apart from chronomere to be calculated in three kinds of models respectively
Plan with table amount position;
Step D:By the metering table type to be calculated being calculated in the time nearest apart from chronomere to be calculated
Unit is planned with table amount, with metering table type to be calculated in historical data apart from chronomere to be calculated it is nearest one
The actual of individual chronomere is compared with table amount, chooses the minimum model of difference as metering table type to be calculated most
Excellent computation model;
Step E:Using the Optimal calculation model of the metering table type to be calculated obtained in step D, will wait to count comprising distance
The historical data in several chronomeres including a nearest chronomere of evaluation time unit, imports metering to be calculated
With the Optimal calculation model of table type, obtain metering table type to be calculated and use table demand in chronomere to be calculated.
2. the metering table demand computational methods according to claim 1 based on data analysis, it is characterised in that:Described
Metering table type to be calculated according to device class and specification of equipment be divided into single-phase electric energy meter, three-phase electric energy meter, transformer and
Acquisition terminal, metering table type to be calculated are divided into new clothes with table, replacing with table and failure table according to equipment purposes.
3. the metering table demand computational methods according to claim 1 based on data analysis, it is characterised in that:Described
In step A, historical data is extracted from sales service Database Systems according to imposing a condition by data pump installation.
4. the metering table demand computational methods according to claim 1 based on data analysis, it is characterised in that:Described
Chronomere is divided into annual and monthly according to measurement period.
5. the metering table demand computational methods according to claim 1 based on data analysis, it is characterised in that described
Exponential smoothing model is:
Si=α xi+(1-α)Si-1;
Wherein, SiIt is preceding i issues according to the smooth value of the i-th issue evidence, Si-1It is preceding i-1 issues according to the smooth of the i-th -1 issue evidence
Value, xiFor the actual observed value of the i-th phase, i is natural number, and α is horizontal smoothing factor, and α span is [0,1].
6. the metering table demand computational methods according to claim 5 based on data analysis, it is characterised in that:Described
Multi-parameter seaconal model is:
Si=α (xi-pi-k)+(1-α)(Si-1+ti-1)
ti=β (Si-Si-1)+(1-β)ti-1
pi=γ (xi-Si)+(1-γ)pi-k
Wherein, tiAnd ti-1Respectively using preceding i phases and preceding i-1 issues according to the smooth value to trend increment, piFor the i-th season phase item
Exponential smoothing value, k is seasonal periodicity length, and α, β and γ distinguish the smoothing factor of horizontal item, trend term and season item, respectively
Referred to as horizontal smoothing factor, trend smoothing factor and seasonal exponential smooth, α, β and γ span are [0,1],For
The predicted value of i+h phases, pi-k+hFor the exponential smoothing value of the season item of the i-th-k+h phases, h is smooth issue backward.
7. the metering table demand computational methods according to claim 6 based on data analysis, it is characterised in that described
Gray model is:
<mrow>
<mfrac>
<mrow>
<msup>
<mi>dx</mi>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</msup>
</mrow>
<mrow>
<mi>d</mi>
<mi>t</mi>
</mrow>
</mfrac>
<mo>+</mo>
<msup>
<mi>ax</mi>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</msup>
<mo>=</mo>
<mi>u</mi>
<mo>;</mo>
</mrow>
Wherein, for a to develop grey number, u is grey actuating quantity.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711077313 | 2017-11-06 | ||
CN2017110773137 | 2017-11-06 |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107862476A true CN107862476A (en) | 2018-03-30 |
Family
ID=61704706
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711268203.9A Pending CN107862476A (en) | 2017-11-06 | 2017-12-05 | A kind of metering table demand computational methods based on data analysis |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107862476A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111506876A (en) * | 2020-04-02 | 2020-08-07 | 沈阳数云科技有限公司 | Data prediction analysis method, system, equipment and readable storage medium |
CN111967684A (en) * | 2020-08-27 | 2020-11-20 | 北京合众伟奇科技有限公司 | Metering asset active distribution method based on big data analysis |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103617468A (en) * | 2013-12-13 | 2014-03-05 | 李敬泉 | Medium and long term predication method for necessities of life |
CN103617548A (en) * | 2013-12-06 | 2014-03-05 | 李敬泉 | Medium and long term demand forecasting method for tendency and periodicity commodities |
CN104268651A (en) * | 2014-09-28 | 2015-01-07 | 江南大学 | Seasonal energy consumption data forecasting method based on wavelet multi-scale cubic exponential smoothing models |
CN104598986A (en) * | 2014-12-12 | 2015-05-06 | 国家电网公司 | Big data based power load prediction method |
CN104899451A (en) * | 2015-06-12 | 2015-09-09 | 哈尔滨工业大学 | Time sequence prediction model selecting method based on statistical signal process |
CN105447594A (en) * | 2015-11-17 | 2016-03-30 | 福州大学 | Electric power system grey load prediction method based on exponential smoothing |
CN105512447A (en) * | 2014-09-26 | 2016-04-20 | 山西云智慧科技股份有限公司 | Bus passenger volume prediction method based on Holt-Winters model |
CN107292428A (en) * | 2017-06-07 | 2017-10-24 | 国网浙江省电力公司物资分公司 | A kind of distribution Power Material procurement demand forecasting system |
-
2017
- 2017-12-05 CN CN201711268203.9A patent/CN107862476A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103617548A (en) * | 2013-12-06 | 2014-03-05 | 李敬泉 | Medium and long term demand forecasting method for tendency and periodicity commodities |
CN103617468A (en) * | 2013-12-13 | 2014-03-05 | 李敬泉 | Medium and long term predication method for necessities of life |
CN105512447A (en) * | 2014-09-26 | 2016-04-20 | 山西云智慧科技股份有限公司 | Bus passenger volume prediction method based on Holt-Winters model |
CN104268651A (en) * | 2014-09-28 | 2015-01-07 | 江南大学 | Seasonal energy consumption data forecasting method based on wavelet multi-scale cubic exponential smoothing models |
CN104598986A (en) * | 2014-12-12 | 2015-05-06 | 国家电网公司 | Big data based power load prediction method |
CN104899451A (en) * | 2015-06-12 | 2015-09-09 | 哈尔滨工业大学 | Time sequence prediction model selecting method based on statistical signal process |
CN105447594A (en) * | 2015-11-17 | 2016-03-30 | 福州大学 | Electric power system grey load prediction method based on exponential smoothing |
CN107292428A (en) * | 2017-06-07 | 2017-10-24 | 国网浙江省电力公司物资分公司 | A kind of distribution Power Material procurement demand forecasting system |
Non-Patent Citations (2)
Title |
---|
卓金武 等: "《MATLAB在数学建模中的应用》", 30 September 2014, 北京:北京航空航天大学出版社 * |
杨东援 等: "《连续数据环境下的交通规则与管理》", 31 January 2014, 上海:同济大学出版社 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111506876A (en) * | 2020-04-02 | 2020-08-07 | 沈阳数云科技有限公司 | Data prediction analysis method, system, equipment and readable storage medium |
CN111506876B (en) * | 2020-04-02 | 2023-09-12 | 沈阳数云科技有限公司 | Data predictive analysis method, system, equipment and readable storage medium |
CN111967684A (en) * | 2020-08-27 | 2020-11-20 | 北京合众伟奇科技有限公司 | Metering asset active distribution method based on big data analysis |
CN111967684B (en) * | 2020-08-27 | 2024-02-02 | 北京合众伟奇科技股份有限公司 | Metering asset active distribution method based on big data analysis |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Meng et al. | Degree-day based non-domestic building energy analytics and modelling should use building and type specific base temperatures | |
Chen et al. | Project delivery system selection of construction projects in China | |
CN104484723B (en) | A kind of power transformer Forecast of Economic Life method based on lifetime data | |
CN105139090A (en) | Power industry safety stock decision analysis method based on consumption prediction | |
Kulikova et al. | Planning of technological development of new products and its impact on the economic performance of the enterprise | |
CN104574002A (en) | Asset management information processing method based on big data analysis and asset management information processing device based on big data analysis | |
CN103985000A (en) | Medium-and-long term typical daily load curve prediction method based on function type nonparametric regression | |
CN110991985A (en) | Project progress data processing method | |
CN111199493A (en) | Arrearage risk identification method based on customer payment information and credit investigation information | |
CN108734567A (en) | A kind of asset management system and its appraisal procedure based on big data artificial intelligence air control | |
CN110322121A (en) | A kind of power supply enterprise's customer satisfaction appraisal procedure | |
Perekrest et al. | Principles of designing and functioning of the expert system of assessing the efficiency of introducing energy conservation measures | |
CN107862476A (en) | A kind of metering table demand computational methods based on data analysis | |
Konior et al. | Course of planned, actual and earned cost curves of diverse construction investments | |
CN116188198A (en) | Water supply network operation state diagnosis method and computer readable storage medium | |
CN106682840A (en) | Short-term power sale quantity predicting method and system based on daily cumulative distributed electric quantity | |
England et al. | Advanced forecasting techniques | |
Sinitsyna et al. | Some practical aspects of electric power consumption time series analysis | |
Klassen et al. | Quantifiying the Business Impact of Information Quality-a Risk-Based Approach | |
CN109598421A (en) | Charging station operation data processing method and processing device | |
CN104156602B (en) | The method and apparatus for creating power grid Calculating model | |
Hassan | Power Plant Operation Optimization: Unit Commitment of Combined Cycle Power Plants Using Machine Learning and MILP | |
Del Bufalo et al. | Innovative Instruments and Techniques for Measurement System Operation Management | |
Sameye et al. | Proposing a mathematical model of parallel machine scheduling with the aim of minimizing the task completion time and energy cost using a meta-heuristic algorithm. | |
Setiawan et al. | Optimization of the Use of Artificial Neural Network Models for Accuracy Data Measurement Palm Oil Production Prediction Rate |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180330 |
|
RJ01 | Rejection of invention patent application after publication |