CN106934482A - Using 24 hours electric power demand forecasting devices and electric power demand forecasting method of the temperature for temporally dividing - Google Patents

Abstract

24 hours electric power demand forecasting devices and electric power demand forecasting method of the temperature for temporally dividing are the present invention relates to the use of, electric power demand forecasting method according to an embodiment of the invention includes：Estimated temperature receiving step, electric power demand forecasting device receives estimated temperature record corresponding with the temperature forecast of target day in the step；Similar day extracts step out, and with the temperature historical data base for having set be compared above-mentioned estimated temperature record by electric power demand forecasting device in the step, and extracts similar day corresponding with the temperature Change characteristic curve of above-mentioned estimated temperature record out；And prediction electricity needs characteristic curve generation step, electric power demand forecasting device extracts the measure electricity needs characteristic curve determined in above-mentioned similar day out from the electricity needs database for having set in the step, and the prediction electricity needs characteristic curve of above-mentioned target day is generated using said determination electricity needs characteristic curve.

Description

Using 24 hours electric power demand forecasting devices and electricity needs of the temperature for temporally dividing Forecasting Methodology

Technical field

The present invention relates to electric power demand forecasting device and electric power demand forecasting method on 24 hours next day (Apparatus and method for 24hour electrical load forecasting), more particularly to utilizes For the temperature forecast for temporally dividing of next day prediction, retrieval is needed with the characteristic past electric power of similar temperature Change Ask, and be used as entering data to the electric power demand forecasting device and electric power demand forecasting method of predicting electricity needs.

Background technology

Electric power demand forecasting is important for stable and smoothly power system operation and power supply and demand planning foundation Key element.Electric power demand forecasting result is used to determine power price or power system operation, thus electric power demand forecasting error It is likely to become the reason for hindering stable power system to run and cause huge economic losses.So as to be needed to reduce electric power Seek the error of prediction, it is proposed that have time series analysis method, regression analysis, ANN, Knowledge based engineering expert be The various electric power demand forecasting methods such as system.

However, conventional electric power demand forecasting method major part is merely using only the highest temperature for dividing by date and most Low temperature is predicted, therefore is difficult the characteristic of the fully reflection electricity needs sensitive to temperature.For example, in the gas for making an exception In the case of warm variation characteristic curve, electricity needs characteristic curve is possible to different according to temperature Change, but only by conventional Forecasting Methodology be difficult consider this temperature Change come the electricity needs that calculates to a nicety.So as to exist and make an exception in the past Temperature Change date occur the big grade of electric power demand forecasting error problem.

The content of the invention

The present invention provides a kind of electricity needs that electricity needs is predicted using the temperature Change characteristic curve for temporally dividing Prediction meanss and electric power demand forecasting method.

Electric power demand forecasting method according to an embodiment of the invention includes：Estimated temperature receiving step, electricity in the step Power demand-prediction device receives estimated temperature record corresponding with the temperature forecast of target day；Similar day extracts step, the step out With the temperature historical data base for having set be compared above-mentioned estimated temperature record by middle electric power demand forecasting device, and extract out with The temperature Change characteristic curve corresponding similar day of above-mentioned estimated temperature record；And prediction electricity needs characteristic curve generation step Suddenly, electric power demand forecasting device extracts the survey determined in above-mentioned similar day out from the electricity needs database for having set in the step Determine electricity needs characteristic curve, and the prediction electric power of above-mentioned target day generated using said determination electricity needs characteristic curve to need Seek characteristic curve.

Here, in extracting step above-mentioned similar day out, model is extracted out using the similar day for having set, storage is calculated above-mentioned Each in temperature historical data base is pressed by the similar degree between the temperature record and above-mentioned estimated temperature record that determine bu The measure day for having set number is extracted out as above-mentioned similar day according to above-mentioned similar degree order high.

Here, above-mentioned similar day extracts model out being

In any one, and calculate Euclidean distance using model is extracted above-mentioned similar day out, thus determine similar degree. This, Δ T_max=T^f _max-T^p _max, Δ T_min=T^f _min-T^p _min, T^f _maxIt is the highest temperature of target day, T^p _maxIt is the highest for determining day Temperature, T^f _minIt is the lowest temperature of target day, T^p _minIt is the lowest temperature for determining day.And, TC=1-R, R are targets day and survey Coefficient correlation between the temperature on settled date, Δ slope_m=| slope^f _m-slope^p _m|, Δ slope_a=| slope^f _a-slope^p _a |, slope^f _mIt is the morning temperature slope of target day, slope^p _mIt is the morning temperature slope for determining day, slope^f _aIt is target day Afternoon temperature slope, slope^p _aIt is the temperature slope in afternoon for determining day.And, Δ H_max=H^f _max-H^p _max, Δ H_min=H^f _min- H^p _min, H^f _maxIt is the highest temperature time of origin of target day, H^p _maxIt is the highest temperature time of origin for determining day, H^f _minIt is target The lowest temperature time of origin of day, H^p _minIt is the lowest temperature time of origin for determining day, T^f _hIt is the h point temperature of target day, T^p _hIt is The h point temperature of day is determined, h is the positive number of satisfaction 0≤h≤24.In addition, ω₁₁、ω₁₂、ω₁₃、ω₁₄、ω₁₅It is weighted value, it is above-mentioned Weighted value is applicable each regression equation and is calculated by least square method.

Here, above-mentioned prediction electricity needs characteristic curve generation step includes：Exponential smoothing process, it is for above-mentioned extraction Multiple electricity needs characteristic curves that determine be applicable exponential smoothings, generate exponential smoothing demand characteristics curve；Normalization process, It carries out normalization to above-mentioned exponential smoothing demand characteristics curve, generates regular demand characteristic curve；And characteristic curve generation Process, it receives the maximum power requirement forecasting value and minimum power requirement forecasting value about above-mentioned target day, and for above-mentioned Regular demand characteristic curve is applicable above-mentioned maximum power requirement forecasting value and minimum power requirement forecasting value, generates above-mentioned prediction electricity Power demand characteristics curve.

Here, above-mentioned exponential smoothing process generates above-mentioned exponential smoothing demand characteristics curve using following calculating formula, i.e.,

Wherein, SDL^p _hIt is the temporal exponential smoothing requirements of h of target day, SDLⁿ _hIt is class that similar degree is n-th high Like the temporal electricity needs values of h of day, α is exponential smoothing weighted value.

Here, above-mentioned normalization process generates above-mentioned regular demand characteristic curve using following calculating formula, i.e.,

Wherein, PU_SDL^p _hBe target day the temporal normalizations of h after regular demand value, SDL^p _hIt is the h of target day Temporal exponential smoothing requirements, SDL^p _maxIt is the maximum of the exponential smoothing requirements of target day, SDL^p _minIt is target day The minimum value of exponential smoothing requirements.

Above-mentioned characteristic curve generating process generates above-mentioned prediction electricity needs characteristic curve using following calculating formula, i.e.,

Wherein,It is the temporal prediction electricity needs values of h of target day, PU_SDL^p _hWhen being the h of above-mentioned target day Between on normalization after regular demand value,It is above-mentioned maximum power requirement forecasting value,It is above-mentioned minimum power Requirement forecasting value.

Electric power demand forecasting device according to an embodiment of the invention, including：Data reception portion, it is received and target day The corresponding estimated temperature record of temperature forecast；Similar day search part, it goes through above-mentioned estimated temperature record and the temperature for having set History database is compared, and extracts similar day corresponding with the temperature Change characteristic curve of above-mentioned estimated temperature record out；And Demand characteristics curve generating unit, it extracts the measure electric power determined in above-mentioned similar day out from the electricity needs database for having set Demand characteristics curve, and the prediction electricity needs characteristic curve of above-mentioned target day is generated using said determination electricity characteristic curve.

Additionally, the solution of above-mentioned problem is not to enumerate all features of the invention.Various features of the invention And its advantage brought and effect will become understanding in more detail by referring to following specific embodiment.

Electric power demand forecasting device according to an embodiment of the invention and electric power demand forecasting method, it is contemplated that temperature Change is come the electricity needs that calculates to a nicety.So as to even if on the date of the temperature Change for making an exception, it is also possible to perform accurate Electric power demand forecasting.

Brief description of the drawings

Fig. 1 is the block diagram for representing electric power demand forecasting device according to an embodiment of the invention.

Fig. 2 is the characteristic curve map of temperature Change for representing general meteorological day according to an embodiment of the invention.

Fig. 3 is the characteristic curve map of temperature Change for representing exception according to an embodiment of the invention meteorological day.

Fig. 4 is the expression characteristic curve map of electricity needs according to an embodiment of the invention.

Fig. 5 predicts the characteristic curve map of electricity needs to represent according to an embodiment of the invention.

Fig. 6 to Fig. 8 is the precedence diagram for representing electric power demand forecasting method according to an embodiment of the invention.

Description of reference numerals is as follows：

100：Electric power demand forecasting device 110：Data reception portion

120：Similar day search part 130：Demand characteristics curve generating unit

200：Temperature historical data base 300：Electricity needs database

S100：Estimated temperature receiving step S200：Similar day extracts step out

S300：Prediction electricity needs characteristic curve generation step

S310：Exponential smoothing process S320：Normalization process

S330：Characteristic curve generating process

Specific embodiment

Hereinafter, preferred embodiment is explained in detail with reference to the accompanying drawings, so that those skilled in the art easily implement the present invention.Its In, when the preferred embodiments of the present invention are described in detail, when the detailed description for judging to be directed to related known function or composition has Idea of the invention may be caused it is unnecessary obscure in the case of, will description is omitted.Additionally, similar on playing Function and the part of effect, use identical reference in all of the figs.

And, throughout the specification, when certain part is with another part " connection ", not only represent the feelings of " being directly connected to " Condition, also including between the two across the situation of another element " being indirectly connected with ".And, " including " certain inscape is that do not having During special opposite record, it is meant that be not precluded from other inscapes but be also possible that other inscapes.And, say Terms such as " portions ", " module " described in bright book refers to the unit for processing at least one function or action, and this can be by hardware Or software, or the combination of hardware and software realizes.

In general, 24 hours temperature Change characteristic curves, as shown in Fig. 2 after the time in morning reaches the lowest temperature, with Sunrise is gradually increasing, and declines again after reaching the highest temperature before and after 15 points~16 points.That is, general meteorology day is unrelated with season Ground, with the time in morning occur the lowest temperature and in the afternoon the time generation highest temperature temperature Change characteristic curve.

But, difference according to circumstances, it is also possible to there is the highest temperature and occur before sunrise or after the sunset.Example Such as, as Fig. 3 (a) on December in 2014 16 Soul the temperature Change divided on a time period shown in, betide icepro in the highest temperature In the case of time period in morning, temperature Change characteristic curve has the form of continuous decrease.Additionally, such as in the December, 2014 of Fig. 3 (b) Shown in the temperature Changes divided on a time period of Soul on the 14th, when the highest temperature is betided after the sunset, it is possible to which temperature occurs Situations such as variation characteristic curve has the form of lasting rising.

In addition, Fig. 4 is represented occurring 24 hours of 16 days December in 2014 and its periphery date for declining atmospheric temperature property curve Electricity needs and temperature.In the case of 16 days December in 2014, there is the special report of cold current in the whole nation, there occurs big portion it is cloudy or according to The difference of region occurs in that severe snow.On the contrary, on December 14th, 2014, there is southern low pressure, continue for compared to other days Warm weather for phase.Two determine days on the period for belong to winter, but there occurs unusual weather, quite In the situation that there occurs temperature continuous decrease or rising, this temperature Change is produced to the electricity needs characteristic curve for temporally dividing Considerable influence.

Specifically, reference picture 4,16 days December in 2014 for occurring to decline atmospheric temperature property curve is maximum power demand hair It is born at 18 points.But, 15 days December in 2014, on December 19,17 days to 2014 December in 2014 in addition to the date Maximum power demand betides at 10 points and 11 points.That is, the electricity needs characteristic curve on December 16 is holding because of temperature within 2014 Continuous decline increased the electricity needs in afternoon.Furthermore, it is possible to it is general to confirm that on December 16th, 2014 there occurs again afterwards Atmospheric temperature property curve, electricity needs characteristic curve returns to the form that maximum power demand occurs in the morning.So as to, if will If the electricity needs on December 16 is predicted according to conventional same way within 2014, then cannot calculate to a nicety because of exception The electricity needs come of atmospheric temperature property curve belt, therefore Occurrence forecast error is the problems such as be significantly increased.In order to tackle this generation The situation of the temperature Change of exception is, it is necessary to a kind of in view of the characteristic electric power demand forecasting side of temperature Change for temporally dividing Method.

Fig. 1 is the block diagram for representing electric power demand forecasting device according to an embodiment of the invention.

Reference picture 1, electric power demand forecasting device 100 according to an embodiment of the invention includes data reception portion 110, is similar to Day search part 120 and demand characteristics curve generating unit 130.

Hereinafter, reference picture 1 illustrates electric power demand forecasting device according to an embodiment of the invention.

Data reception portion 110 receives estimated temperature number corresponding with the temperature forecast of the target day of electricity needs to be predicted According to.Herein, it is contemplated that temperature record is 24 hours temperature forecasts of the target day received from meteorological observatory, according to the difference of embodiment, Can also be transformed to weather forecast in 24 hours with interpolation etc. for the 3 hours unit temperature forecasts received from meteorological observatory As a result.Data reception portion 110 is connected by wired or wireless data communication with meteorological observatory, and the Open provided with meteorological observatory API etc. receives above-mentioned estimated temperature record.

With the temperature historical data base 200 for having set be compared estimated temperature record by similar day search part 120. This, temperature historical data base 200 is stored with the result that is measured of 24 hours temperature Changes of the measure day determined to the past.Example Such as, 24 hours ambient temperature curve figures of past Soul region shown in Fig. 2 to Fig. 4 etc. are included in above-mentioned temperature historical data base 200 In.So as to similar day search part 120 is compared by with temperature historical data base 200, and can extract out has and estimated temperature The temperature Change characteristic curve of data similar temperature Change characteristic measure day, and the measure day of above-mentioned extraction is set It is the similar day for target day.

Specifically, it is similar to day search part 120 and is extracted out with similar temperature using model is extracted the similar day for having set out The similar day of variation characteristic curve.That is, calculate storage in temperature historical data base 200 each by determine bu temperature Similar degree between data and above-mentioned estimated temperature record, and extract the measure day for having set number out according to similar degree order high As similar day.Now, in order to calculate the characteristic similar degree of temperature Change, 8 kinds of temperature of following tables can be used to want Element.

Here, T^f _maxIt is the highest temperature of target day, T^p _maxIt is the highest temperature for determining day, T^f _minIt is the minimum of target day Temperature, T^p _minIt is the lowest temperature for determining day, R is the coefficient correlation between the temperature of target day and measure day.In addition, subscript m It is morning time (1 point~12 points), a is (13 points~24 points), slope the time in the afternoon^f _mIt is the morning temperature slope of target day, slope^p _mIt is the morning temperature slope for determining day, slope^f _aIt is the temperature slope in afternoon of target day, slope^p _aDetermine under day Noon temperature slope.T^f _hIt is the h point temperature of target day, T^p _hIt is the h point temperature for determining day, h is the positive number of satisfaction 0≤h≤24, H^f _max It is the highest temperature time of origin of target day, H^p _maxIt is the highest temperature time of origin for determining day, H^f _minIt is the minimum gas of target day Warm time of origin, H^p _minIt is the lowest temperature time of origin for determining day.

Similar day search part 120 calculates similar degree with extraction model of similar day, it is also possible to according to embodiment by Europe Family name's distance model is used as similar day to extract model out.Now, day extraction model is similar to be generated using above-mentioned 8 kinds of temperature key elements, and The model of various species is generated according to the temperature key element for being used.

Specifically, day search part 120 is similar to any one in following 4 kinds similar days extraction models to extract class out Like day.

Here, EDM1 to EDM3 is the model for being applicable weighted euclidean distance, EDM4 is the model for being applicable general Euclidean distance. That is, the end value for being calculated by each calculating formula is Euclidean distance, it is meant that the above-mentioned smaller similar degree of Euclidean distance is higher. This, ω_ijIt is i-th j-th weighted value of temperature key element of model.

In addition, electricity needs reacts sensitive to temperature Change, therefore similar day retrieval is performed on the basis of identical season.From And, nearest month (30 days) on the basis of target day are set as and target day identical season or weather conditions.And, ± 30 days on the basis of past phase same date the year before, ± 30 days on the basis of the phase same date before past 2 years are wrapped Include in similar day retrieval is interval to expand the female collection size for being similar to day retrieval.For example, being set by January 21st, 2010 In the case of being set to target day, will be from December 22,19 days to 2009 January in 2010, from 20 days to 2008 2 months 2009 December 22,2 months 2008 on December 22nd, 20 days 1 are set as that above-mentioned similar day retrieval is interval.

In addition, the weighted value ω of temperature key element_ijThe recurrence by each model point can be constructed using past data After equation, applicable least square method is calculated.Here, being constructed and least square for the regression equation that weighted value is calculated Method is applicable the content for belonging to and being widely known by the people, therefore omits the explanation about specific usability methods.

Any one in the similar applicable similar day extraction model of day extraction unit 120 is included in what is set to calculate Extract the Europe between the temperature Change characteristic curve of the measure day in interval and the prediction temperature record of above-mentioned target day similar day out Family name's distance.Afterwards, in day is determined, according to the nearest order of the Euclidean distance between target day, extraction has set number Extraction day as similar day.According to the difference of embodiment, it is also possible to determine Euclidean distance most short 3 and extract out days as above-mentioned Similar day.

Demand characteristics curve generating unit 130 is extracted out in the measure electricity for determining above-mentioned similar day from electricity needs database 300 Power demand characteristics curve, and it is bent that the prediction electricity needs characteristic of target day is generated using said determination electricity needs characteristic curve Line.Here, it is bent in the electricity needs characteristic temporally divided for determining past measure day to be stored with electricity needs database 300 Line, above-mentioned electricity needs database 300 can be provided from exchange of Korea Electric Power etc..

First, demand characteristics curve generating unit 130 is applicable finger for extracted out multiple electricity needs characteristic curves that determine The smooth demand characteristics curve of number exponential smoothing next life exponentially.Following calculating formula represents the finger extracted out in the case of 3 similar days The smooth demand characteristics curve of number.

Here, SDL^p _hIt is the temporal electricity needs of h of target day, SDLⁿ _hIt is the h on the date that similar degree is n-th high Temporal electricity needs, α is exponential smoothing weighted value.

Then, demand characteristics curve generating unit 130 carries out the normalization regular need of generation to exponential smoothing demand characteristics curve Characteristic curve is sought, and maximum power requirement forecasting value and minimum power requirement forecasting are applicable for above-mentioned regular demand characteristic curve Value generation prediction electricity needs characteristic curve.

Specifically, demand characteristics curve generating unit 130 is utilized

Generation regular demand characteristic curve, here, PU_SDL^p _hBe target day the temporal normalizations of h after regular need Evaluation, SDL^p _maxIt is the maximum of the exponential smoothing requirements of target day, SDL^p _minBe the exponential smoothing requirements of target day most Small value.

Then, demand characteristics curve generating unit 130 is utilized

Ultimately generate prediction electricity needs characteristic curve.Here,The temporal prediction electric power of h for being target day need to Evaluation,It is maximum power requirement forecasting value,It is minimum power requirement forecasting value.Here, maximum power requirement forecasting Value and minimum power requirement forecasting value are calculated by conventional mode, and above-mentioned maximum power requirement forecasting value and minimum are electric Power requirement forecasting value can directly be calculated by electric power demand forecasting device, it is also possible to be provided from outside.

Fig. 5 represents using electric power demand forecasting device according to an embodiment of the invention to perform electric power demand forecasting The curve map of result.Specifically, the situation of Fig. 5 (a) is the electricity needs of the decline atmospheric temperature property curve that temperature is gradually reduced Predict the outcome, Fig. 5 (b) is the electric power demand forecasting result risen on atmospheric temperature property curve that temperature is gradually increasing.Such as Fig. 5 institutes Show, predicated error can be improved compared to conventional mode in the case of being able to confirm that with Forecasting Methodology of the invention.

And, reference picture 5 is able to confirm that the difference according to each atmospheric temperature property curve, the class that worst error at least occurs It is different to extract model out like the similar day for extracting model day out with mean error at least occurs.So as to, according to the difference of embodiment, According to each atmospheric temperature property curve, or according to the species of the error to be improved, can also be taken out to be optionally applicable similar day Go out model to minimize predicated error.

Fig. 6 and Fig. 7 are the precedence diagrams for representing electric power demand forecasting method according to an embodiment of the invention.

Reference picture 6 and Fig. 7, electric power demand forecasting method according to an embodiment of the invention include estimated temperature receiving step S100, similar day extract step S200 and prediction electricity needs characteristic curve generation step S300 out, and prediction electricity needs characteristic is bent Line generation step S300 includes exponential smoothing process S310, normalization process S320 and characteristic curve generating process S330.

Hereinafter, reference picture 6 and Fig. 7 illustrate electric power demand forecasting method according to an embodiment of the invention.

In estimated temperature receiving step S100, electric power demand forecasting device receives corresponding pre- with the temperature forecast of target day Meter temperature record.Herein, it is contemplated that temperature record can be 24 hours temperature forecasts of the target day received from meteorological observatory, according to The difference of embodiment, or 3 hours temperature forecasts of unit for being received from meteorological observatory are converted with interpolation etc. It is 24 hours results of temperature forecast.Electric power demand forecasting device is connected by wired or wireless data communication with meteorological observatory, and Open API etc. for there is provided with meteorological observatory receive above-mentioned estimated temperature record.

Similar day is extracted out in step S200, and electric power demand forecasting device is by above-mentioned estimated temperature record and the temperature for having set Historical data base is compared.Here, temperature historical data base is the 24 hours gas for determining day determined for the past that are stored with The result that temperature change is measured.Extract out similar day in step S200 by the comparing with temperature historical data base, extract out with it is pre- The temperature Change characteristic curve corresponding measure day of temperature record is counted as similar day.

Specifically, it is similar in day extraction step S200, is extracted out with similar using model is extracted the similar day for having set out Temperature Change characteristic similar day.That is, calculate store in temperature historical data base each by the gas for determining bu Similar degree between warm data and above-mentioned estimated temperature record, and extract the measure for having set number out according to similar degree order high Day is used as similar day.Now, in order to calculate the characteristic similar degree of temperature Change, can be with multiple temperature key elements.Cause Having been described that temperature key element before, therefore in this detailed.

According to the difference of embodiment, in similar day extracts step S200 out, can be applicable with Euclidean distance model to count That calculates similar degree extracts model similar day out, and extracts mould out according to the different and applicable various similar day of the temperature key element for being used Type.Specifically, it is similar in day extraction step S200, can extracts models out extract similar day out with following 4 kinds similar days.

Here, EDM1 to EDM3 is the model for being applicable weighted euclidean distance, EDM4 is the model for being applicable general Euclidean distance. That is, the end value for being calculated in each calculating formula is Euclidean distance, it is meant that the above-mentioned smaller similar degree of Euclidean distance is higher. This, ω ij are i-th j-th weighted values of temperature key element of model.

Similar day retrieval can be performed on the basis of identical season, it is also possible to by nearest one on the basis of target day It is set as the moon (30 days) and target day identical season and weather conditions.Here, can be by by with past the year before identical It is included within ± 30 days on the basis of date, ± 30 days on the basis of the phase same date before past 2 years similar day retrieval interval interior Mode expand the female collection size for being similar to day retrieval.The weighted value ω ij of temperature key element can use past data After constructing the regression equation by each model point, applicable least square method is calculated.

In addition, being similar in day extraction step S200, any one in applicable similar day extraction model is calculated and is included in Between the temperature Change characteristic curve and the prediction temperature record of target day of the measure day in retrieval of similar day interval for having set Euclidean distance.Then, in day is determined, according to the order nearest with the Euclidean distance of target day, extraction has set number Day is extracted out as similar day.According to the difference of embodiment, nearest 3 of the above-mentioned Euclidean distance for calculating can be determined and made days For above-mentioned similar day extracts out.

In prediction electricity needs characteristic curve generation step S300, extracted out in class from the electricity needs database for having set Like the measure electricity needs characteristic curve that day determines, and the pre- of target day is generated using said determination electricity needs characteristic curve Survey electricity needs characteristic curve.

Specifically, as shown in fig. 7, prediction electricity needs characteristic curve generation step S300 includes exponential smoothing process S310, normalization process S320 and characteristic curve generating process S330.

In exponential smoothing process S310, the multiple electricity needs characteristic curves that determine for above-mentioned extraction are applicable exponential smoothing The smooth demand characteristics curve of method next life exponentially.At this point it is possible to utilize

To generate above-mentioned exponential smoothing demand characteristics curve.Here, SDL^p _hThe temporal exponential smoothings of h for being target day need to Evaluation, SDLⁿ _hIt is the temporal electricity needs values of h of the similar day that similar degree is n-th high, α is exponential smoothing weighted value.

In normalization process S320, it is special to generate regular demand to carry out normalization to above-mentioned exponential smoothing demand characteristics curve Linearity curve.Utilized in normalization process

To generate above-mentioned regular demand characteristic curve.Here, PU_SDL^p _hBe target day the temporal normalizations of h after Regular demand value, SDL^p _hIt is the temporal exponential smoothing requirements of h of target day, SDL^p _maxIt is the exponential smoothing demand of target day The maximum of value, SDL^p _minIt is the minimum value of the exponential smoothing requirements of target day.

And, in characteristic curve generating process S330, receiving the maximum power requirement forecasting value for above-mentioned target day After minimum power requirement forecasting value, for above-mentioned regular demand characteristic curve be applicable above-mentioned maximum power requirement forecasting value and Minimum power requirement forecasting value, ultimately generates prediction electricity needs characteristic curve.

Specifically, prediction electricity needs characteristic curve is utilizedCome Generation.Here,It is the temporal prediction electricity needs values of h of target day, PU_SDL^p _hIt is the h times of above-mentioned target day On normalization after regular demand value,It is above-mentioned maximum power requirement forecasting value,It is above-mentioned minimum power demand Predicted value.Maximum power requirement forecasting value and minimum power requirement forecasting value are calculated by conventional mode, above-mentioned maximum Electric power demand forecasting value and minimum power requirement forecasting value can directly be calculated by electric power demand forecasting device, it is also possible to from Outside is provided.

In addition, electric power demand forecasting method according to an embodiment of the invention is performed according to the precedence diagram of Fig. 8.

Reference picture 8, user extract out in the past determine date on according to 24 hours electricity needs characteristics of temperature Change Curve (S1).That is, relevant temperature Change characteristic curve and the characteristic letter of electricity needs by each date point shown in Fig. 4 Breath, can extract out from database or meteorological observatory, electricity transaction institute for having set etc..

Then, user's input will predict the target day of electricity needs, and collect temperature forecast for the target day etc. (S2).Here, temperature forecast is the information being predicted for 24 hours temperature Changes of target day, can be by from meteorological observatory The modes such as reception are collected.

User has the temperature Change characteristic similar day similar with the temperature forecast of target day to extract out, if Similar day extract model (S3) out calmly.Specifically, any one in EDM1, EDM2, EDM3, the EDM4 that will can illustrate before It is set as that above-mentioned similar day extracts model out.But, according to the difference of embodiment, it is also possible to preset be according to various situations or Condition etc. selects mutually different similar day to extract model out.For example, the temperature forecast that can be set as target day includes rising EDM3 is used during characteristic curve, and EDM4 is used during including negative characteristic curve, or be set as according to each season or regulation Cycle selection similar day set in advance extracts model out.

After target day has determined that, in order to extract out with the temperature Change characteristic date similar with target day, The range of search (S4) on date in setting past.For example, nearest month (30 days) on the basis of target day are regarded as and target Identical season day and weather conditions, and set it to range of search.But, in this case, it is possible to because of range of search It is narrow and be difficult to extract out and the target day abundant similar date.It is thus possible to be by by the phase same date with the past the year before ± 30 days of benchmark, ± 30 days on the basis of the phase same date before past 2 years be included in side in similar day retrieval interval Formula collects size to expand for being similar to the female of day retrieval.

After similar day extracts model and range of search out to be decided, temperature forecast for target day and retrieval is included in In the range of the characteristic similar degree of temperature Change on date in past be compared.Specifically, mould is extracted out using similar degree Type is come the characteristic Euclidean distance of the relevant temperature Change (S5) that calculates target day with pass by the date.Here, meaning Euclidean Distance is nearer, and similar degree is higher.It is thus possible to be determined by comparing Euclidean distance about the temperature Change of target day and each The similar degree of the individual temperature Change on date in the past.

Then, the result of calculation of Euclidean distance, closest upper number 3 dates as similar day is extracted out (S6).That is, It is included in the multiple dates in the past in range of search, extracts out with special with the most similar temperature Change of the temperature forecast of target day 3 similar days of linearity curve.Here, extraction 3 examples of similar day of upper number are shown, but can also be according to embodiment not With the similar day for extracting different numbers out, this is obvious.

When extracting for 3 similar day out, the special according to 24 hours electricity needs of temperature Change of similar day for being extracted out is utilized Linearity curve, calculates 24 hours electricity needs characteristic curves (S7) of target day.Electricity needs is sensitive for temperature Change, therefore 24 hours electricity needs characteristic curves and realistic objective with the temperature Change characteristic similar day similar with target day 24 hours electricity needs characteristic curves of day are extremely similar to.So as to if by 24 hours electricity needs for 3 similar days The mode that characteristic curve is applicable exponential smoothing calculates 24 hours electricity needs characteristic curves of target day, then be obtained in that tool There are 24 hours electricity needs characteristic curves of pinpoint accuracy.But, according to the difference of embodiment, it is also possible to put down with except index Various methods beyond sliding method.

In addition, after 24 hours electricity needs characteristic curves about target day are calculated, it is possible to use the curve is pre- Survey 24 hours electricity needs of target day, it is possible to use predict the outcome (S8) in all fields.

The present invention is not limited to foregoing embodiment and accompanying drawing.To those skilled in the art, this is not being departed from In the range of the technology design of invention, inscape of the invention can be entered line replacement, deformation and become do not say even more and Analogy.

Claims (11)

1. a kind of electric power demand forecasting method, it is characterised in that including：

Estimated temperature receiving step, electric power demand forecasting device receives corresponding with the temperature forecast of target day estimated in the step Temperature record；

Similar day extracts step out, and electric power demand forecasting device goes through above-mentioned estimated temperature record and the temperature for having set in the step History database is compared, and extracts similar day corresponding with the temperature Change characteristic curve of above-mentioned estimated temperature record out；And

Prediction electricity needs characteristic curve generation step, electric power demand forecasting device is from the electricity needs number for having set in the step According to the measure electricity needs characteristic curve extracted out in storehouse in above-mentioned similar day measure, and using said determination electricity needs characteristic song Line generates the prediction electricity needs characteristic curve of above-mentioned target day.

2. electric power demand forecasting method according to claim 1, it is characterised in that

Above-mentioned similar day is extracted out in step, and model is extracted out using the similar day for having set, and calculates storage in above-mentioned temperature history The similar degree between each temperature record and above-mentioned estimated temperature record by measure bu in database, and according to above-mentioned class Seemingly spend order high and extract the measure day for having set number out as above-mentioned similar day.

3. electric power demand forecasting method according to claim 2, it is characterised in that

It is using following calculating formula to calculate Euclidean distance and determine similar degree to extract model above-mentioned similar day out, i.e.,

$\begin{array}{c}EDM1=\sqrt{D_1+D_2+D_3+D_4+D_5}\\ =\sqrt{{\mathrm{\ω}}_{11}{{\mathrm{\ΔT}}_{\max }}^2+{\mathrm{\ω}}_{12}{{\mathrm{\ΔT}}_{\min }}^2+{\mathrm{\ω}}_{13}{\mathrm{TC}}^2+{\mathrm{\ω}}_{14}{{\mathrm{\Δslope}}_m}^2+{\mathrm{\ω}}_{15}{{\mathrm{\Δslope}}_a}^2}\end{array},$

Wherein, Δ T_max=T^f _max-T^p _max, Δ T_min=T^f _min-T^p _min, T^f _maxIt is the highest temperature of target day, T^p _maxIt is to determine day The highest temperature, T^f _minIt is the lowest temperature of target day, T^p _minIt is the lowest temperature for determining day, TC=1-R, R are targets day and survey Coefficient correlation between the temperature on settled date, Δ slope_m=| slope^f _m-slope^p _m|, Δ slope_a=| slope^f _a-slope^p _a |, slope^f _mIt is the morning temperature slope of target day, slope^p _mIt is the morning temperature slope for determining day, slope^f _aIt is target day Afternoon temperature slope, slope^p _aIt is the temperature slope in afternoon for determining day, ω₁₁、ω₁₂、ω₁₃、ω₁₄、ω₁₅It is weighted value, it is above-mentioned Weighted value is applicable each regression equation and is calculated by least square method.

4. electric power demand forecasting method according to claim 2, it is characterised in that it is to utilize that above-mentioned similar day extracts model out Following calculating formula calculates Euclidean distance and determines similar degree, i.e.,

$\begin{array}{c}EDM2=\sqrt{D_1+D_2+D_3+D_4+D_5+D_7+D_8}\\ =\sqrt{\begin{array}{c}{\mathrm{\ω}}_{21}{{\mathrm{\ΔT}}_{\max }}^2+{\mathrm{\ω}}_{22}{{\mathrm{\ΔT}}_{\min }}^2+{\mathrm{\ω}}_{23}{\mathrm{TC}}^2+{\mathrm{\ω}}_{24}{{\mathrm{\Δslope}}_m}^2\\ +{\mathrm{\ω}}_{25}{{\mathrm{\Δslope}}_a}^2+{\mathrm{\ω}}_{26}{{\mathrm{\ΔH}}_{\max }}^2+{\mathrm{\ω}}_{27}{{\mathrm{\ΔH}}_{\min }}^2\end{array}}\end{array},$

Wherein, Δ T_max=T^f _max-T^p _max, Δ T_min=T^f _min-T^p _min, T^f _maxIt is the highest temperature of target day, T^p _maxIt is to determine day The highest temperature, T^f _minIt is the lowest temperature for determining day, T^p _minIt is the lowest temperature for determining day, TC=1-R, R are targets day and survey Coefficient correlation between the temperature on settled date, Δ slope_m=| slope^f _m-slope^p _m|, Δ slope_a=| slope^f _a-slope^p _a |, slope^f _mIt is the morning temperature slope of target day, slope^p _mIt is the morning temperature slope for determining day, slope^f _aIt is target day Afternoon temperature slope, slope^p _aIt is the temperature slope in afternoon for determining day, Δ H_max=H^f _max-H^p _max, Δ H_min=H^f _min-H^p _min, H^f _maxIt is the highest temperature time of origin of target day, H^p _maxIt is the highest temperature time of origin for determining day, H^f _minBe target day most Low temperature time of origin, H^p _minIt is the lowest temperature time of origin for determining day, ω₁₁、ω₁₂、ω₁₃、ω₁₄、ω₁₅It is weighted value, on Applicable each regression equation of weighted value is stated to be calculated by least square method.

5. electric power demand forecasting method according to claim 2, it is characterised in that it is to utilize that above-mentioned similar day extracts model out Following calculating formula calculates Euclidean distance and determines similar degree, i.e.,

$EDM3=\sqrt{D_1+D_2+D_3+D_7+D_8}$

$=\sqrt{{\mathrm{\ω}}_{31}{{\mathrm{\ΔT}}_{\max }}^2+{\mathrm{\ω}}_{32}{{\mathrm{\ΔT}}_{\min }}^2+{\mathrm{\ω}}_{33}{\mathrm{TC}}^2+{\mathrm{\ω}}_{34}{{\mathrm{\ΔH}}_{\max }}^2+{\mathrm{\ω}}_{35}{{\mathrm{\ΔH}}_{\min }}^2},$

Wherein, Δ T_max=T^f _max-T^p _max, Δ T_min=T^f _min-T^p _min, T^f _maxIt is the highest temperature of target day, T^p _maxIt is to determine day The highest temperature, T^f _minIt is the lowest temperature for determining day, T^p _minIt is the lowest temperature for determining day, TC=1-R, R are targets day and survey Coefficient correlation between the temperature on settled date, Δ H_max=H^f _max-H^p _max, Δ H_min=H^f _min-H^p _min, H^f _maxIt is the highest gas of target day Warm time of origin, H^p _maxIt is the highest temperature time of origin for determining day, H^f _minIt is the lowest temperature time of origin of target day, H^p _min It is the lowest temperature time of origin for determining day, ω₁₁、ω₁₂、ω₁₃、ω₁₄、ω₁₅It is weighted value, above-mentioned weighted value is to be applicable each Regression equation is calculated by least square method.

6. electric power demand forecasting method according to claim 2, it is characterised in that it is to utilize that above-mentioned similar day extracts model out Following calculating formula calculates Euclidean distance and determines similar degree, i.e.,

$EDM4=\sqrt{D_6}=HD=\sqrt{\underset{h=1}{\overset{24}{\Σ}}{(T_h^f-T_h^p)}^2},$

Wherein, T^f _hIt is the h point temperature of target day, T^p _hIt is the h point temperature for determining day, h is the positive number of satisfaction 0≤h≤24.

7. electric power demand forecasting method according to claim 2, it is characterised in that above-mentioned prediction electricity needs characteristic curve Generation step includes：

Exponential smoothing process, it is applicable exponential smoothing, generation for multiple electricity needs characteristic curves that determine of above-mentioned extraction Exponential smoothing demand characteristics curve；

Normalization process, it carries out normalization to above-mentioned exponential smoothing demand characteristics curve, generates regular demand characteristic curve；With And

Characteristic curve generating process, it receives pre- about the maximum power requirement forecasting value and minimum power demand of above-mentioned target day Measured value, and it is applicable above-mentioned maximum power requirement forecasting value and minimum power requirement forecasting for above-mentioned regular demand characteristic curve Value, generates above-mentioned prediction electricity needs characteristic curve.

8. electric power demand forecasting method according to claim 7, it is characterised in that above-mentioned exponential smoothing process is using following Calculating formula generates above-mentioned exponential smoothing demand characteristics curve, i.e.,

${\mathrm{SDL}}_h^p={\mathrm{\αSDL}}_h^1+\mathrm{\α}(1-\mathrm{\α}){\mathrm{SDL}}_h^2+\mathrm{\α}{(1-\mathrm{\α})}^2{\mathrm{SDL}}_h^3+{(1-\mathrm{\α})}^3\frac{{\mathrm{SDL}}_h^1+{\mathrm{SDL}}_h^2+{\mathrm{SDL}}_h^3}{3},$

Wherein, SDL^p _hIt is the temporal exponential smoothing requirements of h of target day, SDLⁿ _hIt is similar day that similar degree is n-th high The temporal electricity needs values of h, α is exponential smoothing weighted value.

9. electric power demand forecasting method according to claim 7, it is characterised in that above-mentioned normalization process utilizes following meter Formula generates above-mentioned regular demand characteristic curve, i.e.,

$PU\_{\mathrm{SDL}}_h^p=\frac{{\mathrm{SDL}}_h^p-{\mathrm{SDL}}_{\min }^p}{{\mathrm{SDL}}_{\max }^p-{\mathrm{SDL}}_{\min }^p},$

Wherein, PU_SDL^p _hBe target day the temporal normalizations of h after regular demand value, SDL^p _hIt is on the h times of target day Exponential smoothing requirements, SDL^p _maxIt is the maximum of the exponential smoothing requirements of target day, SDL^p _minIt is that the index of target day is put down The minimum value of sliding requirements.

10. electric power demand forecasting method according to claim 7, it is characterised in that above-mentioned characteristic curve generating process profit Above-mentioned prediction electricity needs characteristic curve is generated with following calculating formula, i.e.,

${\hat{L}}_d\[h\]=({\hat{L}}_d^{\max }-{\hat{L}}_d^{\min })\×PU\_{\mathrm{SDL}}_h^p+{\hat{L}}_d^{\min },$

Wherein,It is the temporal prediction electricity needs values of h of target day, PU_SDL^p _hIt is on the h times of above-mentioned target day Normalization after regular demand value,It is above-mentioned maximum power requirement forecasting value,It is above-mentioned minimum power demand Predicted value.

A kind of 11. electric power demand forecasting devices, it is characterised in that including：

Data reception portion, it receives estimated temperature record corresponding with the temperature forecast of target day；

Similar day search part, with the temperature historical data base for having set be compared above-mentioned estimated temperature record by it, and extracts out Similar day corresponding with the temperature Change characteristic curve of above-mentioned estimated temperature record；And

Demand characteristics curve generating unit, it extracts the measure determined in above-mentioned similar day out from the electricity needs database for having set Electricity needs characteristic curve, and it is bent using the prediction electricity needs characteristic that said determination electricity characteristic curve generates above-mentioned target day Line.