CN106408119A - Large power grid load prediction method based on comprehensive evaluation of analog-meteorological subareas - Google Patents

Abstract

The invention discloses a large power grid load prediction method based on comprehensive evaluation of analog-meteorological subareas. The large power grid load prediction method includes the steps of obtaining a load prediction value and resent historical data of each subarea, dividing the whole network into N meteorological subareas by using synchronous back substitution eliminating technology, calculating average proportional coefficient of moment t in N subareas recently, predicting proportional coefficient of the moment t at the day to be predicted in the N subareas, constructing the comprehensive evaluation index of the moment t, selecting q subareas with higher priority, and predicting the whole network load of the moment t respectively, establishing an optimal comprehensive model for q different prediction values, obtaining the prediction value of the whole network load at the moment t, respectively establishing optimal comprehensive models for T moments in a day to obtain load prediction sequences, providing G load prediction sequences for G prediction schemes for different values q, and reestablishing an optimal comprehensive model for each moment to obtain final prediction result of the whole network load at the day to be predicted. Through the method, the short-period load prediction accuracy of a power system is improved.

Description

Bulk power grid load forecasting method based on class meteorology subregion overall merit

Technical field

The present invention relates to Load Prediction In Power Systems correlative technology field, specifically, it is related to a kind of class meteorology that is based on and divides The bulk power grid load forecasting method of area's overall merit, for power-system short-term load forecasting.

Background technology

For ensureing the dynamic equilibrium of power system generated output and load power it is necessary to make science to power system load Prediction.Load prediction is the important process that control centre and power network development plan department, and the result of load prediction is to electricity The aspects such as network operation, control, scheduling, planning, construction have important guiding value, and it is electrical network scientific development and scientific dispatch Basis.

Improve load prediction technical merit, be conducive to planned supply and use of electric power manage, be conducive to reasonable arrangement power system operating mode and Unit maintenance scheduling, is conducive to economizing on coal, fuel-economizing and reduce cost of electricity-generating, is conducive to improving economic benefit and the society of power system Benefit.Therefore, load prediction has become the important content realizing power system management modernization.

Currently with region load prediction the whole network system loading method be " subnet summation ", " subnet summation " pre- Flow gauge is as shown in Figure 1.Its prediction is segmented into following three steps substantially：

1st, take into full account the factor of various impact short-term load forecastings in regional, select suitable Forecasting Methodology, Then short-term load forecasting is carried out to regional according to each regional historical data；Draw 96 point load prediction knots of each subnet Really.

2nd, each region load prediction results are collected, the 96 point prediction data in each region are carried out adding up obtains each moment Point cumulative and.

3rd, 96 points of the station service of day to be predicted and network loss are calculated, and to cumulative and be modified drawing final Anhui electricity Net load prediction results.

Because the method will be predicted to each region subnet, and for short-term load forecasting, due to each area The steady load degree in domain is different, and prediction difficulty is totally different, and meanwhile, station service and network loss data are also required to predict, therefore, work as profit When carrying out subnet with the prediction load of Zone Full and adding up, the accuracy effect of the whole network system loading prediction may be not ideal enough.

In order to solve the above problems, the applicant in the application for a patent for invention of Application No. CN201310648023.9, Propose a kind of the whole network load forecasting method based on region predicted load overall merit accuracy rate, the method is to a certain degree On improve the accuracy of power-system short-term load forecasting.But said method is when carrying out region division, it is with geographical position Residing administrative region divides to region, but the variation tendency of each administrative region meteorological condition is different, especially In meteorological condition change violent summer it is thus possible to lead to the load of each administrative region and the ratio of the load of the whole network connecting Change greatly in continuous several days, be unfavorable for the prediction of each administrative region and the whole network load proportion coefficient；According to comprehensive evaluation index Result select q regional prediction the whole network load, due to different q values, will have different prediction effects, therefore arrange different pre- Survey scheme and its number of regions of selection, can avoid the presence of prediction limits error.

Content of the invention

The present invention be directed to existing when bulk power grid load being predicted based on region predicted load overall merit, adopt Administrative region residing for geographical position carries out the problems of region division to bulk power grid, provides one kind based on class meteorology subregion The bulk power grid load forecasting method of overall merit, to improve power-system short-term load forecasting accuracy rate.Meanwhile, this method is same Selected part region predicted load, to predict the whole network load, can avoid some region load predictions specific responsibilitys to report regional prediction Result affects the whole network load prediction not in time.

The technical problem solving required for the present invention, can be achieved through the following technical solutions：

A kind of bulk power grid load forecasting method based on class meteorology subregion overall merit is it is characterised in that include following walking Suddenly：

(1) obtain the predicted load of the M administrative region having reported in daily load prediction value to be predicted；Obtain recent one As historical data sample space, the load data in described historical data is the whole network and M to historical data in the individual sample period The actual load of individual administrative region and prediction load；

(2) at least 2 history meteorological condition data in M administrative region in the sample period are obtained, meteorological by history Condition data, is divided into the weather region that N number of meteorological condition differs greatly；

(3) calculate the average proportions coefficient in moment point t for the described N number of weather regionUse exponential smoothing dynamic prediction Day to be predicted, N number of weather region, in the proportionality coefficient of identical moment point t, obtained the proportionality coefficient in moment point t for N number of weather region Matrix C_t；

(4) by the weight of multiple single evaluation indexs and each single evaluation index, build moment point t overall merit and refer to Mark F_AL,t；

(5) N number of weather region is pressed to the comprehensive evaluation index F in each region in moment point t_AL,tPriority from small to large Sequence, and with comprehensive evaluation index F_AL,tMinimum highest priority.

(6) (q is between 1 to arrange the weather region number q selecting in G (1≤G≤N) individual prediction scheme and each prediction scheme Numerical value to N), and the numerical value of the q of each prediction scheme is different；

(7) for certain prediction scheme g, select q higher weather region of moment point t priority, with selected q The whole network system loading of weather region difference prediction time point t, obtains q different predicting the outcome；

(8) q different predicting the outcome is set up with the optimal synthesis model of moment point t, calculates selected each meteorologic district The optimal weights in domain, obtain moment point t the whole network system loading finally predict the outcome for

Wherein,It is the whole network system loading of being predicted by the weather region k of prediction scheme g optimum in moment point t Weight,The whole network system loading of moment point t predicting out with k-th weather region for prediction scheme g；

(9) for day to be predicted whole day T moment point, set up optimal synthesis model respectively, obtain the whole day of prediction scheme g Load prediction sequence

(10) for different the predicting the outcome of G prediction scheme G, the optimal synthesis of whole day T moment point are set up respectively Model, obtains the final load prediction results (L of day to be predicted₁,L₂,…,L_T)：

Wherein,The whole network system loading predicting for prediction scheme g is in the optimal weights of moment point t.

In the present invention, described at least 2 history meteorological condition data are the highest temperature of administrative region in the sample period 2 in degree, minimum temperature, mean temperature, human comfort, weather pattern, wind speed, comfort index, coldness index or 2 Above any combination.

Described at least 2 history meteorological condition data are the maximum temperature of administrative region, lowest temperature in the sample period Degree and mean temperature, the history meteorological condition data according to M administrative region forms meteorological data matrix X：

X=(X₁,X₂,…X_i,…,X_M)

Wherein,

X_iFor i-th administrative region within a sample period meteorological condition data acquisition system of n days,

For i-th administrative region within a sample period maximum temperature of n days；

For i-th administrative region within a sample period minimum temperature of n days；

For i-th administrative region within a sample period mean temperature of n days.

In the present invention, using the synchronous back substitution technology for eliminating based on probability metrics, matrix X is polymerized, is divided into N number of The weather region that meteorological condition differs greatly：

First, a Probability p occurring is specified to M column vector in matrix X_s=1/M (s=1,2 ..., M).Set S table Show initial administrative region, make ξ_sS-th administrative region in representing matrix X, DT_s,s'Represent administrative region s and administrative region s' Distance, its value is vector 2 norm between administrative region s and administrative region s', the vectorial DA=[1,2 ..., M] of setting^T, DB It is all the square formation of 0 M × M for all elements.

The basic step that synchronous back substitution eliminates is as follows：

1) calculate the distance between each pair administrative region DT_s,s', DT_s,s'=| | ξ_s,ξ_s'||.

2) for each administrative region k, find out the administrative region r, i.e. DT the shortest with administrative region k distance_k(r)= minDT_k,s'.

3) calculate PD_k(r)=p_k*DT_k(r), k ∈ S, finds out administrative region index h so that PD in k_h=minPD_k,k∈ S.

4) make S=S- ξ_DA(h), and p_r=p_r+p_h.DB (h, h)=DA (r), DA (h) are set to sky.

5) 2 are repeated) -4), till remaining administrative region number is N.

DA is only left N number of element, corresponding square formation DB (DA (i), DA (i))=0, i=1,2 ... N, that is, need to retain N number of gas As the weather region that condition difference is larger.

The administrative region search step of each auto polymerization in weather region that in DA, N number of meteorological condition differs greatly is as follows：

1) i=1, P=DA (i) are made.

2) find out and meet the m requiring so that DB (m, m)=P, then the administrative region comprising in the i of weather region have P=[P, m].

3) find out and meet the l requiring so that DB (l, l)=m, P=[P, l], this search procedure is up to the diagonal element of DB Till not being equal to l；

4) i=i+1, P=DA (i), repeat 2) -4), until i=N.

In the present invention, in described step (1), described historical data is pre-processed as follows：

Order：L (d, t) is the load value of the d days t, L (d, t₁) and L (d, t₂) it is the d days two adjacent with t Moment t₁、t₂Load value, L (d₁, t) with L (d₂, t) it is the load value in moment point t for two adjacent with d day；

A) for the process of missing data

If load value L (d, the t) disappearance of the d days t, then formula (1) is utilized to obtain L (d, t)：

L (d, t)=α L (d, t₁)+αL(d,t₂)+βL(d₁,t)+βL(d₂,t) (1)

In formula (1), α and β is coefficient, α>β, alpha+beta=1；

B) for the process of bad point data

Define the bias ratio that allows for load of ε, ρ (d, t) is the actual bias ratio of the d days moment point t, as ρ (d, t) >=ε, Judge that L (d, t) is bad data, bad data is usedReplaced：

In the present invention, described step (3) is expressed as：

C_t=(C_1,t,C_2,t,…,C_N,t)

Wherein,

For the 1st administrative region historical data moment point t average proportions coefficient；

For the 2nd administrative region historical data moment point t average proportions coefficient；

For n-th administrative region historical data moment point t average proportions coefficient；

C_1,tFor the 1st administrative region day to be predicted moment point t proportionality coefficient；

C_2,tFor the 2nd administrative region day to be predicted moment point t proportionality coefficient；

C_N,tFor n-th administrative region day to be predicted moment point t proportionality coefficient.

Exponential smoothing in described step (3) is that onset index smoothing model is：

In formula (3)：Represent that weather region i accounts for the predicted value of the whole network system loading ratio in moment t,J before expression It weather region i accounts for the actual value of the whole network system loading ratio in moment t；N is the number of days within a sample period；λ_jTable Show weight coefficient, λ_j=λ (1- λ)^j-1, λ is constant, and 0<λ<1.

In the present invention, in described step (4), build multiple single evaluation index systems F of moment point t_t, using three lists One evaluation index, then the multiple criteria system F of each weather region_tFor：

F_t=(F_1,t,F_2,t,F_3,t)^T

Described three single evaluation indexs are respectively：Certain weather region load refers in the comprehensive stability degree evaluation of moment point t Mark F_1,t, certain weather region load is in prediction synthesis accuracy rate evaluation index F of moment point t_2,t, certain weather region load proportion system Number is in stability evaluation index F of moment point t_3,t；

Build moment point t comprehensive evaluation index F_AL,tFor；

F_AL,t=ω × Y

Wherein：ω=(ω₁,ω₂,ω₃), ω is the weight matrix in moment point t for three single evaluation indexs；ω₁,ω₂, ω₃For multiple single evaluation index systems F_tIn each single evaluation index moment point t weight coefficient；Y is according to multiple lists One assessment indicator system F_tThe decision matrix in moment point t being formed.

Certain weather region load described is in comprehensive stability degree evaluation index F of moment point t_1,tObtained by formula (4)：

In formula (4)：RSD_tFor certain weather region moment t weather region load relative standard deviation,For certain Weather region accounts for the average proportions coefficient of system loading in moment t time domain load；

Certain weather region load described is in prediction synthesis accuracy rate evaluation index F of moment point t_2,tObtain by formula (5)

?：

In formula (5)：Q_tConsensus forecast accuracy rate for moment t in selected sample space for certain weather region；

Certain weather region load proportion coefficient described is in stability evaluation index F of moment point t_3,tObtain by formula (6)：

In formula (6)：For the sample canonical variance of the proportionality coefficient in moment t for certain weather region,For certain gas Sample mathematic expectaion as the proportionality coefficient in moment t for the region；

Or, each single evaluation index described is in the weight coefficient ω of moment point t₁,ω₂,ω₃And decision matrix Y presses Following method obtains：

1) the decision matrix Y is made to be：Y=(yil)_3×N, wherein：

In formula (7)：F_i,lL-th evaluation index for weather region i, min_i{F_i,lIt is in N number of weather region l-th The minimum of a value of evaluation index, max_i{F_i,lBe l-th evaluation index in N number of weather region maximum；

2) then have：

In formula (8)：Wherein s_lFor the standard variance of l item index in decision matrix Y,For l item index Mathematic expectaion, ω_lIt is l-th evaluation index weight coefficient in moment point t.

In the present invention, in described step (7), q different predicting the outcome is obtained for certain prediction scheme g, is expressed as：

Wherein,

It is the whole network system loading of the 1st moment point t that weather region predicts using prediction scheme g；

It is the whole network system loading of the 2nd moment point t that weather region predicts using prediction scheme g；

It is the whole network system loading of moment point t being predicted using q-th weather region of prediction scheme g, andL_q,tFor q-th weather region moment point tt predicted load, C_q,tFor q-th weather region when The proportionality coefficient predicted value of punctum t.

In the present invention, in described step (8), optimal synthesis model is set up as follows：

Characterize the object function of moment point t the whole network system loading predicted value with formula (9), formula (10)-(11) are object function Constraints：

In formula (9)：Represent that the whole network system that jth sky moment point t of prediction scheme g is predicted by weather region k is born Lotus predicted value, L_all,t,jRepresent the whole network system loading actual value of jth sky moment point t；

Q weather region selected in prediction scheme g is calculated in moment point t by formula (9), formula (10) and formula (11) Optimal weightsAfterwards, further according toThe whole network system that weighting obtains moment point t of prediction scheme g is born Lotus predicted value.

In the present invention, in described step (4), the optimal synthesis model of each prediction scheme is set up as follows：

Characterize the object function of moment point t the whole network system loading predicted value with formula (12), formula (13)-(14) are object function Constraints：

In formula (12)：Represent the whole network system loading predicted value that jth sky moment point t is predicted by prediction scheme g, L_all,t,jRepresent the whole network system loading actual value of jth sky moment point t.

The optimal weights in moment point t for the G weather region are calculated by formula (12), formula (13) and formula (14)Afterwards, then According toWeighting obtains the final predicted value of the whole network system loading of day to be predicted moment point t.

, there is following beneficial effect in the bulk power grid load forecasting method based on class meteorology subregion overall merit for the present invention：

1st, to reduce load fluctuation in subnet summation in prior art larger and be difficult to the area predicting for the inventive method The impact that domain is caused to the whole network load prediction；Avoid the prediction to power plant's electricity consumption and grid loss.

2nd, need in subnet summation to know all regions predicted load in advance, and the inventive method selected part area Domain predicted load is predicting the whole network load, it is to avoid some region load predictions specific responsibilitys report regional prediction result not in time and The whole network load prediction at impact province regulation and control center.

3rd, the inventive method adopts the method that overall target is evaluated that each area priorities are sorted, and can consider each The impact to the whole network predicted load for the region predicted load, is conducive to improving the accuracy rate of the whole network load prediction.

4th, the division in region is no longer simply carried out using administrative region, by being based on meteorological condition similarity to administrative region Processed, it is to avoid when each administrative region meteorological condition changes greatly, especially in the summer that meteorological condition change is violent The ratio of season, the load of each administrative region that may lead to and the load of the whole network changed greatly the shadow of generation in continuously several days Ring, by the division to weather region, be conducive to the prediction of each weather region and the whole network load proportion coefficient；Commented according to comprehensive The result of valency index selects q regional prediction the whole network load, due to different q values, will have different prediction effects, therefore setting is not Same prediction scheme and its number of regions of selection, can avoid the presence of prediction limits error.

Brief description

To further illustrate the present invention below in conjunction with the drawings and specific embodiments.

Fig. 1 is subnet accumulation algorithm flow process.

Fig. 2 is the inventive method flow chart.

Fig. 3 is that 6 prediction scheme that certain saves electrical network on May 18th, 2016 predict the 6 the whole network load predictions obtaining respectively Curve.

Fig. 4 is load prediction curve and the realized load curve comparison diagram of certain province's electrical network on May 18th, 2016 final whole day.

Specific embodiment

In order that the technological means of the present invention, creation characteristic, reached purpose and effect are easy to understand, with reference to tool Body illustrates, and the present invention is expanded on further.

Idea of the invention is that, by basic operation being predicted to bulk power grid load to based on region predicted load The problems of the analysis of mode situation, find region to be carried out divide using the administrative region residing for geographical position, provide A kind of bulk power grid load forecasting method based on class meteorology subregion overall merit is accurate to improve power-system short-term load forecasting Rate.

Referring to Fig. 2, based on the bulk power grid load forecasting method of class meteorology subregion overall merit, comprise the following steps：

The historical data obtaining (such as n days) a recent sample period Nei is as historical data sample space, history number According in load data be the actual load of M administrative region and predict load in the whole network and the whole network, here, administrative region refers to The region being divided with geographical position, for example, in the case that the scope of the whole network is a province, administrative region can be this inside the province The administrative division being carried out in units of city.

For described historical data, can be pre-processed as follows：

Order：L (d, t) is the load value of the d days t, L (d, t₁) and L (d, t₂) it is the d days two adjacent with t Moment t₁、t₂Load value, L (d₁, t) with L (d₂, t) it is the load value in moment point t for two adjacent with d day；

A) for the process of missing data

If load value L (d, the t) disappearance of the d days t, then formula (1) is utilized to obtain L (d, t)：

L (d, t)=α L (d, t₁)+αL(d,t₂)+βL(d₁,t)+βL(d₂,t) (1)

In formula (1), α and β is coefficient, α>β, alpha+beta=1；

B) for the process of bad point data

Define the bias ratio that allows for load of ε, ρ (d, t) is the actual bias ratio of the d days moment point t, as ρ (d, t) >=ε, Judge that L (d, t) is bad data, bad data is usedReplaced：

Then at least 2 history meteorological condition data in M administrative region in the sample period, history meteorological condition are obtained Data preferably adopts these three history of the maximum temperature of administrative region, minimum temperature and mean temperature in the sample period meteorological Condition data, certainly, for lifting prediction accuracy demand, history meteorological condition data can also using human comfort, Weather pattern, wind speed, comfort index, coldness index etc., are not the history meteorological condition data of data for weather pattern etc., can So that data is converted into using data processing.For the ease of example, in present embodiment, adopt the maximum temperature of administrative region, minimum Temperature and these three history meteorological condition data of mean temperature are illustrative.

History meteorological condition data according to M administrative region forms meteorological data matrix X：

X=(X₁,X₂,…X_i,…,X_M)

Wherein,

X_iFor i-th administrative region within a sample period meteorological condition data acquisition system of n days,

For i-th administrative region within a sample period maximum temperature of n days；

For i-th administrative region within a sample period minimum temperature of n days；

For i-th administrative region within a sample period mean temperature of n days.

So, the row vector number of X matrix is 3*n, and column vector number is M.

In the present invention, using the synchronous back substitution technology for eliminating based on probability metrics, matrix X is polymerized, divides N number of gas As the weather region that condition difference is larger：

First, a Probability p occurring is specified to M column vector in matrix X_s=1/M (s=1,2 ..., M).Set S table Show initial administrative region, make ξ_sS-th administrative region in representing matrix X, DT_s,s' represent administrative region s and administrative region The distance of s', its value is vector 2 norm between administrative region s and administrative region s', the vectorial DA=[1,2 ..., M] of setting^T, DB is the square formation of 0 M × M for all elements.

The basic step that synchronous back substitution eliminates is as follows：

1) calculate the distance between each pair administrative region DT_s,s', DT_s,s'=| | ξ_s,ξ_s'||.

2) for each administrative region k, find out the administrative region r, i.e. DT the shortest with administrative region k distance_k(r)= minDT_k,s'.

3) calculate PD_k(r)=p_k*DT_k(r), k ∈ S, finds out administrative region index h so that PD in k_h=minPD_k,k∈ S.

4) make S=S- ξ_DA(h), and p_r=p_r+p_h.DB (h, h)=DA (r), DA (h) are set to sky.

5) 2 are repeated) -4), till remaining administrative region number is N.

DA is only left N number of element, corresponding square formation DB (DA (i), DA (i))=0, i=1,2 ... N, that is, need to retain N number of gas As the weather region that condition difference is larger.

The administrative region search step of each auto polymerization in weather region that in DA, N number of meteorological condition differs greatly is as follows：

1) i=1, P=DA (i) are made.

2) find out and meet the m requiring so that DB (m, m)=P, then the administrative region comprising in the i of weather region have P=[P, m].

3) find out and meet the l requiring so that DB (l, l)=m, P=[P, l], this search procedure is up to the diagonal element of DB Till not being equal to l；

4) i=i+1, P=DA (i), repeat 2) -4), until i=N.

In fact, the essence being divided into the weather region that N number of meteorological condition differs greatly is, by gas in M administrative region It is polymerized during prediction as the less region of condition difference and (will be considered as together the less administrative region of meteorological condition difference One weather region)).Specifically when being operated, can be M administrative region is divided into based on the difference of meteorological condition N number of Weather region, comprises one or several administrative regions in each weather region.

After completing the acquisition of weather region, calculate the average proportions coefficient in moment point t for N number of weather regionUse index Exponential smoothing dynamic prediction day to be predicted N number of weather region identical moment point t proportionality coefficient, obtain N number of weather region when The proportionality coefficient Matrix C of punctum t_t；It is expressed as：

C_t=(C_1,t,C_2,t,…,C_N,t)

Wherein,

For the 1st administrative region historical data moment point t average proportions coefficient；

For the 2nd administrative region historical data moment point t average proportions coefficient；

For n-th administrative region historical data moment point t average proportions coefficient；

C_1,tFor the 1st administrative region day to be predicted moment point t proportionality coefficient；

C_2,tFor the 2nd administrative region day to be predicted moment point t proportionality coefficient；

C_N,tFor n-th administrative region day to be predicted moment point t proportionality coefficient.

Exponential smoothing is that onset index smoothing model is：

In formula (3)：Represent that weather region i accounts for the predicted value of the whole network system loading ratio in moment t,J before expression It weather region i accounts for the actual value of the whole network system loading ratio in moment t；N is the number of days within a sample period；λ_jTable Show weight coefficient, λ_j=λ (1- λ)^j-1, λ is constant, and 0<λ<1, the weight for ensureing Recent data is big, and weight at a specified future date is little, λ Generally take the constant between 0.7～0.9.

By the weight of multiple single evaluation indexs and each single evaluation index, build moment point t comprehensive evaluation index F_AL,t, concrete operations are：Build multiple single evaluation index systems F of moment point t_t, using three single evaluation indexs, then often The multiple criteria system F of individual weather region_tFor：

F_t=(F_1,t,F_2,t,F_3,t)^T

Described three single evaluation indexs are respectively：Certain weather region load refers in the comprehensive stability degree evaluation of moment point t Mark F_1,t, certain weather region load is in prediction synthesis accuracy rate evaluation index F of moment point t_2,t, certain weather region load proportion system Number is in stability evaluation index F of moment point t_3,t；

Due to containing three kinds of single evaluation indexs, three evaluation indexes in selected sample space in the multiple criteria system It is possible that different evaluation results, therefore how these three indexs of overall merit, therefrom choose most suitable several meteorology Region then becomes crucial predicting the whole network load, builds moment point t comprehensive evaluation index F_AL,tFor；

F_AL,t=ω × Y

Wherein：ω=(ω₁,ω₂,ω₃), ω is the weight matrix in moment point t for three single evaluation indexs；ω₁,ω₂, ω₃For multiple single evaluation index systems F_tIn each single evaluation index moment point t weight coefficient；Y is according to multiple lists One assessment indicator system F_tThe decision matrix in moment point t being formed.

Certain weather region load is in comprehensive stability degree evaluation index F of moment point t_1,tObtained by formula (4)：

In formula (4)：RSD_tFor certain weather region moment t weather region load relative standard deviation,For certain Weather region weather region load in moment t accounts for the average proportions coefficient of system loading；Load comprehensive stability degree index F_1,t's Physical significance is：When with certain weather region load prediction the whole network load, the size of weather region load fluctuation is born in the whole network The embodiment of lotus.When with load comprehensive stability degree index F_1,tEach moment point is ranked up to each weather region, numerical value is Little region, when representing in this moment point with load prediction the whole network load of this weather region, deviation is minimum.

Certain weather region load is in prediction synthesis accuracy rate evaluation index F of moment point t_2,tObtain by formula (5)：

In formula (5)：Q_tConsensus forecast accuracy rate for moment t in selected sample space for certain weather region；Load Comprehensive stability degree index F_2,tPhysical significance be：When with certain weather region load prediction the whole network load, weather region load The size of prediction deviation is in the embodiment of the whole network load.When with load comprehensive stability degree index F_2,tTo each in each moment point Weather region is ranked up, the minimum region of numerical value, and predictablity rate highest represents and is taken in this meteorologic district in this moment point The predicted load in domain predicts the whole network load, and the predicated error of the whole network load is minimum.

When with weather region load prediction the whole network load, the whole network load not only quality with weather region predicted load Relevant, and to account for the whole network load proportion coefficient predictors relevant with the weather region load in day to be predicted, therefore proposes gas here As region load accounts for the whole network load proportion coefficient stabilization degree index.Certain weather region load proportion coefficient stablizing in moment point t Degree evaluation index F_3,tObtain by formula (6)：

In formula (6)：For the sample canonical variance of the proportionality coefficient in moment t for certain weather region,For certain gas Sample mathematic expectaion as the proportionality coefficient in moment t for the region.

Each single evaluation index is in the weight coefficient ω of moment point t₁,ω₂,ω₃Obtain as follows：

Because the dimension of three single evaluation indexs is different with the order of magnitude, carry out each single evaluation indices non-dimension first Process obtains decision matrix Y：Y=(yil)_3×N, wherein：

In formula (7)：F_i,lL-th evaluation index for weather region i, min_i{F_i,lIt is in N number of weather region l-th The minimum of a value of evaluation index, max_i{F_i,lBe l-th evaluation index in N number of weather region maximum.

Then have：

In formula (8)：Wherein s_lFor the standard variance of l item index in decision matrix Y,For l item index Mathematic expectaion, ω_lIt is l-th evaluation index weight coefficient in moment point t.

N number of weather region is pressed to the comprehensive evaluation index F in each region in moment point t_AL,tPriority row from small to large Sequence, and with comprehensive evaluation index F_AL,tMinimum highest priority.

(q is between 1 to N to the weather region number q selecting in setting G (1≤G≤N) individual prediction scheme and each prediction scheme Numerical value), and the numerical value of the q of each prediction scheme is different.For prediction scheme g, select moment point t priority higher Q weather region, with the whole network system loading of selected q weather region difference prediction time point t, obtains prediction scheme g Q different predicting the outcome, be expressed as：

Wherein,

It is the whole network system loading of the 1st moment point t that weather region predicts using prediction scheme g；

It is the whole network system loading of the 2nd moment point t that weather region predicts using prediction scheme g；

It is the whole network system loading of moment point t being predicted using q-th weather region of prediction scheme g,

AndL_q,tFor q-th weather region moment point t predicted load, C_q,tFor q-th gas As region is in the proportionality coefficient predicted value of moment point t.

The optimal synthesis model of moment point t is set up in q different the predicting the outcome to prediction scheme g, calculates selected The optimal weights of each weather region, obtain moment point t the whole network system loading finally predict the outcome for

Wherein,It is the whole network system loading of being predicted by the weather region k of prediction scheme g optimum in moment point t Weight,The whole network system loading of moment point t predicting out with k-th weather region for prediction scheme g；；

The moment point t optimal synthesis model of prediction scheme g is set up as follows：

Characterize the object function of moment point t the whole network system loading predicted value with formula (9), formula (10)-(11) are object function Constraints：

In formula (9)：Represent that the whole network system that jth sky moment point t of prediction scheme g is predicted by weather region k is born Lotus predicted value, L_all,t,jRepresent the whole network system loading actual value of jth sky moment point t.

For the solution of formula (9)-(11), define the whole network system loading predicted value that each weather region predicts first Virtual prognostication residual error v_kjt, virtual prognostication residual sum of squares (RSS)And certain two weather region predicts the virtual of the whole network system loading The covariance predicting the outcomeAs follows：

Then formula (9)-(11) object function in the t period is converted into following matrix form：

Wherein：

Formula (12) is the canonical form of quadratic programming problem, directly calculates the selected q meteorologic district of prediction scheme g Domain is in the optimal weights of moment point tAfterwards, further according toWeighting obtains prediction side The whole network system loading predicted value of case g moment point t.

For the whole network system loading that each weather region predicted load predicts, present in different moment point Go out different prediction effects, therefore treat " each moment point of day to be predicted " with a certain discrimination, set up collective model respectively so that each The whole network system loading that weather region predicted load predicts is all different in the weight in each moment, each comfortable different to embody The prediction effect of moment point.For day to be predicted whole day T moment point, set up optimal synthesis model respectively, obtain prediction scheme g Whole day load prediction sequence

Similarly, G whole day load prediction sequence, moment point t of each prediction scheme will be had for G prediction scheme Optimal synthesis model is set up as follows：

Characterize the object function of moment point t the whole network system loading predicted value with formula (13), formula (14)-(15) are object function Constraints：

In formula (13)：Represent the whole network system loading predicted value that jth sky moment point t is predicted by prediction scheme g, L_all,t,jRepresent the whole network system loading actual value of jth sky moment point t.

The optimal weights in moment point t for the G weather region are calculated by formula (13), formula (14) and formula (15)Afterwards, then According toWeighting obtains the final predicted value of the whole network system loading of day to be predicted moment point t.

For the solution of formula (13)-(15), the solution with formula (9)-(11) is identical, same definition the whole network system loading The virtual prognostication residual error of predicted value, virtual prognostication residual sum of squares (RSS), and prediction the whole network system loading of certain two prediction scheme The covariance of virtual prognostication result, sets up such as the quadratic programming of formula (12), directly calculates prediction scheme g in moment point t Optimal weightsAfterwards, further according toThe whole network system loading weighting moment point t obtaining day to be predicted is finally pre- Measured value.For day to be predicted whole day T moment point, set up the optimal synthesis model of each prediction scheme respectively, obtain prediction side Case g whole day load prediction sequence (L₁,L₂,…,L_T), with described whole day load prediction sequence (L₁,L₂,…,L_T) it is the whole network load Predict the outcome.

Below with specific example, go on to say implementation process of the present invention：

In the methods of the invention, the meteorology letter to each administrative region using the synchronous back substitution technology for eliminating based on probability metrics Breath is polymerized, and is divided into the weather region that N number of meteorological condition differs greatly.G prediction scheme of setting and each Forecasting Methodology Selection weather region number q.Further according to the comprehensive evaluation index of each weather region, select and evaluate q optimum meteorologic district Domain.Taking predict that certain saves the whole network load as a example, this province includes 16 districts and cities and (uses districts and cities 1, districts and cities 2 ... respectively, districts and cities 16 represent, ground City's numbering criterion is：So that from north orientation south, geographical position from east to west is numbered to each districts and cities), according to meteorological condition phase Like degree determination methods, 16 districts and cities are divided into the weather region that 6 meteorological conditions differ greatly, select 6 prediction scheme, often The weather region number of individual prediction scheme corresponding selection is respectively 1,2,3,4,5,6, is embodied as carrying out according to the following steps：

1st, digital independent：Selection on May 18th, 2016 is day to be predicted, obtains prediction 30 working day the whole networks a few days ago respectively And in the actual load data of prefectures and cities, prefectures and cities' temperature record of first 30 days (highest temperature, the lowest temperature) and prefectures and cities The daily load prediction value to be predicted of report.

2nd, class weather region divides：Using the meteorology letter to 16 districts and cities for the synchronous back substitution technology for eliminating based on probability metrics Breath is polymerized, and forms the weather region that N=6 meteorological condition differs greatly.Through synchronous back substitution technology for eliminating, DA=[ Districts and cities 14 of 10 districts and cities of 8 districts and cities of 3 districts and cities of 2 districts and cities of city 11], the diagonal element of 16 × 16 matrix D B is：

(1,1)	(2,2)	(3,3)	(4,4)	(5,5)	(6,6)	(7,7)	(8,8)
								2	0	0	1	7	8	3	0
(9,9)	(10,10)	(11,11)	(12,12)	(13,13)	(14,14)	(15,15)	(16,16)
								8	0	0	11	10	0	13	14

Therefore 6 weather region of formation are as shown in the table：

3rd, the multiple criteria system solves：Three single evaluation indexs are respectively：Certain weather region load is in moment point t Comprehensive stability degree evaluation index F_1,t, certain weather region load is in prediction synthesis accuracy rate evaluation index F of moment point t_2,t, certain gas As region load proportion coefficient is in stability evaluation index F of moment point t_3,t.

Following table be in implementation steps 26 weather region in three evaluation index results of t=1 moment point.

Region	F1,t	F2,t	F3,t
				Weather region 1	0.3431	0.3779	0.0381
Weather region 2	0.1949	0.1781	0.0349
				Weather region 3	0.1998	0.0930	0.0239
Weather region 4	0.3367	0.2099	0.0294
				Weather region 5	0.1605	0.1266	0.0328
Weather region 6	0.9666	0.4578	0.0555

4th, build comprehensive evaluation index F_AL,t：(F as described in description of the invention, first to 6 weather region_1,t, F_2,t,F_3,t) dimensionless process, obtain decision matrix Y, then seek the variation weights omega=(ω of each single index₁,ω₂, ω₃), finally solve comprehensive evaluation index F_AL,t=ω × Y：

By calculating, this 6 weather region are after the comprehensive evaluation index sort result of t=1 moment point：

5th, prediction scheme is set：According to 6 weather region, without loss of generality, 6 prediction scheme, each prediction side are set Case corresponding weather region number is respectively 1,2,3,4,5,6.

6th, choose weather region and predict the whole network load respectively：Here taking prediction scheme 3 as a example, its weather region number q=3. For comprehensive evaluation index F_AL,t, its numerical value is less, illustrates that the load of this weather region is more stable, is more conducive to using this meteorology The predicted load in region is predicting the whole network load.Here according to F_AL,tRanking results, corresponding respectively is weather region 3, gas As region 5, weather region 2.ByThe 3 the whole network load prediction results obtaining in t=1 moment point are as follows：

7th, the optimal synthesis model of single moment point：The whole network predicted load of prediction t=1 moment point.To formula (9)-(11) Solved, solving the optimal weights obtaining is：

The whole network predicted load of the therefore t=1 moment point of prediction scheme 3 And be 15069.9MW in the actual load of May t=1 moment point on the 18th, the precision of prediction of the t=1 moment point of prediction scheme 3 reaches To 98.83%.

8th, the load prediction of whole day multiple spot.For t in one day day of prediction from 2 to 96 other load predictions put, repeat Step 1 is to step 6 it is possible to obtain the load prediction value sequence of whole day.

9th, the load prediction of the whole day T moment point of each prediction scheme：Similar to step 6, step 7, step 8, each Prediction scheme all carries out load prediction to 96 moment of one day day to be predicted, obtains the whole day predicted load of each prediction scheme Sequence.

10th, the optimal synthesis model of each prediction scheme：Each prediction scheme is set up respectively for each moment point Excellent collective model, solves to the planning problem of formula (13)-(15), obtains the optimum in t=1 moment point for each prediction scheme Weight is：

Final the whole network predicted load of t=1 moment point therefore is：

Fig. 3 is that this saves 6 prediction scheme of on May 18th, 2016 and predicts the 6 the whole network load prediction curves obtaining respectively, figure 4 is load prediction curve and the realized load curve contrast of the final whole day through each prediction scheme optimal synthesis model solution Figure.By calculating, the day accuracy rate of the inventive method is 98.87%.

General principle, principal character and the advantages of the present invention of the present invention have been shown and described above.The technology of the industry , it should be appreciated that the present invention is not restricted to the described embodiments, the simply explanation described in above-described embodiment and specification is originally for personnel Invention principle, without departing from the spirit and scope of the present invention the present invention also have various changes and modifications, these change Change and improvement both falls within scope of the claimed invention.Claimed scope by appending claims and its Equivalent defines.

Claims (10)

1. the bulk power grid load forecasting method based on class meteorology subregion overall merit is it is characterised in that comprise the steps：

(1) obtain the predicted load of the M administrative region having reported in daily load prediction value to be predicted；Obtain a recent sample Historical data in this period, the load data in described historical data is the actual load and in advance of the whole network and M administrative region Survey load；

(2) at least 2 history meteorological condition data in M administrative region in the sample period are obtained, by history meteorological condition Data, is divided into the weather region that N number of meteorological condition differs greatly；

(3) calculate the average proportions coefficient in moment point t for the described N number of weather regionTreated pre- with exponential smoothing dynamic prediction Survey day N number of weather region in the proportionality coefficient of identical moment point t, obtain the proportionality coefficient matrix in moment point t for N number of weather region C_t；

(4) by the weight of multiple single evaluation indexs and each single evaluation index, build moment point t comprehensive evaluation index F_AL,t；

(5) N number of weather region is pressed to the comprehensive evaluation index F in each region in moment point t_AL,tPrioritization from small to large, And with comprehensive evaluation index F_AL,tMinimum highest priority；

(6) the weather region number q selecting in G prediction scheme and each prediction scheme is set, and the number of the q of each prediction scheme Value is different, and described G meets 1≤G≤N, and described q meets 1≤q≤N；

(7) for certain prediction scheme g, select q higher weather region of moment point t priority, meteorological with selected q The whole network system loading of region difference prediction time point t, obtains q different predicting the outcome；

(8) q different predicting the outcome is set up with the optimal synthesis model of moment point t, calculates selected each weather region Optimal weights, obtain moment point t the whole network system loading finally predict the outcome for

$L_{all,t}^g=\underset{k=1}{\overset{q}{\&Sigma;}}{w}_t^{k,g}{L}_{all,t}^{k,g},k=1,2,...,q$

Wherein,It is the whole network system loading of being predicted by the weather region k of prediction scheme g optimal weights in moment point t,The whole network system loading of moment point t predicting out with k-th weather region for prediction scheme g；

(9) for day to be predicted whole day T moment point, set up optimal synthesis model respectively, obtain the whole day load of prediction scheme g Forecasting sequence

(10) for different the predicting the outcome of G prediction scheme G, set up the optimal synthesis model of whole day T moment point respectively, Obtain the final load prediction results (L of day to be predicted₁,L₂,…,L_T)：

$L_t=\underset{g=1}{\overset{G}{\&Sigma;}}{w}_t^g{L}_{all,t}^g,g=1,2,...,G$

Wherein,The whole network system loading predicting for prediction scheme g is in the optimal weights of moment point t.

2. the bulk power grid load forecasting method based on class meteorology subregion overall merit according to claim 1, its feature exists In：Described at least 2 history meteorological condition data are in the sample period, the maximum temperature of administrative region, minimum temperature, flat Any group of in equal temperature, human comfort, weather pattern, wind speed, comfort index, coldness index 2 or more than 2 Close.

3. the bulk power grid load forecasting method based on class meteorology subregion overall merit according to claim 1, its feature exists In：Described at least 2 history meteorological condition data be a sample period in, the maximum temperature of administrative region, minimum temperature and Mean temperature, the history meteorological condition data according to M administrative region forms meteorological data matrix X：

X=(X₁,X₂,…X_i,…,X_M)

$X_i={(W_{i,n}^{\max },W_{i,n}^{\min },W_{i,n}^{ave})}^T$

Wherein,

X_iFor i-th administrative region within a sample period meteorological condition data acquisition system of n days,

For i-th administrative region within a sample period maximum temperature of n days；

For i-th administrative region within a sample period minimum temperature of n days；

For i-th administrative region within a sample period mean temperature of n days.

4. the bulk power grid load forecasting method based on class meteorology subregion overall merit according to claim 3, its feature exists In：Using the synchronous back substitution technology for eliminating based on probability metrics, matrix X is polymerized, is divided into N number of meteorological condition difference relatively Big weather region：

First, a Probability p occurring is specified to M column vector in matrix X_s=1/M (s=1,2 ..., M).Set S represents First administrative region, makes ξ_sS-th administrative region in representing matrix X, DT_s,s'Represent administrative region s and administrative region s' away from From its value is vector 2 norm between administrative region s and administrative region s', the vectorial DA=[1,2 ..., M] of setting^T, DB is institute There is the square formation that element is all 0 M × M.

The basic step that synchronous back substitution eliminates is as follows：

1) calculate the distance between each pair administrative region DT_s,s', DT_s,s'=| | ξ_s,ξ_s'||.

2) for each administrative region k, find out the administrative region r, i.e. DT the shortest with administrative region k distance_k(r)=min DT_k,s'.

3) calculate PD_k(r)=p_k*DT_k(r), k ∈ S, finds out administrative region index h so that PD in k_h=min PD_k,k∈S.

4) make S=S- ξ_DA(h), and p_r=p_r+p_h.DB (h, h)=DA (r), DA (h) are set to sky.

5) 2 are repeated) -4), till remaining administrative region number is N.

5. according to the arbitrary described bulk power grid load forecasting method based on class meteorology subregion overall merit of Claims 1-4, its It is characterised by：In described step (1), described historical data is pre-processed as follows：

Order：L (d, t) is the load value of the d days t, L (d, t₁) and L (d, t₂) it is the d days two adjacent with t moment t₁、t₂Load value, L (d₁, t) with L (d₂, t) it is the load value in moment point t for two adjacent with d day；

A) for the process of missing data

If load value L (d, the t) disappearance of the d days t, then formula (1) is utilized to obtain L (d, t)：

L (d, t)=α L (d, t₁)+αL(d,t₂)+βL(d₁,t)+βL(d₂,t) (1)

In formula (1), α and β is coefficient, α>β, alpha+beta=1；

B) for the process of bad point data

Define the bias ratio that allows for load of ε, ρ (d, t) is the actual bias ratio of the d days moment point t, as ρ (d, t) >=ε, judgement L (d, t) is bad data, and bad data is usedReplaced：

$\stackrel{\&OverBar;}{L}(d,t)=\frac{L(d_1,t)+L(d_2,t)}{2}---\left(2\right).$

6. according to the arbitrary described bulk power grid load forecasting method based on class meteorology subregion overall merit of Claims 1-4, its It is characterised by：Described step (3) is expressed as：

$\stackrel{\&OverBar;}{C_t}=(\stackrel{\&OverBar;}{C_{1,t}},\stackrel{\&OverBar;}{C_{2,t}},...,\stackrel{\&OverBar;}{C_{N,t}})$

C_t=(C_1,t,C_2,t,…,C_N,t)

Wherein,

For the 1st weather region historical data moment point t average proportions coefficient；

For the 2nd weather region historical data moment point t average proportions coefficient；

For n-th weather region historical data moment point t average proportions coefficient；

C_1,tFor the 1st weather region day to be predicted moment point t proportionality coefficient；

C_2,tFor the 2nd weather region day to be predicted moment point t proportionality coefficient；

C_N,tFor n-th weather region day to be predicted moment point t proportionality coefficient.

7. the bulk power grid load forecasting method based on class meteorology subregion overall merit according to claim 6, its feature exists In：Exponential smoothing in described step (3) is that onset index smoothing model is：

$C_t^i=\underset{j=1}{\overset{n}{\&Sigma;}}{\mathrm{\&lambda;}}_j{C}_{t,j}^i,i=1,2,...,N$

$\left\{\begin{array}{c}\underset{j=1}{\overset{n}{\&Sigma;}}{\mathrm{\&lambda;}}_j=1\\ 0\&le;{\mathrm{\&lambda;}}_j\&le;1\end{array}\right.,j=1,2,...,n---\left(3\right)$

In formula (3)：Represent that weather region i accounts for the predicted value of the whole network system loading ratio in moment t,J days before expression Weather region i accounts for the actual value of the whole network system loading ratio in moment t；N is the number of days within a sample period；λ_jRepresent power Weight coefficient, λ_j=λ (1- λ)^j-1, λ is constant, and 0<λ<1.

8. according to the arbitrary described bulk power grid load forecasting method based on class meteorology subregion overall merit of Claims 1-4, its It is characterised by：In described step (4), build multiple single evaluation index systems F of moment point t_t, referred to using three single evaluation Mark, then the multiple criteria system F of each weather region_tFor：

F_t=(F_1,t,F_2,t,F_3,t)^T

Described three single evaluation indexs are respectively：Certain weather region load is in the comprehensive stability degree evaluation index of moment point t F_1,t, certain weather region load is in prediction synthesis accuracy rate evaluation index F of moment point t_2,t, certain weather region load proportion coefficient Stability evaluation index F in moment point t_3,t；

Build moment point t comprehensive evaluation index F_AL,tFor；

F_AL,t=ω × Y

Wherein：ω=(ω₁,ω₂,ω₃), ω is the weight matrix in moment point t for three single evaluation indexs；ω₁,ω₂,ω₃ For multiple single evaluation index systems F_tIn each single evaluation index moment point t weight coefficient；Y is according to multiple single Assessment indicator system F_tThe decision matrix in moment point t being formed.

9. the bulk power grid load forecasting method based on class meteorology subregion overall merit according to claim 8, its feature exists In：Certain weather region load described is in comprehensive stability degree evaluation index F of moment point t_1,tObtained by formula (4)：

$F_{1,t}=\frac{{\mathrm{RSD}}_t}{\stackrel{\&OverBar;}{C_t}},t=1,2,\mathrm{...},T---\left(4\right)$

In formula (4)：RSD_tFor certain weather region moment t weather region load relative standard deviation,Meteorological for certain Region accounts for the average proportions coefficient of system loading in moment t time domain load；

Certain weather region load described is in prediction synthesis accuracy rate evaluation index F of moment point t_2,tObtain by formula (5)：

In formula (5)：Q_tConsensus forecast accuracy rate for moment t in selected sample space for certain weather region；

Certain weather region load proportion coefficient described is in stability evaluation index F of moment point t_3,tObtain by formula (6)：

$F_{3,t}=\frac{\sqrt{S_t^2}}{\stackrel{\&OverBar;}{X_t}},t=1,2,\mathrm{...},T---\left(6\right)$

In formula (6)：For the sample canonical variance of the proportionality coefficient in moment t for certain weather region,For certain meteorologic district The sample mathematic expectaion of the proportionality coefficient in moment t for the domain；

Or, each single evaluation index described is in the weight coefficient ω of moment point t₁,ω₂,ω₃And decision matrix Y is by as follows Method obtains：

1) the decision matrix Y is made to be：Y=(yil)_3×N, wherein：

$y_{il}=\frac{F_{i,l}-{\min }_i\left\{F_{i,l}\right\}}{{\max }_i\left\{F_{i,l}\right\}-{\min }_i\left\{F_{i,l}\right\}},i=1,2,...,N;l=1,2,3---\left(7\right)$

In formula (7)：F_i,lL-th evaluation index for weather region i, min_i{F_i,lIt is evaluating for l-th in N number of weather region The minimum of a value of index, max_i{F_i,lBe l-th evaluation index in N number of weather region maximum；

2) then have：

In formula (8)：Wherein s_lFor the standard variance of l item index in decision matrix Y,Number for l item index Term hopes, ω_lIt is l-th evaluation index weight coefficient in moment point t.

10. according to the arbitrary described bulk power grid load forecasting method based on class meteorology subregion overall merit of Claims 1-4, It is characterized in that：In described step (7), for certain prediction scheme g, obtain q different predicting the outcome, be expressed as：

$(L_{all,t}^{1,g},L_{all,t}^{2,g},...,L_{all,t}^{q,g})$

Wherein,

It is the whole network system loading of the 1st moment point t that weather region predicts using prediction scheme g；

It is the whole network system loading of the 2nd moment point t that weather region predicts using prediction scheme g；

It is the whole network system loading of moment point t being predicted using q-th weather region of prediction scheme g, andL_q,tFor q-th weather region moment point t predicted load, C_q,tFor q-th weather region when The proportionality coefficient predicted value of punctum t；

In described step (8), optimal synthesis model is set up as follows：

Characterize the object function of moment point t the whole network system loading predicted value with formula (9), formula (10)-(11) are the pact of object function Bundle condition：

$\begin{array}{cc}\underset{w^{k,g}}{min}& \underset{j=1}{\overset{n}{\&Sigma;}}\underset{t=1}{\overset{T}{\&Sigma;}}{\left(\underset{k=1}{\overset{q}{\&Sigma;}}{w}_t^{k,g}{L}_{all,t,j}^{k,g}-L_{all,t,j}\right)}^2\end{array}---\left(9\right)$

$\begin{array}{cc}S.t.& \underset{k=1}{\overset{q}{\&Sigma;}}{w}_t^{k,g}=1\end{array}---\left(10\right)$

$w_t^{k,g}\&GreaterEqual;0,k=1,2,...,q---\left(11\right)$

In formula (9)：Represent that the whole network system loading that jth sky moment point t of prediction scheme g is predicted by weather region k is pre- Measured value, L_all,t,jRepresent the whole network system loading actual value of jth sky moment point t；

Q weather region selected in prediction scheme g is calculated in moment point t by formula (9), formula (10) and formula (11) Excellent weightAfterwards, further according toThe whole network system loading weighting moment point t obtaining prediction scheme g is pre- Measured value；

In described step (10), G prediction scheme will obtain G different predicting the outcome, and optimal synthesis model is as follows Set up：

Characterize the object function of moment point t the whole network system loading predicted value with formula (12), formula (13)-(14) are the pact of object function Bundle condition：

$\begin{array}{cc}\underset{{\mathrm{\&omega;}}^g}{min}& \underset{j=1}{\overset{n}{\&Sigma;}}\underset{t=1}{\overset{T}{\&Sigma;}}{\left(\underset{g=1}{\overset{G}{\&Sigma;}}{w}_t^g{L}_{all,t,j}^g-L_{all,t,j}\right)}^2\end{array}---\left(12\right)$

$\begin{array}{cc}S.t.& \underset{g=1}{\overset{G}{\&Sigma;}}{w}_t^g=1\end{array}---\left(13\right)$

$w_t^g\&GreaterEqual;0,g=1,2,...,G---\left(14\right)$

In formula (12)：Represent the whole network system loading predicted value that jth sky moment point t is predicted, L by prediction scheme g_all,t,j Represent the whole network system loading actual value of jth sky moment point t.

The optimal weights in moment point t for the G weather region are calculated by formula (12), formula (13) and formula (14)Afterwards, further according toWeighting obtains the final predicted value of the whole network system loading of day to be predicted moment point t.