CN103971175A - Short-term load prediction method of multistage substations - Google Patents

Abstract

The invention provides a short-term load prediction method of multistage substations. The short-term load prediction method comprises the steps that the historical data of n stages of substations are acquired, and the historical data are preprocessed; the preprocessed historical data are processed, and the principal factor which influences the load of an nth-stage substation is acquired; a prediction model is established, and the load prediction result of the nth-stage substation at any moment t is acquired according to the meteorological data of the nth-stage substation in the day to be predicted and the predication model; the power loss of all the substations in (n-k+1) stages of substations and the power loss of all the substations in the (n-k) stages of substations are acquired according to the parameter computing trend of the substations and lines; the load prediction result of the nth-stage substation at any moment t is acquired according to historical loads and the meteorological data; the load prediction result of the nth-stage substation at any moment t, the load prediction result of the (n-2)th-stage substation at the moment t... and the load prediction result of the (n-k)th-stage substation at the moment t are acquired according to the load prediction result of the nth-stage substation at the moment t and all power losses.

Description

A kind of short-term load forecasting method of multistage transformer station

Technical field

The invention belongs to transformer station's load prediction technical field, more specifically, relate to a kind of short-term load forecasting method of multistage transformer station.

Background technology

Transformer station's (bus nodes) load prediction is accurately to promote power grid security to check and the important step of congestion management accuracy, be realize that electrical network is reliable, safety, economy and the basic guarantee that efficiently moves.Current, it is ripe that the technical development of region load prediction has been tending towards, but that the technology of transformer station's load prediction is studied is less.Required transformer station's (bus nodes) load in existing method of operation arrangement and Security Checking, be generally by by the load in region in proportion coefficient distribute and obtain.For example, region load has increased by 1.1 times, just the load of each bus is all increased to 1.1 times and carries out mode arrangement and check.But because the load of each transformer station is different, its characteristic is different even may difference very large, the load of for example some transformer may drop to 0.8, and the load of some transformer be increased to 1.5 times even more.In this case, just there will be mode to arrange and check no problem, but partial line road and transformer overload and power supply occurs block in actual moving process have reduced the reliability of power supply.Therefore, current this too extensive dispensing pattern cannot be considered the part throttle characteristics of each transformer station, often make formulated method of operation plan substantial deviation actual motion, even cause partial line road and transformer overload in actual motion and power supply occurs block, thereby greatly reduce power supply reliability and the management and running level of system, need badly in conjunction with the part throttle characteristics of each transformer station and carry out rationally effectively load prediction.

In prior art, transformer station's load forecasting method is less, rare especially for mesolow transformer station (110kV is following) load forecasting method, and mostly applies mechanically region load forecasting method.The method of the load forecasts such as such as trend analysis method, regression analysis, exponential smoothing, unit consumption method, gray model method, load density method and elastic coefficient method.But transformer station load radix is little, do not there is the regular strong feature of region load period, simply directly apply mechanically and be difficult to obtain good prediction effect, and existing forecasting techniques and method are ignored the electrical specification between different electric pressure transformer station.

Summary of the invention

For the defect of prior art, the object of the present invention is to provide a kind of short-term load forecasting method of multistage transformer station, be intended to solve existing transformer station (bus) load forecasting method and do not consider respectively the part throttle characteristics of transformer station's load, do not consider the Electric connection characteristic between different electric pressure transformer station, so that the not enough problem such as transformer station's (bus) short-term load forecasting precision is low.

The short-term load forecasting method that the invention provides a kind of multistage transformer station, comprises the steps:

(1) obtain the historical data of n level transformer station, and described historical data is carried out to pre-service; N is more than or equal to 3 positive integer;

(2) based on grey correlation theory analysis, pretreated historical data is processed, obtained the principal element of the load that affects n level transformer station;

(3) set up forecast model according to the principal element of the load that affects n level transformer station, and obtain the load prediction results of any time t of n level transformer station according to the weather data of n level transformer station to be predicted day and described forecast model;

(4) according to the calculation of parameter trend of transformer station and circuit obtain in n level transformer station in each transformer station and n-1 level transformer station in the power attenuation between each transformer station, n-1 level transformer station the power attenuation between each transformer station in each transformer station and n-2 level transformer station ... and the power attenuation between each transformer station in each transformer station and n-k level transformer station in n-k+1 level transformer station; K=1,2 ... n; Wherein, so that from generating plant, transformer station is farthest as n level, from generating plant, nearest transformer station is the 1st grade;

(5) utilize regression analysis to obtain the load prediction results of any time t of n level transformer station according to historical load and weather data;

(6) according to each power attenuation in load prediction results and the step (4) of any time t of n level transformer station in step (5) obtain the n-1 level t of transformer station moment load prediction results, the n-2 level t of transformer station moment load prediction results ... and the n-k level t of transformer station moment load prediction results.

Wherein, described historical data comprises load data and weather data, be specially the interval 15min 1 year 365 day every day load of totally 96 periods, 1 year 365 day every day daily maximum temperature, daily mean temperature, day lowest temperature, day high humility, per day humidity, day minimum humidity, day high wind speed, per day wind speed and daily rainfall.

Wherein, for g transformer station in n level transformer station, g=1,2 ..., G _n, G _nbe the sum of transformer station in n level transformer station, described forecast model is c ₁, c ₂..., c _mfor the perunit value of the major influence factors of selected load, represent the per unit value of n level transformer station at the load in t moment, A _{1, t}, A _{2, t}..., A _{m, t}, A _{0, t}for the weights of each influence factor.

Wherein, described load data and weather data are carried out to Least Square in Processing, obtain the weights A of each influence factor _{1, t}, A _{2, t}..., A _{m, t}, A _{0, t}concrete grammar as follows:

Adopt the load data p of 365 days in historical t moment of transformer station _{n, g, t}, known each weather data x _i(i=1,2 ..., m) and formula

$\left[\begin{array}{cccc}m+1& \underset{i=0}{\overset{m}{\mathrm{\Σ}}}{x}_i& ...& \underset{i=0}{\overset{m}{\mathrm{\Σ}}}{x}_i^m\\ \underset{i=0}{\overset{m}{\mathrm{\Σ}}}{x}_i& \underset{i=0}{\overset{m}{\mathrm{\Σ}}}{x}_i^2& ...& \underset{i=0}{\overset{n}{\mathrm{\Σ}}}{x}_i^{m+1}\\ .& .& .& .\\ .& .& .& .\\ .& .& .& .\\ \underset{i=0}{\overset{m}{\mathrm{\Σ}}}{x}_i^m& \underset{i=0}{\overset{m}{\mathrm{\Σ}}}{x}_i^{m+1}& ...& \underset{i=0}{\overset{m}{\mathrm{\Σ}}}{x}_i^{2m}\end{array}\right]\left[\begin{array}{c}{A}_{0,t}\\ A_{1,t}\\ .\\ .\\ .\\ A_{m,t}\end{array}\right]=\left[\begin{array}{c}\underset{i=0}{\overset{m}{\mathrm{\Σ}}}{P}_{n,g,t}\\ \underset{i=0}{\overset{m}{\mathrm{\Σ}}}{x}_i{P}_{n,g,t}\\ .\\ .\\ .\\ \underset{i=0}{\overset{m}{\mathrm{\Σ}}}{x}_i^m{P}_{n,g,t}\end{array}\right];(i=\mathrm{1,2},...,m;t=\mathrm{1,2},...,96)$ Obtain the weights of each influence factor.

Wherein, the parameter of transformer station and circuit is calculated to the transmission power loss between transformer station and transformer station by trend.

Wherein, in described n-k+1 level transformer station, predicting the outcome of g the t of transformer station moment equals predicting the outcome and mutual loss sum between them of the next stage t of all transformer stations moment of being directly connected with g transformer station in n-k+1 level transformer station, that is:

$P_{n-k+1,g,t}=\underset{1\∈g}{\overset{G_{n-k+2}}{\mathrm{\Σ}}}(P_{n-k+2,l,t}+{\mathrm{\ΔP}}_{n-k+2,l,t})$

Wherein, k=2 ..., n, l ∈ g is the l of transformer station being directly connected with g transformer station, G _n-k+2be the sum of transformer station in n-k+2 level transformer station, P _{n-k+1, g, t}be the prediction load in the t of the transformer station moment of g in n-k+1 level transformer station, P _{n-k2, l, t}be the prediction load in the t of the transformer station moment of l in n-k+2 level transformer station, △ P _{n-k+2, l, t}be in transformer station and the n-k+2 level transformer station of g in n-k+1 level transformer station the individual transformer station of l at the active power loss in t moment.

The present invention is in conjunction with the Electric connection characteristic between different electric pressure transformer station, utilizes the trend relation between different electric pressure transformer station, realizes the Accurate Prediction to multistage transformer station short term, and precision of prediction is high; Simultaneously for the lean analytical calculations such as the reasonable arrangement method of operation, dynamical state estimation, idle work optimization and power grid security check provide data basis and ensure.

Brief description of the drawings

Fig. 1 is the realization flow figure of the short-term load forecasting method of multistage transformer station provided by the invention;

The structural representation of the multistage transformer station that Fig. 2 provides for the embodiment of the present invention;

The multistage short-term load forecasting method realization flow figure of transformer station that Fig. 3 provides for the embodiment of the present invention.

Embodiment

In order to make object of the present invention, technical scheme and advantage clearer, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein, only for explaining the present invention, is not intended to limit the present invention.

The present invention proposes a kind of step Forecasting Methodology, fully excavating and utilizing on the basis of transformer station's part throttle characteristics, in conjunction with the Electric connection characteristic between different electric pressure transformer station, utilize the trend relation between different electric pressure transformer station, realize the Accurate Prediction to multistage transformer station short term, for the lean analytical calculations such as the reasonable arrangement method of operation, state estimation, idle work optimization and power grid security check provide data basis and ensure.

The embodiment of the present invention provides a kind of step Forecasting Methodology of the multistage transformer station short-term load forecasting for " 500kV-220kV-110kV-35kV (10kV 6kV) ", specifically comprises the steps:

The weather data of known to be predicted day, predicts the load of multistage transformer station based on step forecasting techniques, comprising:

1. obtain the historical data of transformer station and data are carried out to pre-service

1.1 obtain the historical data of transformer station

1.1.1 the canonical topology structure of involved multistage transformer station is as shown in Figure 2:

The electric pressure of the multistage transformer station of typical case includes 500kV, 220kV, 110kV, 35kV and 10kV, and there are 10.5/500kV primary substation, 500/220kV step-down substation, 500/110kV step-down substation, 220/35kV step-down substation, 220/10kV step-down substation, 110/35kV step-down substation and 110/10kV step-down substation in the transformer station comprising.The relationship between superior and subordinate of each transformer station is defined as: the primary substation of 10.5/500kV is first order transformer station, its next stage transformer station is 500/220kV and 500/110kV transformer station (second level transformer station), the next stage transformer station of 500/220kV transformer station is 220/35kV, 220/10kV, and the next stage transformer station of 500/110kV transformer station is 110/35kV and 110/10kV transformer station (third level transformer station).

In embodiments of the present invention, for convenience of description, a simplification is done to by transformer station, be specially: by 10.5/500kV transformer station referred to as 500kV transformer station, by 500/220kV transformer station referred to as 220kV transformer station, by 500/110kV transformer station referred to as 110kV transformer station, by 220/35kV, 110/35kV referred to as 35kV transformer station, by 220/10kV, 110/10kV transformer station referred to as 10kV transformer station.As shown in Figure 2:

1.1.2 predict that 10kV and the 35kV transformer station required load obtaining of load and weather data comprise:

The interval 15min 1 year 365 day every day load of totally 96 periods, 1 year 365 day every day daily maximum temperature, daily mean temperature, day lowest temperature, day high humility, per day humidity, day minimum humidity, day high wind speed, per day wind speed and daily rainfall.

1.2 data pre-service

1.2.1 get first day in historical load data day peak load as reference value, be designated as y _b; Get the meteorological numerical value of first day in historical meteorologic factor as reference value, be designated as x _{i, B}, i represents i meteorologic factor.

1.2.2 data are normalized, the historical data of obtaining, respectively divided by selected reference value, are calculated by formula (1) and (2):

${x_i}^{\′}=\frac{x_i\left(j\right)}{x_{i,B}},i=\mathrm{1,2},...,q,j,=\mathrm{1,2},...,T---\left(1\right)$

${y_i}^{\′}=\frac{y\left(j\right)}{y_B},i=\mathrm{1,2},...,n,j,=\mathrm{1,2},...,96\×T---\left(2\right)$

Wherein x _i(j) represent i meteorologic factor data; Y (j) represents j day peak load data; M represents the number of meteorologic factor; T represents total number of days.

2. utilize grey relational grade analysis (Grey Relational Analysis) to select to affect the load principal element of 10kV and 35kV transformer station; Grey Correlation Analysis Theory is as a kind of statistical analysis technique, the effective ways of analyzing correlation degree between many factors, can be in system in the indefinite or incomplete situation of data, to carrying out correlation analysis between various factors, the degree of association between various factors shows that more greatly their correlativity is larger; Otherwise, less.It has a wide range of applications at Load Prediction In Power Systems with in affect load correlation analysis, and the present invention utilizes this theory can be from the load of transformer station being affected to larger meteorologic factor on selecting the influential multiple meteorologic factor of loading.

2.1 historical datas based on after normalization, determine analytical sequence, formation sequence matrix

Using every light peak load of the 1st step gained as characteristic sequence matrix, be designated as Y=[y _max(1) ..., y _max(j) ..., y _max(T)], j=1,2 ..., T, T represents total number of days, wherein y _max(1) to y _max(T) be the day peak load of the 1st day to T days; The meteorologic factor of every day, as subsequence matrix, is designated as X=[X ₁..., X _i..., X _q], wherein i=1,2 ..., q, q represents the number of meteorologic factor, X _i=[x _i(1) ..., x _i(j) ... x _i(T)], x _i(j) be i the meteorologic factor weather data of j days, form the matrix as shown in formula (3):

$(Y,X)=\left[\begin{array}{c}{y}_{\max }\left(1\right)\\ .\\ .\\ .\\ y_{\max }\left(j\right)\\ .\\ .\\ .\\ y_{\max }\left(T\right)\end{array}\right|\left.\begin{array}{ccccc}{x}_1\left(1\right)& ...& x_i\left(1\right)& ...& x_q\left(1\right)\\ .& .& .& .& .\\ .& .& .& .& .\\ .& .& .& .& .\\ x_1\left(j\right)& ...& x_i\left(j\right)& ...& x_q\left(j\right)\\ .& .& .& .& .\\ .& .& .& .& .\\ .& .& .& .& .\\ x_1\left(T\right)& ...& x_i\left(T\right)& ...& x_q\left(T\right)\end{array}\right]---\left(3\right)$

The sequence of differences matrix Δ of 2.2 calculated loads and meteorologic factor _i(j)

Get respectively j days day peak load and the difference of each meteorologic factor value of j days form difference sequence matrix, each element △ in matrix _i(j) calculate and press formula (4), △ _i(j) represent the day peak load and i the meteorologic factor value of j days poor of j days:

△ _i(j)＝|y _max(j)-x _i(j)|，i＝1，2，…，q，j＝1，2，…，T (4)

Thereby formula (3) is converted to suc as formula the sequence of differences matrix shown in (5):

${\mathrm{\Δ}}_i\left(j\right)=\begin{array}{c}\left[\begin{array}{ccccc}{\mathrm{\Δ}}_1\left(1\right)& ...& {\mathrm{\Δ}}_i\left(1\right)& ...& {\mathrm{\Δ}}_q\left(1\right)\\ .& .& .& .& .\\ .& .& .& .& .\\ .& .& .& .& .\\ {\mathrm{\Δ}}_1\left(j\right)& ...& {\mathrm{\Δ}}_i\left(j\right)& ...& {\mathrm{\Δ}}_q\left(j\right)\\ .& .& .& .& .\\ .& .& .& .& .\\ .& .& .& .& .\\ {\mathrm{\Δ}}_1\left(T\right)& ...& {\mathrm{\Δ}}_i\left(T\right)& ...& {\mathrm{\Δ}}_q\left(T\right)\end{array}\right]---\left(5\right)\end{array}$

Maximal value △ in selection matrix (5) _maxwith minimum value △ _min,

${\mathrm{\Δ}}_{\max }=\underset{i}{\max }\underset{j}{\max }{\mathrm{\Δ}}_i\left(j\right),{\mathrm{\Δ}}_{\min }=\underset{i}{\min }\underset{j}{\min }{\mathrm{\Δ}}_i\left(j\right)---\left(6\right)$

Wherein i=1,2 ..., q, j=1,2 ..., T.

The incidence coefficient matrix of 2.3 calculated loads and meteorologic factor

By the △ in matrix (5) _iand △ (j) _max, △ _minin substitution formula (7), obtain the degree of association between day peak load and meteorologic factor, form suc as formula the degree of association matrix shown in (8):

${\mathrm{\λ}}_i\left(j\right)=\frac{{\mathrm{\Δ}}_{\min }+{\mathrm{\ρ\Δ}}_{\max }}{{\mathrm{\Δ}}_i\left(j\right)+{\mathrm{\ρ\Δ}}_{\max }},i=\mathrm{1,2},...,q,j=\mathrm{1,2},...,T---\left(7\right)$

In formula, ρ ∈ (0,1) is resolution ratio, conventionally gets 0.5.

${\mathrm{\λ}}_i\left(j\right)=\begin{array}{c}\left[\begin{array}{ccccc}{\mathrm{\λ}}_1\left(1\right)& ...& {\mathrm{\λ}}_i\left(1\right)& ...& {\mathrm{\λ}}_q\left(1\right)\\ .& .& .& .& .\\ .& .& .& .& .\\ .& .& .& .& .\\ {\mathrm{\λ}}_1\left(j\right)& ...& {\mathrm{\λ}}_i\left(j\right)& ...& {\mathrm{\λ}}_q\left(j\right)\\ .& .& .& .& .\\ .& .& .& .& .\\ .& .& .& .& .\\ {\mathrm{\λ}}_1\left(T\right)& ...& {\mathrm{\λ}}_i\left(T\right)& ...& {\mathrm{\λ}}_q\left(T\right)\end{array}\right]---\left(8\right)\end{array}$

Wherein i=1,2 ..., q; J=1,2 ..., T.

The weighted association degree of 2.4 calculated loads and meteorologic factor, being loads calculates the weighted association degree r of day peak load and i meteorologic factor by formula (9) with meteorological correlation coefficient _i:

$r_i=\underset{j=1}{\overset{T}{\mathrm{\Σ}}}{\mathrm{\ω}}_i\left(j\right){\mathrm{\λ}}_i\left(j\right),i=\mathrm{1,2},...,q;j=\mathrm{1,2},...,T---\left(9\right)$

Wherein, ω _i(j) be weighted value, represent the meteorologic factor i of j days and the correlation coefficient λ of day peak load _i(j) weight, gets in the present invention

2.5 select front 80% the influence factor that correlation coefficient is larger, as the principal element that affects this transformer station's load.

So far, obtain affecting the principal element of 10kV and 35kV transformer station load.

3. set up the relation of transformer station's load and influence factor, as the load forecasting model of 10kV and 35kV transformer station.

3.1 set up 10kV and any time t of 35kV transformer station (t=1,2 ..., 96) load and the relation of each major influence factors as forecast model, shape is as shown in (10):

$P_{35/10,t}^{*}=A_{1,t}{c}_1+A_{2,t}{c}_2+...+A_{m,t}{c}_m+A_{0,t}---\left(10\right)$

In formula, c ₁, c ₂..., c _mfor the major effect load factor of selecting, represent 10kV or the 35kV transformer station per unit value at the load in t moment, A _{1, t}, A _{2, t}..., A _{m, t}, A _{0, t}for the weights of each influence factor, ask for by 3.2 introduction methods.

3.2 load and weather datas based on historical, adopt least square method, can ask for and obtain parameter A _{1, t}, A _{2, t}... .A _{m, t}, A _{0, t}.

So far, can obtain any t moment of 10kV and 35kV transformer station (t=1,2 ..., 96) load forecasting model.

3.3 press the 10kV of to be predicted day and the weather data of 35kV transformer station after formula (1) normalization, and substitution formula (10) can obtain the per unit value of the prediction load of 10kV and 35kV transformer station respectively; The per unit value of prediction load is converted by formula (11), can obtain the famous value of 10kV and 35kV transformer station prediction load.

$P_{35/10,t}=P_{35/10,t}^{*}\×y_B---\left(11\right)$

4. adopt conventional tidal current computing method to carry out trend calculating to the system being formed by multistage transformer station, obtain the loss between each electric pressure transformer station.Based on this, adopt cascade coordination technology, obtain predicting the outcome of n-1 level transformer station according to predicting the outcome of n level transformer station, obtained the load prediction results of 110kV, 220kV and 500kV transformer station by the load prediction results recursion of 10kV and 35kV transformer station.

4.1 according to the calculation of parameter trend of transformer station and circuit, can obtain any t moment (t=1,2, ..., 96) under, power attenuation between 10kV and 110kV transformer station, between 10kV and 220kV transformer station, between 35kV and 110kV transformer station, between 35kV and 220kV transformer station, between 110kV and 500kV transformer station and between 220kV and 500kV transformer station, is designated as respectively Δ P _{110-10, t}, △ P _{220-10, t}, Δ P _{110-35, t}, Δ P _{220-35, t}, Δ P _{500-110, t}, △ P _{500-220, t}.

The 10kV that 4.2 comprehensive the 3.3rd steps obtain and any time t (t=1 of 35kV transformer station, 2,96) load prediction results and 4.1 calculate loss, according to the mode increasing by degrees, increase by degrees and obtain predicting the outcome of n-1 level transformer station the load of 110kV, 220kV and 500kV transformer station is predicted according to predicting the outcome of n level transformer station, and claim that this technology is step forecasting techniques.

Any time t (the t=1 of 110kV, 220kV and 500kV transformer station, 2,96) result of load prediction equals to add the burden with power loss between them with him 10kV and the load prediction results in corresponding moment of 35kV transformer station of direct connected next stage, shown in (12)～(14).

The t of the 110kV transformer station moment predicts the outcome: P _{110, t}=P _{110-35, t}+ Δ P _{11035, t}+ P _{110-10, t}+ Δ P _{110-10, t}(12)

The t of the 220kV transformer station moment predicts the outcome: P _{220, t}=P _{220-35, t}+ △ P _{220-35, t}+ P _{220-10, t}+ △ P _{220-10, t}(13)

The t of the 500kV transformer station moment predicts the outcome: P _{500, t}=P _{220, t}+ Δ P _{500-220, t}+ P _{110, t}+ Δ P _{500-110, t}(14)

Wherein P _{110, t}represent the prediction load in the 110kV t of transformer station moment, P _{220, t}represent the prediction load in the 220kV t of transformer station moment, P _{500, t}represent the prediction load in the 500kV t of transformer station moment; P _{110-35, t}, P _{110-10, t}represent the 35kV and the prediction in the 10kV t of the transformer station moment load that are directly connected with 110kV transformer station, P _{220-35, t}, P _{220-10, t}represent the 35kV and the load prediction results in the 10kV t of transformer station moment that are directly connected with 220kV transformer station.

So far, completed from any time t of 500kV～10kV transformer station (t=1,2 ..., 96) load prediction.

The step Forecasting Methodology that the present invention proposes is fully being excavated and is being utilized on the basis of 10kV and 35kV transformer station part throttle characteristics, sets up the relation of load and major influence factors, obtains forecast model; Based on this, system is carried out to trend calculating, utilize the trend relation between different electric pressure transformer station, recursion obtains the load prediction results of 110kV, 220kV and 500kV transformer station; With dispensing method in proportion with directly apply mechanically region load forecasting method and compare, the present invention has not only considered that different meteorologic factors is different from the correlativity of electric load, choose the meteorologic factor larger with electric load correlativity according to grey relation theory, again selected meteorologic factor and historical Power system load data are carried out to load prediction to each transformer station, this has reduced greatly because blindly choosing the predicated error that meteorologic factor is brought, also considered the Electric connection characteristic between transformer station and transformer station, therefore the inventive method has higher precision of prediction simultaneously.

For the Forecasting Methodology that further illustrates that the embodiment of the present invention provides, below in conjunction with accompanying drawing, the invention process case is described in further detail the process of multistage transformer station load prediction.

The weather datas such as known daily maximum temperature to be predicted day, daily mean temperature, day lowest temperature, day high humility, per day humidity, day minimum humidity, day high wind speed, per day wind speed and daily rainfall, to this day interval 15min totally 96 periods loads predict.

Implementation step 1: obtain needed historical data as shown in table 1.

The load that table 1 comprises and weather data

Implementation step 2: according to step 2, load and weather data are carried out to the calculating of the grey degree of association, choose front 80% the meteorologic factor larger with the correlation coefficient of loading as major influence factors.Selection result is as shown in table 2, obtains the highest temperature, temperature on average, the lowest temperature and high humility and load and is related to maximum.

The load of table 22010 year and the meteorological degree of association

Implementation step 3: set up any t (t=1,2 ..., 96) and 10kV and the 35kV transformer station load forecasting model in moment.

For example, the load forecasting model in the 1st moment is as follows:

220/35kV transformer station primary mold: P _220/15,1=0.373-0.044x ₁+ 0.135x ₂+ 0.03x ₃-0.110x ₄;

220/10kV transformer station primary mold: P _220/10,1=0.186-0.022x ₁+ 0.067x ₂+ 0.015x ₃-0.055x ₄;

110/35kV transformer station primary mold: P _110/35,1=0.186-0.022x ₁-0.067x ₂+ 0.015x ₃-0.055x ₄;

110/10kV transformer station primary mold: P _110/10,1=0.093-0.011x ₁+ 0.034x ₂+ 0.008x ₃-0.028x ₄;

The weather data of to be predicted day is normalized to the above-mentioned 10kV of rear substitution and 35kV transformer station forecast model, can obtain this day any time t (t=1,2 ... 96) per unit value of load prediction, obtains the famous value that prediction is loaded according to formula (11) after converting.

For example, the load prediction results in the 1st moment, carry out renormalization according to formula (11) after result as shown in table 3:

Table 310kV and 35kV transformer station load prediction results

Implementation step 4: system is carried out to trend calculating, obtain the burden with power loss between adjacent substations, utilize step forecasting techniques, recursion obtains the prediction load of 110kV, 220kV and 500kV transformer station.

For example, adopt conventional tidal current computing method or software, system is carried out to trend calculating, the loss calculating between the 1st moment transformer stations at different levels is:

ΔP _110-10,1＝2.7MW、ΔP _110-35,1＝2.7MW、ΔP _220-10,1＝2.7MW、ΔP _220-35,1＝2.7MW、ΔP _500-110,1＝1.6MW、ΔP _500-220,1＝1.6MW。

Thereby obtain corresponding the 1st period, the prediction load of 110kV, 220kV and 500kV transformer station is as follows:

P _110，1＝P _110-35,1+P _110-10,1+ΔP _110-35,1+ΔP _110-10,1＝63.01+31.50+2.70+2.70＝99.91；

P _220,1＝P _220-35,1+P _220-10,1+ΔP _220-35,1+ΔP _220-10,1＝126.03+63.01+2.70+2.70＝194.44；

P _500,1＝P _110，1+P _220,1+ΔP _500-110,1+ΔP _500-220,1＝99.91+194.44+1.60+1.60＝297.55。

To sum up, the prediction of 110kV, 220kV and 500kV transformer station load is as shown in table 4.

The load prediction results of table 4110kV, 220KV and 500kV

Implementation step 5: the precision of prediction to the inventive method is tested, chooses the data in 7 days weeks and carries out predicated error analysis, and predicated error computing formula is pressed formula (15) and calculated.

For example, the test result of 500kV transformer station is as shown in table 5:

Actual value and the predicted value in table 7 day the 1st moment of 5500kV transformer station

Actual value (MW)	Predicted value (MW)	Relative error (%)
			298.20	328.69	10.22
328.63	335.99	2.24
			336.12	351.00	4.43
350.72	334.14	4.73
			335.37	317.43	5.35
329.15	311.37	5.40
			344.53	328.08	4.77

The test result of 220/35kV transformer station is as shown in table 6:

Actual value and the predicted value in table 7 day the 1st moment of 6220/35kV transformer station

Actual value (MW)	Predicted value (MW)	Relative error (%)
			113.31	123.47	8.97
124.88	127.80	2.34
			127.72	127.80	0.06
133.27	124.42	6.64
			127.44	124.25	2.50
125.07	126.46	1.11
			130.92	126.64	3.27

Find through test, the load prediction error of transformer station is substantially in 10%, and method provided by the present invention has higher precision of prediction.

Those skilled in the art will readily understand; the foregoing is only preferred embodiment of the present invention; not in order to limit the present invention, all any amendments of doing within the spirit and principles in the present invention, be equal to and replace and improvement etc., within all should being included in protection scope of the present invention.

Claims (6)

1. a short-term load forecasting method for multistage transformer station, is characterized in that, comprises the steps:

(1) obtain the historical data of n level transformer station, and described historical data is carried out to pre-service; N is more than or equal to 3 positive integer;

(2) based on grey correlation theory analysis, pretreated historical data is processed, obtained the principal element of the load that affects n level transformer station;

(3) set up forecast model according to the principal element of the load that affects n level transformer station, and obtain the load prediction results of any time t of n level transformer station according to the weather data of n level transformer station to be predicted day and described forecast model;

(4) according to the calculation of parameter trend of transformer station and circuit obtain in n level transformer station in each transformer station and n-1 level transformer station in the power attenuation between each transformer station, n-1 level transformer station the power attenuation between each transformer station in each transformer station and n-2 level transformer station ... and the power attenuation between each transformer station in each transformer station and n-k level transformer station in n-k+1 level transformer station; K=1,2 ... n; Wherein, so that from generating plant, transformer station is farthest as n level, from generating plant, nearest transformer station is the 1st grade;

(5) utilize regression analysis to obtain the load prediction results of any time t of n level transformer station according to historical load and weather data;

(6) according to each power attenuation in load prediction results and the step (4) of any time t of n level transformer station in step (5) obtain the n-1 level t of transformer station moment load prediction results, the n-2 level t of transformer station moment load prediction results ... and the n-k level t of transformer station moment load prediction results.

2. short-term load forecasting method as claimed in claim 1, it is characterized in that, described historical data comprises load data and weather data, be specially the interval 15min 1 year 365 day every day load of totally 96 periods, 1 year 365 day every day daily maximum temperature, daily mean temperature, day lowest temperature, day high humility, per day humidity, day minimum humidity, day high wind speed, per day wind speed and daily rainfall.

3. short-term load forecasting method as claimed in claim 1, is characterized in that, for g transformer station in n level transformer station, and g=1,2 ..., G _n, G _nbe the sum of transformer station in n level transformer station, described forecast model is c ₁, c ₂..., c _mfor the perunit value of the major influence factors of selected load, represent the per unit value of n level transformer station at the load in t moment, A _{1, t}, A _{2, t}..., A _{m, t}, A _{0, t}for the weights of each influence factor.

4. short-term load forecasting method as claimed in claim 3, is characterized in that, described load data and weather data are carried out to Least Square in Processing, obtains the weights A of each influence factor _1,t, A _{2, t}..., A _{m, t}, A _{0, t}concrete grammar as follows:

Adopt the load data P of 365 days in historical t moment of transformer station _{n, g, t}, known each weather data x _i(i=1,2 ..., m) and formula

$\left[\begin{array}{cccc}m+1& \underset{i=0}{\overset{m}{\mathrm{\Σ}}}{x}_i& ...& \underset{i=0}{\overset{m}{\mathrm{\Σ}}}{x}_i^m\\ \underset{i=0}{\overset{m}{\mathrm{\Σ}}}{x}_i& \underset{i=0}{\overset{m}{\mathrm{\Σ}}}{x}_i^2& ...& \underset{i=0}{\overset{n}{\mathrm{\Σ}}}{x}_i^{m+1}\\ .& .& .& .\\ .& .& .& .\\ .& .& .& .\\ \underset{i=0}{\overset{m}{\mathrm{\Σ}}}{x}_i^m& \underset{i=0}{\overset{m}{\mathrm{\Σ}}}{x}_i^{m+1}& ...& \underset{i=0}{\overset{m}{\mathrm{\Σ}}}{x}_i^{2m}\end{array}\right]\left[\begin{array}{c}{A}_{0,t}\\ A_{1,t}\\ .\\ .\\ .\\ A_{m,t}\end{array}\right]=\left[\begin{array}{c}\underset{i=0}{\overset{m}{\mathrm{\Σ}}}{P}_{n,g,t}\\ \underset{i=0}{\overset{m}{\mathrm{\Σ}}}{x}_i{P}_{n,g,t}\\ .\\ .\\ .\\ \underset{i=0}{\overset{m}{\mathrm{\Σ}}}{x}_i^m{P}_{n,g,t}\end{array}\right];i=\mathrm{1,2},...,m;t=\mathrm{1,2},...,96;$ Obtain the weights of each influence factor.

5. short-term load forecasting method as claimed in claim 1, is characterized in that, utilizes the parameter of transformer station and circuit to calculate the transmission power loss between transformer station and transformer station by trend.

6. short-term load forecasting method as claimed in claim 1, it is characterized in that, in described n-k+1 level transformer station, predicting the outcome of the t of the transformer station moment of g equals predicting the outcome and mutual loss sum between them of the next stage t of all transformer stations moment of being directly connected with the transformer station of g in n-k+1 level transformer station, that is:

$P_{n-k+1,g,t}=\underset{1\∈g}{\overset{G_{n-k+2}}{\mathrm{\Σ}}}(P_{n-k+2,l,t}+{\mathrm{\ΔP}}_{n-k+2,l,t})$

Wherein, k=2 ..., n, l ∈ g is the l of transformer station being directly connected with g transformer station, G _n-k+2be the sum of transformer station in n-k+2 level transformer station, P _{n-k+1, g, t}be the prediction load in the t of the transformer station moment of g in n-k+1 level transformer station, P _{n-k+2, l, t}the prediction in the t of the transformer station moment of l load in n-k+2 level transformer station, △ P _{n-k+2, l, t}be in transformer station and the n-k+2 level transformer station of g in n-k+1 level transformer station the individual transformer station of l at the active power loss in t moment.