Summary of the invention
The object of the present invention is to provide a kind of high-voltage fence ability to transmit electricity quick calculation method with Spatio-Temporal Label based on procedure-oriented theory.The ability to transmit electricity quick calculation method with Spatio-Temporal Label is built according to procedure-oriented theory, set up the ability to transmit electricity electric flux model towards time course and power module, characteristic based on pattern-recognition is extracted, and finds the maximal value and minimum value that characterize section ability to transmit electricity, calculates fast.Adopt computing method of the present invention to overcome existing computing method and can only adopt historical data, can not real-time online process mass data and computational problem, there is strong adaptability, use simple, the feature that computing velocity is fast, can be applied to the quick calculating of processor-oriented online high-voltage fence ability to transmit electricity.
For achieving the above object, the invention provides a kind of high-voltage fence ability to transmit electricity quick calculation method with Spatio-Temporal Label set up based on procedure-oriented theory, described method adopts mode identification method to carry out compression process to mass data, introduce Lagrange factor, be optimized algorithm, these computing method comprise the steps:
Steps A: set up electrical network to be asked has Spatio-Temporal Label ability to transmit electricity electric flux model and power module towards time course, realize transmitting capacity of the electric wire netting and calculate in real time online, calculating data acquisition WAMS device is distant copies remote measurement real time data;
Step B: build the ability to transmit electricity concept based on time effects, the concept of definition time transmission cross-section, time course, described ability to transmit electricity is section ability to transmit electricity and process ability to transmit electricity, pass between described time course and time transmission cross-section is that the set of time transmission cross-section forms time course, realizes the infinitesimal analysis that section ability to transmit electricity and process ability to transmit electricity calculate and changes;
Step C: the characteristic set up based on pattern-recognition extracts fast algorithm, first adopts Karhunen-Loeve transformation method that the magnanimity real time data characterizing section ability to transmit electricity is carried out dimensionality reduction compression; Adopt power method and inverse power method maximizing and minimum value in the set characterizing section ability to transmit electricity again, make other section ability to transmit electricity numerical value between these two numerical value, improve computing velocity;
Step D: introduce Lagrange multiplier and set up the optimized algorithm based on section ability to transmit electricity and process ability to transmit electricity with Spatio-Temporal Label, for improve computing velocity and calculating accuracy rate.
Wherein, in described step B, following formula is adopted to carry out definition time transmission cross-section:
A
i(t)={N
i(t),P
i(t),Ctr
i(t),G
i(t)} (1)
In formula: use A
it () represents time transmission cross-section, N
it () represents time dependent network parameter, P
it () represents time varying system status information amount, Ctr
it () represents time dependent control information amount, G
it () represents time dependent relation/logic.
Wherein, in described step C, first adopt Karhunen-Loeve transformation method that the magnanimity real time data characterizing section ability to transmit electricity is carried out dimensionality reduction compression, then adopt the concrete methods of realizing of power method and inverse power method maximizing and minimum value in the set of sign section ability to transmit electricity as follows:
P=(p
1,p
2,......p
n)
T
Wherein: X is power to be asked, P is realtime power proper vector,
n × n square formation,
it is n dimensional vector; Matrix
by the column vector of n Line independent
composition, if so only use m feature, then error delta X (m) is
Wherein, X is power to be asked,
m () is the estimator of X
By the criterion of mean square deviation as the validity of tolerance m proper vector subset, have
Will
m () is set to b
iwith
function; In order to by the realtime power characterization time section ability to transmit electricity being mapped to the lower feature space of dimension, and not influence time section ability to transmit electricity size or reduce gap as far as possible, then ask for making
the b of the best of (m) minimalization
iwith
value,
Try to achieve
In other words for the component p of those P do not retained
i, replace just obtaining best b with mean value
ivalue, best b
iafter trying to achieve, try to achieve
m () is as follows
∑
X=E[(X-E(X))(X-E(X))
T] (8)
Formula (8) is the covariance matrix of X, introduces Lagrange multiplier λ in this formula
i, obtain formula (9)
m () minimum necessary condition is:
Calculating is tried to achieve:
Make
become the covariance matrix ∑ of X
xlatent vector, and λ
ibecome i-th eigenvalue of covariance matrix, so the best
after trying to achieve, tried to achieve by following formula (12)
(m):
Wherein, the concrete grammar setting up optimized algorithm in described step D is as follows:
Objective function:
Equality constraint:
H
i(t)=0 (14)
Inequality constrain condition:
N
min≤N
i(t)≤N
max(15)
Ctr
min≤Ctr
i(t)≤Ctr
max
G
min≤G
i(t)≤G
max
Wherein: f (t) represents electric flux procedure function, P (t) represents not the effective value of power samples in the same time, P
avgrepresent average power in one-period, N
it () represents time dependent network parameter, Ctr
it () represents time dependent control information amount, G
it () represents time dependent relation/logic; Slack variable n is introduced in three inequality constrains of formula (15)
1i(t), n
2i(t), c
1i(t), c
2i(t), g
1i(t), g
2it (), all slack variables are all greater than 0, then inequality constrain is separately converted to equality constraint
N
i(t)+n
1i(t)=N
max(16)
N
i(t)-n
2i(t)=N
min
Ctr
i(t)+c
1i(t)=Ctr
max
Ctr
i(t)-c
2i(t)=Ctr
min
G
i(t)+g
1i(t)=G
max
G
i(t)-g
2i(t)=G
min
Introduce barrier parameter u, v, w, and barrier parameter is all greater than zero, transform objective function as barrier function, this barrier function is similar to former objective function f (t feasible zone planted agent simultaneously
i), and become very large when border, therefore obtain objective function:
maxf
i(t)-u(∑linn
1i(t)+∑linn
2i(t))-v(∑linc
1i(t)+∑linc
2i(t))-
w(∑ling
1i(t)+∑ling
2i(t)) (17)
Make U
i(t)=u ∑ linn
1i(t)+u ∑ linn
2i(t) (18)
V
i(t)=v∑linc
1i(t)+v∑linc
2i(t)
W
i(t)=w∑ling
1i(t)+w∑ling
2i(t)
Then objective function is reduced to maxf
i(t)-U
i(t)-V
i(t)-W
i(t) (19)
Equality constraint
H
i(t)=0 (20)
N
i(t)+n
1i(t)=N
max
N
i(t)-n
2i(t)=N
min
Ctr
i(t)+c
1i(t)=Ctr
max
Ctr
i(t)-c
2i(t)=Ctr
min
G
i(t)+g
1i(t)=G
max
G
i(t)-g
2i(t)=G
min
Solving of this function can be converted into optimization extreme-value problem, introduces Lagrange multiplier λ
1, λ
2, λ
3, λ
4, λ
5, λ
6, order
X
i(t)=λ
1(N(t
i)+n
1(t
i)-N
max)+λ
2(N(t
i)-n
2(t
i)-N
min) (21)
Y
i(t)=λ
3(Ctr(t
i)+c
1(t
i)-Ctr
max)+λ
4(Ctr(t
i)-c
2(t
i)+Ctr
min)
Z
i(t)=λ
5(G(t
i)+g
1(t
i)-G
max)+λ
6(G(t
i)+g
2(t
i)+G
min)
Obtain Lagrangian function maxf
i(t)-U
i(t)-V
i(t)-W
i(t)-X
i(t)-Y
i(t)-Z
i(t) (22)
F
i(t)=maxf
i(t)-U
i(t)-V
i(t)-W
i(t)-X
i(t)-Y
i(t)-Z
i(t) (23)
This problem maximum value existent condition is the partial derivative of Lagrangian function to all variablees and Lagrange multiplier is zero, asks local derviation, make derivative be zero, solve above-mentioned formula.
Beneficial effect of the present invention: a kind of high-voltage fence ability to transmit electricity quick calculation method with Spatio-Temporal Label set up based on procedure-oriented theory of the present invention overcomes existing computing method and can only adopt historical data, can not real-time online process mass data and computational problem, there is strong adaptability, use simple, the feature that computing velocity is fast, can be applied to the quick calculating of processor-oriented online high-voltage fence ability to transmit electricity.
Embodiment
As shown in Figure 3, a kind of step with the high-voltage fence ability to transmit electricity quick calculation method of Spatio-Temporal Label based on the foundation of procedure-oriented theory of the present invention is as follows:
Steps A: set up electrical network to be asked has Spatio-Temporal Label ability to transmit electricity electric flux model and power module towards time course, realize transmitting capacity of the electric wire netting and calculate in real time online, calculating data acquisition WAMS device is distant copies remote measurement real time data;
Step B: build the ability to transmit electricity concept based on time effects, the concept of definition time transmission cross-section, time course, described ability to transmit electricity is section ability to transmit electricity and process ability to transmit electricity, pass between described time course and time transmission cross-section is that the set of time transmission cross-section forms time course, realizes the infinitesimal analysis that section ability to transmit electricity and process ability to transmit electricity calculate and changes;
Step C: the characteristic set up based on pattern-recognition extracts fast algorithm, first adopts Karhunen-Loeve transformation method that the magnanimity real time data characterizing section ability to transmit electricity is carried out dimensionality reduction compression; Adopt power method and inverse power method maximizing and minimum value in the set characterizing section ability to transmit electricity again, make other section ability to transmit electricity numerical value between these two numerical value, improve computing velocity;
Step D: introduce Lagrange multiplier and set up the optimized algorithm based on section ability to transmit electricity and process ability to transmit electricity with Spatio-Temporal Label, for improve computing velocity and calculating accuracy rate.
What procedure-oriented of the present invention theory was set up have in the step B of the high-voltage fence ability to transmit electricity quick calculation method of Spatio-Temporal Label proposes formula and derivation is as follows:
As shown in Figure 1, transmission cross-section divides according to Different factor the one group of interconnection summation formed.For reflecting the real shape in transmission cross-section region, adopt projective representation's method of section, cutting plane perpendicular to by the axis of cutting object or outline line, only should represent the shape in cross sectional area, and must mark section symbols in cross sectional area during projection.The curve of ability to transmit electricity can matching, and its calculating is made up of a series of parameter that is mutually related, quantity of state, controlled quentity controlled variable etc., and therefore any one transmission cross-section is all made up of a series of parameter that is mutually related, quantity of state, controlled quentity controlled variable etc.
A
i(t)={N
i(t),P
i(t),Ctr
i(t),G
i(t)} (1)
In formula: use A
it () represents time transmission cross-section, N
it () represents time dependent network parameter, P
it () represents time varying system status information amount, Ctr
it () represents time dependent control information amount, G
it () represents time dependent relation/logic.
From geometric meaning, by procedure function f (t) (t
b≤ t≤t
e) the trapezoidal area in the bent limit that forms, equal certain section amount f (t
ξ) area of each piecemeal rectangle corresponding to horizontal linear section, as shown in Figure 1, wherein horizontal ordinate t represents the time, and ordinate f (t) represents procedure function, P
avg(ξ) average power in one-period is represented.
Therefore ask for ability to transmit electricity key be to locate P (t) characterization functional integration and an important section.
Then based on the section ability to transmit electricity objective function of time course:
Equality constraint:
H
i(t)=0 (14)
Inequality constrain condition:
N
min≤N
i(t)≤N
max(15)
Ctr
min≤Ctr
i(t)≤Ctr
max
G
min≤G
i(t)≤G
max
Wherein: f (t) represents electric flux procedure function, P (t) represents not the effective value of power samples in the same time, P
avgrepresent average power in one-period.N
it () represents time dependent network parameter, Ctr
it () represents time dependent control information amount, G
it () represents time dependent relation/logic.Slack variable n is introduced in above three inequality constrains
1i(t), n
2i(t), c
1i(t), c
2i(t), g
1i(t), g
2it (), all slack variables are all greater than 0, then inequality constrain is separately converted to equality constraint
N
i(t)+n
1i(t)=N
max(16)
N
i(t)-n
2i(t)=N
min
Ctr
i(t)+c
1i(t)=Ctr
max
Ctr
i(t)-c
2i(t)=Ctr
min
G
i(t)+g
1i(t)=C
max
G
i(t)-g
2i(t)=G
min
Introduce barrier parameter u, v, w, and barrier parameter is all greater than zero simultaneously.Transform objective function as barrier function.This function is similar to former objective function f (t feasible zone planted agent
i), and become very large when border.Therefore objective function can be obtained:
maxf
i(t)-u(∑linn
1i(t)+∑linn
2i(t))-v(∑linc
1i(t)+∑linc
2i(t))-
w(∑ling
1i(t)+∑ling
2i(t)) (17)
Make U
i(t)=u ∑ linn
1i(t)+u ∑ linn
2i(t) (18)
V
i(t)=v∑linc
1i(t)+v∑linc
2i(t)
W
i(t)=w∑ling
1i(t)+w∑ling
2i(t)
Then objective function is reduced to maxf
i(t)-U
i(t)-V
i(t)-W
i(t) (19)
Equality constraint
H
i(t)=0 (20)
N
i(t)+n
1i(t)=N
max
N
i(t)-n
2i(t)=N
min
Ctr
i(t)+c
1i(t)=Ctr
max
Ctr
i(t)-c
2i(t)=Ctr
min
G
i(t)+g
1i(t)=G
max
G
i(t)-g
2i(t)=G
min
Solving of this function can be converted into optimization extreme-value problem, introduces Lagrange multiplier λ
1, λ
2, λ
3, λ
4, λ
5, λ
6, order
X
i(t)=λ
i(N(t
i)+n
1(t
i)-N
max)+λ
2(N(t
i)-n
2(t
i)-N
min) (21)
Y
i(t)=λ
3(Ctr(t
i)+c
1(t
i)-Ctr
max)+λ
4(Ctr(t
i)-c
2(t
i)+Ctr
min)
Z
i(t)=λ
5(G(t
i)+g
1(t
i)-G
max)+λ
6(G(t
i)+g
2(t
i)+G
min)
Obtain Lagrangian function maxf
i(t)-U
i(t)-V
i(t)-W
i(t)-X
i(t)-Y
i(t)-Z
i(t) (22)
F
i(t)=maxf
i(t)-U
i(t)-V
i(t)-W
i(t)-X
i(t)-Y
i(t)-Z
i(t) (23)
This problem maximum value existent condition is the partial derivative of Lagrangian function to all variablees and multiplier is zero.Local derviation is asked to above-mentioned formula, makes derivative be zero, solve.
So far, the ability to transmit electricity electric flux algorithm model based on time course has been built successfully.Because the object of building the ability to transmit electricity electric flux algorithm of time course being in order to ensureing that the ability to transmit electricity under power grid security prerequisite is maximum, therefore mass data being compressed, extreme value is asked to above-mentioned model.
Its basic thought be exactly high dimension vector is mapped to the lower feature space of dimension and not Effect Mode classification.Namely from original n-dimensional space N to a mapping of the lower m dimensional feature space M of dimension, m < n.Here discontinuity surface pattern when random vector X represents is tieed up with n.If g:N → M, X/ → P=g (x) here P is exactly a realtime power proper vector chosen.G is the mapping from N to M.Object is by the not Effect Mode classification to the lower feature space of dimension of higher-dimension instantaneous power maps feature vectors.
P=(p
1,p
2,......p
n)
T
Wherein: P is realtime power proper vector,
n × n square formation,
it is n dimensional vector.Matrix
by the column vector of n Line independent
composition, if so we are only by m feature, then error is
By the criterion of mean square deviation as the validity of tolerance m proper vector subset, have
By son
m () is set to b
iwith
function, in order to by the realtime power characterization time section ability to transmit electricity being mapped to the lower feature space of dimension, and not influence time section ability to transmit electricity size or reduce gap as far as possible, then ask for making
the b of the best of (m) minimalization
iwith
value.
Try to achieve
In other words for the component p of those P do not retained
i, replace just obtaining best b with mean value
ivalue.Best b
iafter trying to achieve, try to achieve
m () is as follows
∑
X=E[(X-E(X))(X-E(X))
T] (8)
Formula (8) is the covariance matrix of X, introduces Lagrange multiplier λ in this formula
i, obtain formula (9)
m () minimum necessary condition is:
Calculating is tried to achieve:
Make
become the covariance matrix ∑ of X
xlatent vector, and λ
ibecome i-th eigenvalue of covariance matrix, so the best
after trying to achieve, tried to achieve by following formula (12)
(m):
In pattern-recognition, p
1, p
2..., p
nregard the feature chosen from original measurement X as, these features have following feature:
(1) each feature p
irepresenting the validity in X by the eigenvalue λ corresponding with it
idetermined.If a feature p
ibe omitted, then square error increases λ
i.Feature large for those eigenvalues should be remained as much as possible.If eigenvalue descending order is numbered, make
λ
1≥λ
2≥......λ
n>0 (24)
Then it can be used as the foundation of feature selecting.
(2) each feature is uncorrelated mutually, and that is the covariance matrix of P is diagonal angle, this is because:
When X be normal distribution in particular cases, each feature pi is separate.
(3) ∑
xlatent vector make
m () is minimum in all selections of orthonormal base vector.
Below that the present invention calculates on the basis of data at IEEE30 node power transmission network, adopt the high-voltage fence ability to transmit electricity quick calculation method with Spatio-Temporal Label to calculate, adopt Karhunen-Loeve transformation method that the magnanimity real time data characterizing section ability to transmit electricity is carried out the example of dimensionality reduction compression.
Fig. 2 is that the present invention adopts Karhunen-Loeve transformation method the magnanimity real time data characterizing section ability to transmit electricity to be carried out the IEEE30 node electrical transmission network systems figure of dimensionality reduction compression employing.
If branch power P
ithe instantaneous power proper vector of a sign section ability to transmit electricity chosen, can find out P exactly
ibe 41 × 41 rank matrixes, namely can obtain P by original table
41 × 41be a sparse matrix, can replace with X.Can be reduced to
X/→P
41×41(26)
By 41 × 41 rank matrix compressions of P, replace P with lower-order approximate matrix
41 × 41.If
for unit matrix, so
In this example, the IEEE30 node transmission system parameter of IEEE30 node power transmission network alternator data and data processing is as shown in Table 1 and Table 2:
Table 1IEEE30 node power transmission network alternator data
The IFFE30 node transmission system parameter of table 2 data processing
Ergodic algorithm is adopted to find maximum power value P
maxwith minimal power values P
min, ask for
with this number percent for reference value, setting one is greater than the number percent of this reference value as the cut off value judging to affect ability to transmit electricity size, selecting in this example 20% is cut off value: respectively by the active power of each branch road divided by maximum power value, by the absolute value of acquired results respectively compared with 20%, be less than or equal to 20%, namely this branch road is little on ability to transmit electricity impact; Be greater than 20%, namely this branch road is larger on ability to transmit electricity impact.
Error analysis: replace with previously selected constant those component subset not retaining P, form following estimator
The variance of X: E (X)=1.46 as calculated
get unit matrix:
Can obtain: b
i=1.46
The above-mentioned active power data by each branch road are divided into little on ability to transmit electricity impact and larger on ability to transmit electricity impact two parts, select to form new matrix to the larger part data of ability to transmit electricity impact and replace former P in this example
41 × 41matrix, then error is
Then former P
41 × 41matrix only replaces just to calculate ability to transmit electricity fast with the new matrix that dimension ability to transmit electricity being affected to larger part data composition is lower.
Procedure schema feature extraction: construct this part example raw data.Suppose that certain system divides is 4 regions, wherein region 1 is the interregional interconnection section be connected with its exterior, and region 2, region 3, region 4 are internal system section.For simplifying, consider that structure time course is one hour, gathered section active power data every 15 minutes, fraction part is omitted, and only retains integral part.Its section active power data see the following form 3:
Table 3 section active power data (MW)
Its vector table is:
If there is a number λ and non-vanishing vector, then meet: px=λ x (29)
Utilize power method iterative equation formula and iterative program, calculating is tried to achieve proper vector and is referred to table 4:
The characteristic value data that table 4 is tried to achieve by dark method
Utilize inverse power method iterative equation formula and iterative program, calculating is tried to achieve proper vector and is referred to table 5:
The characteristic value data that table 5 is tried to achieve by anti-dark method
Therefore power method is utilized to obtain eigenvalue of maximum λ
1=2976.426, the value 2036 in corresponding matrix P is exactly maximum vector, i.e. maximum transmission cross-section.Inverse power method is utilized to obtain eigenvalue of maximum λ
1=78.8769, the value 189 in corresponding matrix P is exactly minimum vector, i.e. minimum transmission cross-section.
Convergence:
Because of the proper vector x of P
1..., x
nbetween be linear independence,
power method is adopted to carry out first time iteration
can obtain when kth walks
due to
as k → ∞, vectorial y
(k)will with proper vector x
1direction reach unanimity, and due to the impact of round-off error, to make along x
1direction has nonzero element to there is (initial vector x
0except) therefore the method be convergence.
It is more than the detailed description of the present invention being carried out for the ease of understanding the present invention; but it may occur to persons skilled in the art that; not departing from the scope that claim of the present invention contains and can also make other changes and modifications, these changes and amendment are all in protection scope of the present invention.