CN106443566A - Big data deduction based electric energy metering device error detection method and system - Google Patents
Big data deduction based electric energy metering device error detection method and system Download PDFInfo
- Publication number
- CN106443566A CN106443566A CN201611092500.8A CN201611092500A CN106443566A CN 106443566 A CN106443566 A CN 106443566A CN 201611092500 A CN201611092500 A CN 201611092500A CN 106443566 A CN106443566 A CN 106443566A
- Authority
- CN
- China
- Prior art keywords
- electric energy
- epsiv
- error
- virtual
- formula
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R35/00—Testing or calibrating of apparatus covered by the other groups of this subclass
- G01R35/04—Testing or calibrating of apparatus covered by the other groups of this subclass of instruments for measuring time integral of power or current
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a big data deduction based electric energy metering device error detection method and system. The method includes the steps: acquiring multiple sets of electric energy data of each circuit, calculating bus electricity quantity non-balance ratio, building a mapping model of electric energy and bus electricity quantity non-balance ratio, calculating close degree of electric energy data and a virtual load model, selecting a plurality of sets of electric energy data according to the close degree to form a training sample set, taking the bus electricity quantity non-balance ratio as a teacher sample set to train the artificial neural network, and performing deduction and reconfiguration to obtain whole error. The system comprises an electric energy data acquisition unit, a bus electricity quantity non-balance ratio calculation unit, a virtual load model building unit, a close degree calculation unit, a sample set generation unit, an artificial neural network training unit and a whole error calculation unit. The method and system can monitor whole metering error of electric energy metering devices in real time at the same time, and has the advantages of avoidance of field test, simplicity in realization method, real-time performance and quickness in detection, high efficiency and safety.
Description
Technical field
The present invention relates in power equipment electric power meter state estimation technology, specifically disclose a kind of based on several greatly
According to the electric power meter error detection method deduced and system.
Background technology
The accurate metering of electric energy is to ensure that the fair basis of electricity clearing, causes because metering device error is overproof every year
Wrong electric energy is considerable.On the one hand, presently, there are the technological means of multiple detection metering device errors.Traditional method has:Electricity
The laboratory verification of energy table and Site Detection, the laboratory verification of transformer and Site Detection, voltage transformer secondary voltage drop are existing
Field detection etc..Metering device is dismantled laboratory and is carried out by these job demand personnel, or need personnel by standard set-up with
Testing equipment takes scene development to, and some experiments must also have a power failure and carry out, and convenience is not enough.Relatively new technological means have:Electricity
Can the detection of Watch Error remote online, the detection of transformer error remote online, the detection of voltage transformer secondary voltage drop remote online etc..
These relatively new technology can remotely, the error of real-time detection metering device, but exist cost high, arrangement inconvenience, deposit
It is also possible to affect the normal operation of power system the problems such as potential faults.
On the other hand, power information acquisition system has covered all responsible consumers, transformer station, has opened through building for many years
Close station, be widely used in the sides such as automatic data logging management, ordered electric management, electricity consumption Statistic Analysis, line loss and loss on transmission analysis
Face, and bring the electric energy data of magnanimity, provide and support for the analysis of each link of electric power enterprise operation and management, decision-making, for reality
Now intelligent two-way interaction service provision information basis.
Electric energy data derives from metering device, had both reflected electric flux conveying situation, and had also contained the mistake of metering device
Difference information.Because the error of metering device is relatively small, the control information amount containing in electric energy data also very faint, very in the past
Hardly possible realizes error this defect state overproof by a small amount of electric energy data.But the electricity that power information acquisition system is provided that now
Energy datum considerably beyond conventional, can be analysed in depth by suitable mathematical model, makes data by these data resources
In the metering device control information that contains integrated, and clearly present, this is the brand-new way evaluating metering device error
Footpath.
Content of the invention
The technical problem to be solved in the present invention is:For the problems referred to above of prior art, provide one kind simultaneously real-time
The overall error in dipping of monitoring electric power meter, is not required to field test, implementation method is simple, detect real-time, efficient peace
The electric power meter error detection method based on big data deduction of full advantage and system.
In order to solve above-mentioned technical problem, the technical solution used in the present invention is:
On the one hand, the present invention provides a kind of electric power meter error detection method based on big data deduction, step bag
Include:
1) obtain multigroup electric energy data of each bar circuit in transformer station's given voltage region to be assessed;
2) corresponding bus Power unbalance rate is calculated according to every group of electric energy data, according to multigroup electric energy data
And corresponding bus Power unbalance rate sets up the mapping model of electric flux and bus Power unbalance rate;
3) create N bar virtual circuit, N be transformer station's given voltage region to be assessed interior lines way, take any two virtual
Railway superstructures circuit pair, by one of circuit to being set to current line pair, sets the metering electricity of a wherein virtual circuit
Being worth for mete-wand, another virtual circuit is comparison other;The electric energy value of the virtual circuit of current line pair is set respectively
For 1 and -1, the electric energy value of the wherein corresponding virtual circuit of mete-wand is 1, the electric energy value of the corresponding virtual circuit of comparison other
For -1, the electric energy value of other virtual circuits is set to 0, forms virtual load model, N × (N-1) individual virtual load mould is obtained
Type set;In each virtual load model, same sequence is pressed in the arrangement of virtual electric energy value, and this is sequentially that each bar virtual circuit is compiled
The ascending order of count word, is defined as " line arrangement order ";
4) by all of electric energy data according to " line arrangement order " arrangement, and one by one with all virtual load model ratios
Relatively, obtain the approach degree of electric energy data and virtual load model;
5) some groups of electric energy datas that approach degree exceedes specified threshold are selected to be replicated as typical sample, after duplication
Sample and original sample form training sample set, corresponding bus Power unbalance rate is as teacher's sample set;
6) create artificial neural network, train artificial neural network, virtual load mould with training sample set, teacher's sample set
Artificial neural network after type is trained is deduced and is obtained virtual bus Power unbalance rate;
7) arrange and deduce all virtual bus Power unbalance rate obtaining, form electric power meter entirety error in dipping
Initial value collection, the error amount reconstruct that overall error in dipping initial value is concentrated, obtain the global error of corresponding electric power meter.
Preferably, step 1) detailed step include:
1.1) obtain in electric energy information acquisition system in transformer station's given voltage region to be assessed each bar circuit from when
Between point t (1) moment start to be spaced the electric energy information of the Each point in time of specified time period Δ T;
1.2) the electric energy value Δ W of each bar circuit in the time period between adjacent moment point is calculated according to formula (1)Tm(n);
ΔWTm(n)=Wt(m+1)(n)-Wt(m)(n) (1)
In formula (1), Δ WTmN () represents that nth bar circuit arrives the electric energy value between t (m+1) time period Tm, W in t (m)t(m)
N () represents the electric energy information in time point t (m) moment for the nth bar circuit, Wt(m+1)N () represents nth bar circuit in time point t (m
+ 1) electric energy information in moment, Δ WTmN () is just to represent that electric flux is sent bus, sent into bus for negative indication electric flux;
1.3) by the electric energy value Δ W of each bar circuit in the time period between adjacent moment pointTmN () constitutes electric energy to be analyzed
Amount data set WTmAs shown in formula (2), by electric energy data group WTmConstitute shown in electric energy data set W such as formula (3);
WTm=[Δ WTm(1),ΔWTm(2),...,ΔWTm(N)] (2)
W=[WT1,WT2,...,WTm]T(3).
Preferably, step 2) in be calculated shown in the function expression such as formula (4) of corresponding bus Power unbalance rate;
Step 2) in set up electric flux and specifically refer to the mapping model of bus Power unbalance rate:Electric energy data set W is obtained
Corresponding bus Power unbalance rate set ε=[εT1,εT2,...,εTm]TConstitute association, form electric flux with bus electricity not
The mapping model of balanced ratio;
In formula (4), ε represents bus Power unbalance rate, and N represents the number of circuit in transformer station's given voltage region to be assessed
Amount, WnRepresent nth bar line transmission electricity, WnIt is worth for just representing that electricity sends bus, negative indication electricity sends into bus;enRepresent
The global error of the corresponding electric quantity metering device of nth bar circuit;Wn' it is that electric quantity metering device shows the electric energy with control information
Amount, Wn' derive from electric energy data group WTm.
Preferably, step 4) detailed step include:
4.1) by all electric energy data groups according to " line arrangement order " arrangement;
4.2) choose one of which in all virtual load models, represent the model of virtual load model in array mode
Feature, in the array of the aspect of model, the electric energy value of the corresponding virtual circuit of mete-wand is 1, and comparison other corresponds to virtual circuit
Electric energy value be -1, the electric energy value of other virtual circuits is set to 0;
4.3) in all groups of electric energy datas, traversal chooses one group of electric energy data as comparing main body Wreal, electric flux
The form of data set is Wreal=[W1,W2,W3,...,WN], wherein WiRepresent i-th line road transmission electricity, i ∈ [1, N], N are
Transformer station's given voltage region to be assessed interior lines way;
4.4) virtual load model of selection and the electric energy data group chosen are compared, inclined according to distance as shown in formula (5)
Difference proximity algorithm calculates its approach degree;
In formula (5), σ (T ', W) represents approach degree, and T ' represents one group of virtual load model array, and W represents that one group have passed through
The electric energy data group of normalized, element W (n) in W=WrealN ()/max (W), n are the individual of this group electric energy data
Number, the bigger typical characteristic illustrating that this group electric energy data more has virtual load model of approach degree;
4.5) judge whether that all electric energy datas have stepped through to finish, if having stepped through finishing, redirecting and executing step
Rapid 5);Otherwise, redirect execution step 4.3).
Preferably, step 5) detailed step include:
5.1) in the comparison that one group of virtual load model participates in, five groups of maximum electric energy datas of approach degree are selected, right
This batch of electric energy data group and corresponding bus Power unbalance rate are replicated, and number of copy times is entered according to formula (6) with approach degree
Row association;
In formula (6), l rounds as number of copy times, and σ is approach degree, and a, b, c are conversion coefficient;
5.2) the electric energy data group of duplication and corresponding bus Power unbalance rate are added to initial electric flux number
According to group set and bus Power unbalance rate data acquisition system, the electric energy data that N × (N-1) individual virtual load model is replicated is then
There is 5 × N × (N-1) group, electric energy data group set becomes training sample set, bus Power unbalance rate data acquisition system becomes religion
Teacher's sample set.
Preferably, step 6) detailed step include:
6.1) initialize BP artificial neural network, give weights and the threshold values of each layer of network by randomly assigne;
6.2) concentrate from training sample and take one group of electric energy data, take corresponding bus electricity uneven from teacher's sample set
Weighing apparatus rate sends into BP artificial neural network as input signal;
6.3) electric energy data positive calculation in BP artificial neural network obtains result, is compared with teacher's sample,
Calculate neutral net output layer output signal error according to formula (7);
δl=-(Tl-Yl)Yl(1-Yl) (7)
In formula (7), δlFor the node error of output layer, TlFor teacher's sample data of output layer, YlNode for output layer
Output signal;
6.4) the node error of each hidden layer is calculated according to formula (8);
In formula (8),For the node error of n-th neuron in kth layer hidden layer,For n-th in kth layer hidden layer
The output signal node of neuron,For the node error of a-th neuron in l layer, l layer is hidden layer or output layer, p
For the neuron population of l layer,Weights for each neuron in a-th connection l layer in kth layer hidden layer;
6.5) adopt newton-Gauss innovatory algorithm BP ANN, update hidden layer and output according to formula (9)
The weights of interlayer, are calculated the node error of each hidden layer by formula (11), update the threshold values of output layer according to formula (12);
In formula (9), ω (k+1) is new weights, and ω (k) is former weights, and Δ ω (k) is right value update amount, HkAnd gkFor centre
Matrix of variables, JT(ω) it is Jacobian matrix, e (ω) is the vector set of single neuron node error composition, single neuron
The set value of node error composition is output layer output signal error Yl, μ is controlling elements, and I is unit matrix;Wherein refined can
Compare matrix JT(ω) shown in function expression such as formula (10);
In formula (10), J represents Jacobian matrix JT(ω), eiRepresent the node error of i-th neuron in this hidden layer, i
∈ [1, M], M are the quantity of this layer of hidden layer neuron, ωijRepresent that i-th neuron of this hidden layer connects next layer of output layer
The weights of j-th neuron, j ∈ [1, O], O represent the neuronal quantity of next layer of output layer;The node of neuron in hidden layer
Error is represented with formula (11) with the weights of corresponding output layer neuron:
In formula (11), eiFor the node error of i-th neuron in this hidden layer, YiFor i-th neuron in this hidden layer
Output signal node δjFor the node error of j-th neuron in next layer of output layer, ωijFor i-th in this hidden layer even
Connect the weights of j-th neuron of next layer of output layer;
In formula (12),For m-th neuron of output layer the t+1 moment threshold values,For m-th god of output layer
Through unit t threshold values, β be correction factor,Node error for m-th neuron in l layer hidden layer;
6.6) judge whether all of training sample set and teacher's sample set is all trained finishes, if not yet trained
Finish, then redirect execution step 6.2), otherwise redirect execution step 6.7);
6.7) error and the study number of times of BP artificial neural network are checked, if the error of artificial neural network is more than or equal to
Setting value and when learning number of times and being not reaching to pre-determined number, redirects execution step 6.2);Otherwise, terminate to train and store currently
The network parameter of BP artificial neural network;
6.8) N × (N-1) individual virtual load model is sent into the BP artificial neural network training to be deduced, obtain void
Intend the individual virtual bus Power unbalance rate of the corresponding N of load model × (N-1);By individual for N × (N-1) virtual bus Power unbalance
Rate is sorted out, and sorting out principle is that corresponding N × (N-1) individual virtual load model has N number of mete-wand, each mete-wand pair
Should there is N-1 comparison other, corresponding for the model of identical mete-wand bus Power unbalance rate is included into same set,
Form N number of set, N-1 element of each set, the element in each set is according to " line arrangement order " arrangement, mete-wand
Corresponding position is empty, using the set obtaining as the overall error in dipping initial value collection of the electric power meter of corresponding mete-wand.
Preferably, step 7) detailed step include:
7.1) concentrate in N number of electric power meter entirety error in dipping initial value and all add zero error, fill up mete-wand
The corresponding evolution initial value collection being empty position, forming as shown in formula (13);
In formula (13), ε represents evolution initial value collection, εi,jRepresent on the basis of i-th set of metering device, jth set metering dress
The Initial value put, εi,xRepresent on the basis of i-th set of metering device, the Initial value of all metering devices (includes first set meter
Amount device) vector, i ∈ [1, N], j ∈ [1, N-1], N represent metering device quantity;
7.2) traversal chooses one of evolution initial value collection, takes the median of each element numerical value in its set;
7.3) by this evolution initial value integrate according to median as translation distance carry out translation obtain shift evolution initial value collection, according to
" line arrangement order " obtains the data that displacement evolution initial value concentration is associated with each metering device, as overall error in dipping;
7.4) judge whether that all evolution initial value collection have stepped through selection and finish, if not yet traversal selection finishes, jump
Turn execution step 7.2);Otherwise, it is determined that all evolution initial value collection are all translated, and often set electric quantity metering device all obtains N
Shown in individual overall error in dipping formula (14);
In formula (14), ε ' represents error reconstruct collection, median (εi,x) represent εi,xMedian, εi′,jRepresent at the beginning of error
Value, εi,jReconstruction value after translation, translational movement is median (εi,x), i ∈ [1, N], j ∈ [1, N], N represents metering device
Quantity;
7.5) make according to the meansigma methodss that function expression shown in formula (15) asks for every set N number of overall error in dipping of metering device
Finally deduce the overall error in dipping obtaining for this set metering device;
εn=[avg (ε 'x,1),avg(ε′x,2),…avg(ε′x,N)] (15)
In formula (14), εnRepresent that n-th set of metering device finally deduces the overall error in dipping obtaining, avg (ε 'x,i) represent
Set ε 'x,iThe meansigma methodss of middle all elements.
On the other hand, the present invention also provides a kind of electric power meter error detecting system based on big data deduction, bag
Include:
Electric energy data data capture unit, for obtaining the many of each bar circuit in transformer station's given voltage region to be assessed
Group electric energy data;
Bus Power unbalance rate computing unit, for being calculated corresponding bus electricity according to every group of electric energy data
Unbalance factor, sets up electric flux according to multigroup electric energy data and corresponding bus Power unbalance rate uneven with bus electricity
The mapping model of weighing apparatus rate;
Virtual load model sets up unit, and for creating N bar virtual circuit, N is transformer station's given voltage region to be assessed
Interior lines way, takes any two virtual circuits to constitute circuit pair, by one of circuit to being set to current line pair, sets wherein
Article one, the metering charge value of virtual circuit be mete-wand, another virtual circuit be comparison other;Void by current line pair
The electric energy value intending circuit is set to 1 and -1, and the electric energy value of the wherein corresponding virtual circuit of mete-wand is 1, comparison other
The electric energy value of corresponding virtual circuit is -1, and the electric energy value of other virtual circuits is set to 0, forms virtual load model, there are
To N × (N-1) individual virtual load model set;In each virtual load model, same sequence is pressed in the arrangement of virtual electric energy value,
This is sequentially the ascending order of each bar virtual circuit numerical digit, is defined as " line arrangement order ";
Approach degree computing unit, for by all of electric energy data according to " line arrangement order " arrangement, and one by one with
All virtual load models compare, and obtain the approach degree of electric energy data and virtual load model;
Sample set signal generating unit, for selecting approach degree to exceed some groups of electric energy datas of specified threshold as typical sample
Originally replicated, the sample after duplication forms training sample set with original sample, and corresponding bus Power unbalance rate is as religion
Teacher's sample set;
Artificial neural network training unit, for creating artificial neural network, with training sample set, the training of teacher's sample set
Artificial neural network, the artificial neural network after virtual load model is trained is deduced and is obtained virtual bus Power unbalance rate;
Global error computing unit, deduces, for arranging, all virtual bus Power unbalance rate obtaining, and forms electric energy
Metering device entirety error in dipping initial value collection, the initial value collection reconstruct that overall error in dipping initial value is concentrated, obtain corresponding electric energy meter
The global error of amount device.
Preferably, described electric energy data data capture unit includes:
Electric energy information acquisition module, for obtaining transformer station to be assessed given voltage in electric energy information acquisition system
In region, each bar circuit starts to be spaced the electric energy information of the Each point in time of specified time period Δ T from time point t (1) moment;
Electric energy magnitude calculation acquisition module, for calculating each bar circuit in the time period between adjacent moment point according to formula (1)
Electric energy value Δ WTm(n);
ΔWTm(n)=Wt(m+1)(n)-Wt(m)(n) (1)
In formula (1), Δ WTmN () represents that nth bar circuit arrives the electric energy value between t (m+1) time period Tm, W in t (m)t(m)
N () represents the electric energy information in time point t (m) moment for the nth bar circuit, Wt(m+1)N () represents nth bar circuit in time point t (m
+ 1) electric energy information in moment, Δ WTmN () is just to represent that electric flux is sent bus, sent into bus for negative indication electric flux;
Electric energy data group builds module, for by the electric energy value Δ of each bar circuit in the time period between adjacent moment point
WTmN () constitutes electric energy data group W to be analyzedTmAs shown in formula (2), by electric energy data group WTmConstitute electric energy data collection
Close shown in W such as formula (3);
WTm=[Δ WTm(1),ΔWTm(2),...,ΔWTm(N)] (2)
W=[WT1,WT2,...,WTm]T(3).
Preferably, described bus Power unbalance rate computing unit is calculated the letter of corresponding bus Power unbalance rate
Shown in number expression formula such as formula (4);Described bus Power unbalance rate computing unit sets up electric flux and bus Power unbalance rate
Mapping model specifically refer to:Electric energy data set W is obtained corresponding bus Power unbalance rate set ε=[εT1,
εT2,...,εTm]TConstitute association, form the mapping model of electric flux and bus Power unbalance rate;
In formula (4), ε represents bus Power unbalance rate, and N represents the number of circuit in transformer station's given voltage region to be assessed
Amount, WnRepresent nth bar line transmission electricity, WnIt is worth for just representing that electricity sends bus, negative indication electricity sends into bus;enRepresent
The global error of the corresponding electric quantity metering device of nth bar circuit;Wn' it is that electric quantity metering device shows the electric energy with control information
Amount, Wn' derive from electric energy data group WTm.
The present invention is had the advantage that based on the electric power meter error detection method tool that big data is deduced:
1) present invention by on-line monitoring obtain electric energy information, by electric energy data calculate bus electric quantity balancing rate and by
Substantial amounts of electric energy data, bus Power unbalance rate set up mapping model, thus assessing the entirety metering of electric power meter
Error condition is it is not necessary to carry out on-site proving or detection to metering device it is not required that detection that equipment is had a power failure, easy to use, high
Effect and safety;Electric energy data with collection is different, and error judgment result also will change accordingly, thus reflecting electric energy in real time
The change of metering device entirety error in dipping, with the timely development trend finding metering device hidden failure or deterioration, Neng Gouyou
Effect ensures the health operation of metering device.
2) present invention adopts newton Gauss innovatory algorithm to train artificial neural network, and newton Gauss innovatory algorithm has
Taking into account global property, training convergence block, avoiding training process to be easily absorbed in the problem jumping out local to a certain extent so that deducing knot
Fruit precision is higher, and the metering device global error of reaction is more accurate.
3) electric energy data and virtual load model are compared by the present invention, using proximity algorithm and fuzzy membership letter
Number realizes the fuzzy strengthening of typical characteristic electric energy data, increases frequency of training under this feature for the artificial neural network, thus
Make artificial neural network have higher diagnostic accuracy and generalization near deduction domain, make deduction result more accurate.
4) the result addition zero error obtaining deduced by artificial neural network by the present invention, forms evolution initial value collection, utilizes
Metering device this feature of synthetic error Normal Distribution, evolution initial value collection is reconstructed, discloses bus Power unbalance rate
With the mapping relations of metering device global error it is achieved that assessment to metering device global error.
The present invention is based on big data based on the electric power meter error detecting system that big data is deduced for the present invention and deduces
The completely corresponding device of electric power meter error detection method, therefore equally also have what the present invention was deduced based on big data
The aforementioned advantages of electric power meter error detection method, therefore will not be described here.
Brief description
Fig. 1 is the basic procedure schematic diagram of present invention method.
Fig. 2 be embodiment of the present invention step 4) schematic flow sheet.
Fig. 3 is the deduction of BP artificial neural network and training principle schematic in the present invention.
Fig. 4 is the present embodiment step 6) newton Gauss innovatory algorithm training schematic flow sheet.
Fig. 5 is the present embodiment step 7) in error reconstruct schematic flow sheet.
Specific embodiment
As shown in figure 1, the step of electric power meter error detection method that the present embodiment is deduced based on big data includes:
1) obtain multigroup electric energy data of each bar circuit in transformer station's given voltage region to be assessed;
2) corresponding bus Power unbalance rate is calculated according to every group of electric energy data, according to multigroup electric energy data
And corresponding bus Power unbalance rate sets up the mapping model of electric flux and bus Power unbalance rate;
3) create N bar virtual circuit, N be transformer station's given voltage region to be assessed interior lines way, take any two virtual
Railway superstructures circuit pair, by one of circuit to being set to current line pair, sets the metering electricity of a wherein virtual circuit
Being worth for mete-wand, another virtual circuit is comparison other;The electric energy value of the virtual circuit of current line pair is set respectively
For 1 and -1, the electric energy value of the wherein corresponding virtual circuit of mete-wand is 1, the electric energy value of the corresponding virtual circuit of comparison other
For -1, the electric energy value of other virtual circuits is set to 0, forms virtual load model, N × (N-1) individual virtual load mould is obtained
Type set;In each virtual load model, same sequence is pressed in the arrangement of virtual electric energy value, and this is sequentially that each bar virtual circuit is compiled
The ascending order of count word, is defined as " line arrangement order ";
4) by all of electric energy data according to " line arrangement order " arrangement, and one by one with all virtual load model ratios
Relatively, obtain the approach degree of electric energy data and virtual load model;
5) some groups of electric energy datas that approach degree exceedes specified threshold are selected to be replicated as typical sample, after duplication
Sample and original sample form training sample set, corresponding bus Power unbalance rate is as teacher's sample set;
6) create artificial neural network, train artificial neural network, virtual load mould with training sample set, teacher's sample set
Artificial neural network after type is trained is deduced and is obtained virtual bus Power unbalance rate;
7) arrange and deduce all virtual bus Power unbalance rate obtaining, form electric power meter entirety error in dipping
Initial value collection, the error amount reconstruct that overall error in dipping initial value is concentrated, obtain the global error of corresponding electric power meter.
In the present embodiment, step 1) detailed step include:
1.1) obtain in electric energy information acquisition system in transformer station's given voltage region to be assessed each bar circuit from when
Between point t (1) moment start to be spaced the electric energy information of the Each point in time of specified time period Δ T;
1.2) the electric energy value Δ W of each bar circuit in the time period between adjacent moment point is calculated according to formula (1)Tm(n);
ΔWTm(n)=Wt(m+1)(n)-Wt(m)(n) (1)
In formula (1), Δ WTmN () represents that nth bar circuit arrives the electric energy value between t (m+1) time period Tm, W in t (m)t(m)
N () represents the electric energy information in time point t (m) moment for the nth bar circuit, Wt(m+1)N () represents nth bar circuit in time point t (m
+ 1) electric energy information in moment, Δ WTmN () is just to represent that electric flux is sent bus, sent into bus for negative indication electric flux;
1.3) by the electric energy value Δ W of each bar circuit in the time period between adjacent moment pointTmN () constitutes electric energy to be analyzed
Amount data set WTmAs shown in formula (2), by electric energy data group WTmConstitute shown in electric energy data set W such as formula (3);
WTm=[Δ WTm(1),ΔWTm(2),...,ΔWTm(N)] (2)
W=[WT1,WT2,...,WTm]T(3).
The electric energy information W in time point t (1) moment for each bar circuit is obtained in electric energy information acquisition systemt(1)
(n);Obtain the electric energy information W in time point t (2) momentt(2)N (), t (1) arrives the electricity of line related between t (2) time period T1
Energy value is Δ WT1(n)=Wt(2)(n)-Wt(1)(n).Further, extend to relation line from t (m) to t (m+1) time period Tm
The electric energy value on road is Δ WTm(n)=Wt(m+1)(n)-Wt(m)(n), wherein T1=T2=...=Tm=Δ T.ΔWTmN () composition is treated
Electric energy data group W of analysisTm=[Δ WTm(1),ΔWTm(2),...,ΔWTm(n)] as shown in formula (2), electric energy data group
WTmConstitute electric energy data set W=[WT1,WT2,...,WTm]TAs shown in formula (3), thus obtain transformer station to be assessed and specify electricity
Multigroup electric energy data of each bar circuit in intermediate pressure section.
In the present embodiment, step 2) in be calculated the function expression such as formula (4) of corresponding bus Power unbalance rate
Shown;Step 2) in set up electric flux and specifically refer to the mapping model of bus Power unbalance rate:By electric energy data set W
Obtain corresponding bus Power unbalance rate set ε=[εT1,εT2,...,εTm]TConstitute association, form electric flux and bus electricity
The mapping model of amount unbalance factor;
In formula (4), ε represents bus Power unbalance rate, and N represents the number of circuit in transformer station's given voltage region to be assessed
Amount, WnRepresent nth bar line transmission electricity, WnIt is worth for just representing that electricity sends bus, negative indication electricity sends into bus (WnDuring > 0
Represent that primary side electric flux sends bus, WnRepresent during < 0 that primary side electric flux sends into bus);enRepresent that nth bar circuit corresponds to
Electric quantity metering device global error;Wn' it is that electric quantity metering device shows the electric flux with control information, Wn' derive from electric energy
Amount data set WTm.For transformer station's identical electric pressure region, same time window enters, sends electricity and the bus electricity of bus
The association of amount unbalance factor can be represented with function expression as shown in formula (4), Wn' it is the electric flux that metering device shows, band
There is the control information of metering device;Derive from electric energy data group WTm;By electric energy data set W=[WT1,WT2,...,
WTm]TObtain corresponding bus Power unbalance rate set ε=[ε furtherT1,εT2,...,εTm]T, and constitute association, form WTm
Mapping model with ε.
In the present embodiment, first confirm that the circuit ingredient in transformer station to be assessed a certain electric pressure region, i.e. this electricity
The N bar circuit that in intermediate pressure section, bus is connected, is then obtained in electric energy information data base by electric energy information acquisition platform
Take electric energy information W in the time periods such as N bar circuitn', built by every group of electric energy data and corresponding bus Power unbalance rate
Vertical metering electric flux Wn' with the mapping model of bus Power unbalance rate ε.The present embodiment takes the electric energy data of nearest 30 days simultaneously
Take the period to be 3 hours, that is, acquire 240 groups of data, every group of data all includes the electric energy data of all N bar circuits, 240 groups
Electric energy data correspondence is calculated 240 bus Power unbalance rates ε.
As shown in Fig. 2 the present embodiment step 4) detailed step include:
4.1) by all electric energy data groups according to " line arrangement order " arrangement;
4.2) choose one of which in all virtual load models, represent the model of virtual load model in array mode
Feature, in the array of the aspect of model, the electric energy value of the corresponding virtual circuit of mete-wand is 1, and comparison other corresponds to virtual circuit
Electric energy value be -1, the electric energy value of other virtual circuits is set to 0, such as one group model feature example be Μ=[1, -1,
0,...,0];
4.3) in all groups of electric energy datas, traversal chooses one group of electric energy data as comparing main body Wreal, electric flux
The form of data set is Wreal=[W1,W2,W3,...,WN], wherein WiRepresent i-th line road transmission electricity, i ∈ [1, N], N are
Transformer station's given voltage region to be assessed interior lines way;
4.4) virtual load model of selection and the electric energy data group chosen are compared, inclined according to distance as shown in formula (5)
Difference proximity algorithm calculates its approach degree;
In formula (5), σ (T ', W) represents approach degree, and T ' represents one group of virtual load model array, and W represents that one group have passed through
The electric energy data group of normalized, element W (n) in W=WrealN ()/max (W), n are the individual of this group electric energy data
Number, the bigger typical characteristic illustrating that this group electric energy data more has virtual load model of approach degree;
4.5) judge whether that all electric energy datas have stepped through to finish, if having stepped through finishing, redirecting and executing step
Rapid 5);Otherwise, redirect execution step 4.3).
In the present embodiment, any two lines road is taken to constitute circuit pair, by one of circuit to being set to current line pair, if
The metering charge value of a fixed wherein circuit is mete-wand, and another is comparison other;Electric energy value by current line pair
It is set to 1 and -1, wherein mete-wand corresponding electric energy value is 1, and comparison other is -1, and the electric energy value of All other routes sets
For 0, form virtual load model;Similar virtual load model composition model set, contains N × (N- in this model set
1) individual virtual load model, N is number of, lines;In each model, same sequence is pressed in the arrangement of the virtual electric energy value of each bar circuit,
This is sequentially defined as " line arrangement order ";By step 1) in multigroup electric energy data of obtaining according to " line arrangement order " row
Row, are compared with virtual load model, calculate its approach degree according to range deviation proximity algorithm as shown in formula (5).
In the present embodiment, step 5) detailed step include:
5.1) in the comparison that one group of virtual load model participates in, five groups of maximum electric energy datas of approach degree are selected, right
This batch of electric energy data group and corresponding bus Power unbalance rate are replicated, and number of copy times is entered according to formula (6) with approach degree
Row association;
In formula (6), l rounds as number of copy times, and σ is approach degree, and a, b, c are conversion coefficient;In the present embodiment, by approach degree
Some groups of larger electric energy datas are replicated as typical sample, and sample replicates follows more former of approach degree more higher duplication
Then, it is distributed the transform of fuzzy membership functions using Cauchy shown in formula (6), the approach degree after processing is associated with number of copy times,
Realize the fuzzy strengthening of typical sample:
5.2) the electric energy data group of duplication and corresponding bus Power unbalance rate are added to initial electric flux number
According to group set and bus Power unbalance rate data acquisition system, the electric energy data that N × (N-1) individual virtual load model is replicated is then
There is 5 × N × (N-1) group, electric energy data group set becomes training sample set, bus Power unbalance rate data acquisition system becomes religion
Teacher's sample set.
In the present embodiment, example takes 5 groups of maximum data of approach degree;This batch of electric energy data group is replicated, number of copy times
It is associated according to formula (6) with approach degree:Virtual load model has N × (N-1) individual, and the electric energy data being replicated then has 5 × N
× (N-1) group, 5 × N × (N-1) organizes the possibility that there is repeat replication in electric energy data.
Using characteristic strengthening type electric energy data group as training sample, corresponding bus Power unbalance rate group conduct
Teacher's sample, described training sample and teacher's sample is inputted to default BP artificial neural network, using newton Gauss
Improved algorithm trains weights and the threshold values of each layer of described BP artificial neural network, obtains the electric energy based on BP artificial neural network
Amount and the mapping model of bus Power unbalance rate.
As shown in figure 3, in the present embodiment the training of BP artificial neural network and deduce when, from virtual load model extract allusion quotation
Type feature, electric energy data is recognized with this typical characteristic, obtains identification result.According to identification result by each group electric energy data
It is divided into typical sample and common sample two class, wherein typical sample is some groups of electric fluxs that feature is pressed close to virtual load model
Data, it is typical sample that this example takes 5 groups of maximum electric energy datas of approach degree, and remaining is common sample.Replicate typical sample, abide by
Follow the more principles of approach degree more higher duplication, be distributed the transform of fuzzy membership functions by approach degree and number of copy times using Cauchy
Association.Sample and the original sample replicating forms training sample set, corresponding bus Power unbalance rate as teacher's sample set,
Obtain training sample and teacher's sample set, training sample and teacher's sample set are sent into BP artificial neural network and is trained.Will
The BP artificial neural network that virtual load model input trains is deduced, and obtains deducing result.In the present embodiment, specifically
Refer to by N × (N-1) individual virtual load model send in the BP artificial neural network completing to train and deduce, obtain N × (N-1) individual
Virtual bus Power unbalance rate, is converted into electric power meter entirety error in dipping initial value.In the present embodiment, N × (N-1) is individual
Virtual bus Power unbalance rate conversion electric power meter entirety error in dipping initial value is processed, the specific implementation method of conversion
For:The individual virtual bus Power unbalance rate of N × (N-1) has corresponded to N × (N-1) individual condition;This condition is N bar line for evaluation object
Road, every circuit all occurs in that N-1 time as mete-wand and comparison other;Selecting a circuit is mete-wand, will meet
N-1 virtual bus Power unbalance rate of this condition is arranged, and carries out according to " line arrangement order ", mete-wand corresponds to
Position be to be empty, obtain electric power meter entirety error in dipping initial value collection, similar initial value collection has N number of.
As shown in figure 4, the present embodiment step 6) detailed step include:
6.1) initialize BP artificial neural network, give weights and the threshold values of each layer of network by randomly assigne;
6.2) concentrate from training sample and take one group of electric energy data, take corresponding bus electricity uneven from teacher's sample set
Weighing apparatus rate sends into BP artificial neural network as input signal;
6.3) electric energy data positive calculation in BP artificial neural network obtains result, is compared with teacher's sample,
Calculate neutral net output layer output signal error according to formula (7);
δl=-(Tl-Yl)Yl(1-Yl) (7)
In formula (7), δlFor the node error of output layer, TlFor teacher's sample data of output layer, YlNode for output layer
Output signal;
6.4) the node error of each hidden layer is calculated according to formula (8);
In formula (8),For the node error of n-th neuron in kth layer hidden layer,For n-th in kth layer hidden layer
The output signal node of neuron,For the node error of a-th neuron in l layer, l layer is hidden layer or output layer, p
For the neuron population of l layer,Weights for each neuron in a-th connection l layer in kth layer hidden layer;
6.5) adopt newton-Gauss innovatory algorithm BP ANN, update hidden layer and output according to formula (9)
The weights of interlayer, are calculated the node error of each hidden layer by formula (11), update the threshold values of output layer according to formula (12);
In formula (9), ω (k+1) is new weights, and ω (k) is former weights, and Δ ω (k) is right value update amount, HkAnd gkFor centre
Matrix of variables, JT(ω) it is Jacobian matrix, e (ω) is the vector set of single neuron node error composition, single neuron
The set value of node error composition is output layer output signal error Yl, μ is controlling elements, and I is unit matrix;Wherein refined can
Compare matrix JT(ω) shown in function expression such as formula (10);
In formula (10), J represents Jacobian matrix JT(ω), eiRepresent the node error of i-th neuron in this hidden layer, i
∈ [1, M], M are the quantity of this layer of hidden layer neuron, ωijRepresent that i-th neuron of this hidden layer connects next layer of output layer
The weights of j-th neuron, j ∈ [1, O], O represent the neuronal quantity of next layer of output layer;The node of neuron in hidden layer
Error is represented with formula (11) with the weights of corresponding output layer neuron:
In formula (11), eiFor the node error of i-th neuron in this hidden layer, YiFor i-th neuron in this hidden layer
Output signal node δjFor the node error of j-th neuron in next layer of output layer, ωijFor i-th in this hidden layer even
Connect the weights of j-th neuron of next layer of output layer;
In formula (12),For m-th neuron of output layer the t+1 moment threshold values,For m-th god of output layer
Through unit t threshold values, β be correction factor,Node error for m-th neuron in l layer hidden layer;
6.6) judge whether all of training sample set and teacher's sample set is all trained finishes, if not yet trained
Finish, then redirect execution step 6.2), otherwise redirect execution step 6.7);
6.7) error and the study number of times of BP artificial neural network are checked, if the error of artificial neural network is more than or equal to
Setting value and when learning number of times and being not reaching to pre-determined number, redirects execution step 6.2);Otherwise, terminate to train and store currently
The network parameter of BP artificial neural network;
6.8) N × (N-1) individual virtual load model is sent into the BP artificial neural network training to be deduced, obtain void
Intend the individual virtual bus Power unbalance rate of the corresponding N of load model × (N-1);By individual for N × (N-1) virtual bus Power unbalance
Rate is sorted out, and sorting out principle is that corresponding N × (N-1) individual virtual load model has N number of mete-wand, each mete-wand pair
Should there is N-1 comparison other, corresponding for the model of identical mete-wand bus Power unbalance rate is included into same set,
Form N number of set, N-1 element of each set, the element in each set is according to " line arrangement order " arrangement, mete-wand
Corresponding position is empty, using the set obtaining as the overall error in dipping initial value collection of the electric power meter of corresponding mete-wand.
As shown in figure 5, the present embodiment step 7) detailed step include:
7.1) concentrate in N number of electric power meter entirety error in dipping initial value and all add zero error, fill up mete-wand
The corresponding evolution initial value collection being empty position, forming as shown in formula (13);
In formula (13), ε represents evolution initial value collection, εi,jRepresent on the basis of i-th set of metering device, jth set metering dress
The Initial value put, εi,xRepresent on the basis of i-th set of metering device, the Initial value of all metering devices (includes first set meter
Amount device) vector, i ∈ [1, N], j ∈ [1, N-1], N represent metering device quantity;
7.2) traversal chooses one of evolution initial value collection, takes the median of each element numerical value in its set;
7.3) by this evolution initial value integrate according to median as translation distance carry out translation obtain shift evolution initial value collection, according to
" line arrangement order " obtains the data that displacement evolution initial value concentration is associated with each metering device, as overall error in dipping;
7.4) judge whether that all evolution initial value collection have stepped through selection and finish, if not yet traversal selection finishes, jump
Turn execution step 7.2);Otherwise, it is determined that all evolution initial value collection are all translated, and often set electric quantity metering device all obtains N
Shown in individual overall error in dipping formula (14);
In formula (14), ε ' represents error reconstruct collection, median (εi,x) represent εi,xMedian, ε 'i,jRepresent at the beginning of error
Value, εi,jReconstruction value after translation, translational movement is median (εi,x), i ∈ [1, N], j ∈ [1, N], N represents metering device
Quantity;
7.5) make according to the meansigma methodss that function expression shown in formula (15) asks for every set N number of overall error in dipping of metering device
Finally deduce the overall error in dipping obtaining for this set metering device;
εn=[avg (ε 'x,1),avg(ε′x,2),…avg(ε′x,N)] (15)
In formula (14), εnRepresent that n-th set of metering device finally deduces the overall error in dipping obtaining, avg (ε 'x,i) represent
Set ε 'x,iThe meansigma methodss of middle all elements.
Because the present embodiment Chinese style (4) considers particular case, taking the 1st article of circuit and the 2nd article of circuit as a example, if only two
Circuit sends into bus and send bus in transmission load, respectively electric flux, then the difference of the error of this two sets of metering devices is
For Power unbalance rate, as shown in formula (16).
In formula (16), ε1,2Represent the Power unbalance rate between the 1st article of circuit and the 2nd article of circuit, W1Represent the 1st bar of line
Electricity, W are transmitted in road2Represent the 2nd article of line transmission electricity, e1Represent the entirety of the 1st article of corresponding electric quantity metering device of circuit by mistake
Difference, e2Represent the global error of the 2nd article of corresponding electric quantity metering device of circuit.Formula (4-1) is by the bus of single input output state
Power unbalance rate is directly connected with metering device error.Artificially a benchmark is set, even e2For 0, then e1=
ε1,2, further genralrlization to the metering device error obtaining bus All other routes, formation error collection is it is possible to obtain all to be evaluated
The global error of estimator device.But the method can be in the face of following problem:1. it is difficult to occur only two during transformer station's actual motion
The situation of bar circuit on-load is it is impossible to directly set up the corresponding model of formula (2) using electric quantity data collection.2. metering device error is subject to
The impact of the random factors such as environment, load is continually changing, and the boundary condition of Power unbalance rate model is unstable.3. it is manually set
A set of metering device error is 0, and practical situation is certainly different, so that whole error collection is offset toward a direction, and side-play amount is not
Fixed.
For solve problem 1., the present embodiment step 3) in adopt with the following method:Create N bar virtual circuit, N is change to be assessed
Power station given voltage region interior lines way, takes any two virtual circuits to constitute circuit pair, one of circuit is worked as to being set to
Front circuit pair, the metering charge value setting a wherein virtual circuit is as mete-wand, another virtual circuit as comparison other;
The electric energy value of the virtual circuit of current line pair is set to 1 and -1, the wherein electric energy of the corresponding virtual circuit of mete-wand
Amount T ' is worth for 1, and the electric energy value of the corresponding virtual circuit of comparison other is -1, and the electric energy value of other virtual circuits is set to 0, is formed
Virtual load model, is obtained N × (N-1) individual virtual load model set;Virtual electric energy value in each virtual load model
Arrangement press same sequence, this is sequentially ascending order of each bar virtual circuit numerical digit, is defined as " line arrangement order ".
For solve problem 2., the present embodiment step 6) in adopt with the following method:Create BP artificial neural network, using nerve
The compatibility of network and error correction solve boundary condition shakiness problem;Using step 5) in the training sample set that obtains and teacher's sample
This collection trains neutral net by newton Gauss innovatory algorithm.
For solve problem 3., the present embodiment step 7) in deduce all virtual bus Power unbalance that obtains by arranging
Rate, forms electric power meter entirety error in dipping initial value collection, and the error amount reconstruct that overall error in dipping initial value is concentrated obtains
The global error of corresponding electric power meter.Each electric power meter entirety error in dipping initial value collection and a zero error knot
Close, form evolution initial value collection, what zero error filled up mete-wand is empty position.Take the median median of evolution initial value collection
[ε], the actual global error of the metering device corresponding to this median is most likely to be close to zero.This is because the synthesis of metering device
Error Normal Distribution ε~N (μ, σ), that is, error be distributed centered on expecting μ, on the other hand, the same voltage regime of transformer station
The configuration consistency of metering device, that is, often set metering device has consistent expectation, and the corresponding actual error of this expectation is in close proximity to
Zero, can be considered zero error.The median of evolution initial value collection is closest with expectation on probability, even if accidental offset, its skew
Amount is also less, it is taken as that the corresponding actual error of median of evolution initial value collection is zero.With median as distance, according to formula
(17) evolution initial value collection is translated, obtain the global error of metering device;
In formula (17), ε represents evolution initial value collection, and ε ' represents error reconstruct collection, and each element in error reconstruct collection ε ' is exactly
The entirety of corresponding metering device deduces error.Due to there being N number of evolution initial value collection, thus every suit metering device all can obtain N number of
Overall deduction error, N number of result equalization is obtained final overall error in dipping and deduces result.
The electric power meter error detection method that the present embodiment is deduced based on big data is particular by computer program
Come to realize, realize the electric power meter error detecting system bag deduced based on big data by this computer program
Include:
Electric energy data data capture unit, for obtaining the many of each bar circuit in transformer station's given voltage region to be assessed
Group electric energy data;
Bus Power unbalance rate computing unit, for being calculated corresponding bus electricity according to every group of electric energy data
Unbalance factor, sets up electric flux according to multigroup electric energy data and corresponding bus Power unbalance rate uneven with bus electricity
The mapping model of weighing apparatus rate;
Virtual load model sets up unit, and for creating N bar virtual circuit, N is transformer station's given voltage region to be assessed
Interior lines way, takes any two virtual circuits to constitute circuit pair, by one of circuit to being set to current line pair, sets wherein
Article one, the metering charge value of virtual circuit be mete-wand, another virtual circuit be comparison other;Void by current line pair
The electric energy value intending circuit is set to 1 and -1, and the electric energy value of the wherein corresponding virtual circuit of mete-wand is 1, comparison other
The electric energy value of corresponding virtual circuit is -1, and the electric energy value of other virtual circuits is set to 0, forms virtual load model, there are
To N × (N-1) individual virtual load model set;In each virtual load model, same sequence is pressed in the arrangement of virtual electric energy value,
This is sequentially the ascending order of each bar virtual circuit numerical digit, is defined as " line arrangement order ";
Approach degree computing unit, for by all of electric energy data according to " line arrangement order " arrangement, and one by one with
All virtual load models compare, and obtain the approach degree of electric energy data and virtual load model;
Sample set signal generating unit, for selecting approach degree to exceed some groups of electric energy datas of specified threshold as typical sample
Originally replicated, the sample after duplication forms training sample set with original sample, and corresponding bus Power unbalance rate is as religion
Teacher's sample set;
Artificial neural network training unit, for creating artificial neural network, with training sample set, the training of teacher's sample set
Artificial neural network, the artificial neural network after virtual load model is trained is deduced and is obtained virtual bus Power unbalance rate;
Global error computing unit, deduces, for arranging, all virtual bus Power unbalance rate obtaining, and forms electric energy
Metering device entirety error in dipping initial value collection, the initial value collection reconstruct that overall error in dipping initial value is concentrated, obtain corresponding electric energy meter
The global error of amount device.
Electric energy data data capture unit described in the present embodiment includes:
Electric energy information acquisition module, for obtaining transformer station to be assessed given voltage in electric energy information acquisition system
In region, each bar circuit starts to be spaced the electric energy information of the Each point in time of specified time period Δ T from time point t (1) moment;
Electric energy magnitude calculation acquisition module, for calculating each bar circuit in the time period between adjacent moment point according to formula (1)
Electric energy value Δ WTm(n);
ΔWTm(n)=Wt(m+1)(n)-Wt(m)(n) (1)
In formula (1), Δ WTmN () represents that nth bar circuit arrives the electric energy value between t (m+1) time period Tm, W in t (m)t(m)
N () represents the electric energy information in time point t (m) moment for the nth bar circuit, Wt(m+1)N () represents nth bar circuit in time point t (m
+ 1) electric energy information in moment, Δ WTmN () is just to represent that electric flux is sent bus, sent into bus for negative indication electric flux;
Electric energy data group builds module, for by the electric energy value Δ of each bar circuit in the time period between adjacent moment point
WTmN () constitutes electric energy data group W to be analyzedTmAs shown in formula (2), by electric energy data group WTmConstitute electric energy data collection
Close shown in W such as formula (3);
WTm=[Δ WTm(1),ΔWTm(2),...,ΔWTm(N)] (2)
W=[WT1,WT2,...,WTm]T(3).
In the present embodiment, described bus Power unbalance rate computing unit is calculated corresponding bus Power unbalance rate
Function expression such as formula (4) shown in;It is uneven with bus electricity that described bus Power unbalance rate computing unit sets up electric flux
The mapping model of weighing apparatus rate specifically refers to:Electric energy data set W is obtained corresponding bus Power unbalance rate set ε=[εT1,
εT2,...,εTm]TConstitute association, form the mapping model of electric flux and bus Power unbalance rate;
In formula (4), ε represents bus Power unbalance rate, and N represents the number of circuit in transformer station's given voltage region to be assessed
Amount, WnRepresent nth bar line transmission electricity, WnIt is worth for just representing that electricity sends bus, negative indication electricity sends into bus;enRepresent
The global error of the corresponding electric quantity metering device of nth bar circuit;Wn' it is that electric quantity metering device shows the electric energy with control information
Amount, Wn' derive from electric energy data group WTm.In sum, the electric power meter error that the present embodiment is deduced based on big data
Detection method and system are capable of the overall error in dipping of simultaneous real-time monitoring electric power meter, have and are not required to field test, reality
Existing method is simple, detection is real-time, the advantage of highly effective and safe.
The above is only the preferred embodiment of the present invention, and protection scope of the present invention is not limited merely to above-mentioned enforcement
Example, all technical schemes belonging under thinking of the present invention belong to protection scope of the present invention.It should be pointed out that for the art
Those of ordinary skill for, some improvements and modifications without departing from the principles of the present invention, these improvements and modifications
Should be regarded as protection scope of the present invention.
Claims (10)
1. a kind of electric power meter error detection method based on big data deduction is it is characterised in that step includes:
1) obtain multigroup electric energy data of each bar circuit in transformer station's given voltage region to be assessed;
2) corresponding bus Power unbalance rate is calculated according to every group of electric energy data, according to multigroup electric energy data and
Corresponding bus Power unbalance rate sets up the mapping model of electric flux and bus Power unbalance rate;
3) create N bar virtual circuit, N is transformer station's given voltage region to be assessed interior lines way, takes any two virtual circuits
Constitute circuit pair, by one of circuit to being set to current line pair, set the metering charge value of a wherein virtual circuit as
Mete-wand, another virtual circuit are comparison other;The electric energy value of the virtual circuit of current line pair is set to 1
With -1, the electric energy value of the wherein corresponding virtual circuit of mete-wand is 1, the electric energy value of the corresponding virtual circuit of comparison other is -
1, the electric energy value of other virtual circuits is set to 0, forms virtual load model, N × (N-1) individual virtual load model collection is obtained
Close;In each virtual load model, same sequence is pressed in the arrangement of virtual electric energy value, and this is sequentially each bar virtual circuit numbering number
The ascending order of word, is defined as " line arrangement order ";
4) by all of electric energy data according to " line arrangement order " arrangement, and compare with all virtual load models one by one,
Obtain the approach degree of electric energy data and virtual load model;
5) some groups of electric energy datas that approach degree exceedes specified threshold are selected to be replicated as typical sample, the sample after duplication
This forms training sample set with original sample, and corresponding bus Power unbalance rate is as teacher's sample set;
6) create artificial neural network, train artificial neural network, virtual load model warp with training sample set, teacher's sample set
Artificial neural network after training is deduced and is obtained virtual bus Power unbalance rate;
7) arrange and deduce all virtual bus Power unbalance rate obtaining, form electric power meter entirety error in dipping initial value
Collection, the error amount reconstruct that overall error in dipping initial value is concentrated, obtain the global error of corresponding electric power meter.
2. the electric power meter error detection method based on big data deduction according to claim 1 it is characterised in that
Step 1) detailed step include:
1.1) obtain in transformer station's given voltage region to be assessed each bar circuit in electric energy information acquisition system from time point t
(1) moment starts to be spaced the electric energy information of the Each point in time of specified time period Δ T;
1.2) the electric energy value Δ W of each bar circuit in the time period between adjacent moment point is calculated according to formula (1)Tm(n);
ΔWTm(n)=Wt(m+1)(n)-Wt(m)(n) (1)
In formula (1), Δ WTmN () represents that nth bar circuit arrives the electric energy value between t (m+1) time period Tm, W in t (m)t(m)(n) table
Show the electric energy information in time point t (m) moment for the nth bar circuit, Wt(m+1)N () represents nth bar circuit when time point t (m+1)
The electric energy information carved, Δ WTmN () is just to represent that electric flux is sent bus, sent into bus for negative indication electric flux;
1.3) by the electric energy value Δ W of each bar circuit in the time period between adjacent moment pointTmN () constitutes electric flux number to be analyzed
According to group WTmAs shown in formula (2), by electric energy data group WTmConstitute shown in electric energy data set W such as formula (3);
WTm=[Δ WTm(1),ΔWTm(2),...,ΔWTm(N)] (2)
W=[WT1,WT2,...,WTm]T(3).
3. the electric power meter error detection method based on big data deduction according to claim 1 it is characterised in that
Step 2) in be calculated shown in the function expression such as formula (4) of corresponding bus Power unbalance rate;Step 2) middle foundation electricity
Energy is specifically referred to the mapping model of bus Power unbalance rate:Electric energy data set W is obtained corresponding bus electricity
Unbalance factor set ε=[εT1,εT2,...,εTm]TConstitute association, form the mapping mould of electric flux and bus Power unbalance rate
Type;
In formula (4), ε represents bus Power unbalance rate, and N represents the quantity of circuit in transformer station's given voltage region to be assessed, Wn
Represent nth bar line transmission electricity, WnIt is worth for just representing that electricity sends bus, negative indication electricity sends into bus;enRepresent nth bar
The global error of the corresponding electric quantity metering device of circuit;W′nIt is that electric quantity metering device shows the electric flux with control information, W 'nCome
Come from electric energy data group WTm.
4. the electric power meter error detection method based on big data deduction according to claim 1 it is characterised in that
Step 4) detailed step include:
4.1) by all electric energy data groups according to " line arrangement order " arrangement;
4.2) choose one of which in all virtual load models, represent that in array mode the model of virtual load model is special
Levy, in the array of the aspect of model, the electric energy value of the corresponding virtual circuit of mete-wand is 1, the corresponding virtual circuit of comparison other
Electric energy value is -1, and the electric energy value of other virtual circuits is set to 0;
4.3) in all groups of electric energy datas, traversal chooses one group of electric energy data as comparing main body Wreal, electric energy data
The form of group is Wreal=[W1,W2,W3,...,WN], wherein WiRepresent i-th line road transmission electricity, i ∈ [1, N], N are to be evaluated
Estimate transformer station's given voltage region interior lines way;
4.4) compare the virtual load model of selection and the electric energy data group chosen, paste according to range deviation as shown in formula (5)
Recency algorithm calculates its approach degree;
In formula (5), σ (T ', W) represents approach degree, and T ' represents one group of virtual load model array, and W represents that have passed through a normalizing
Change the electric energy data group processing, element W (n) in W=WrealN ()/max (W), n are the number of this group electric energy data, patch
The bigger typical characteristic illustrating that this group electric energy data more has virtual load model of recency;
4.5) judging whether that all electric energy datas have stepped through to finish, if having stepped through finished, redirecting execution step 5);
Otherwise, redirect execution step 4.3).
5. the electric power meter error detection method based on big data deduction according to claim 1 it is characterised in that
Step 5) detailed step include:
5.1) in the comparison that one group of virtual load model participates in, five groups of maximum electric energy datas of approach degree are selected, to this batch
Electric energy data group and corresponding bus Power unbalance rate are replicated, and number of copy times is closed according to formula (6) with approach degree
Connection;
In formula (6), l rounds as number of copy times, and σ is approach degree, and a, b, c are conversion coefficient;
5.2) the electric energy data group of duplication and corresponding bus Power unbalance rate are added to initial electric energy data group
Set and bus Power unbalance rate data acquisition system, the electric energy data that N × (N-1) individual virtual load model is replicated then has 5
× N × (N-1) group, electric energy data group set becomes training sample set, and bus Power unbalance rate data acquisition system becomes teacher
Sample set.
6. the electric power meter error detection method based on big data deduction according to claim 1 it is characterised in that
Step 6) detailed step include:
6.1) initialize BP artificial neural network, give weights and the threshold values of each layer of network by randomly assigne;
6.2) concentrate from training sample and take one group of electric energy data, take corresponding bus Power unbalance rate from teacher's sample set
Send into BP artificial neural network as input signal;
6.3) electric energy data positive calculation in BP artificial neural network obtains result, is compared with teacher's sample, according to
Formula (7) calculates neutral net output layer output signal error;
δl=-(Tl-Yl)Yl(1-Yl) (7)
In formula (7), δlFor the node error of output layer, TlFor teacher's sample data of output layer, YlNode output for output layer
Signal;
6.4) the node error of each hidden layer is calculated according to formula (8);
In formula (8),For the node error of n-th neuron in kth layer hidden layer,For n-th nerve in kth layer hidden layer
The output signal node of unit,For the node error of a-th neuron in l layer, l layer is hidden layer or output layer, and p is the
The neuron population of l layer,Weights for each neuron in a-th connection l layer in kth layer hidden layer;
6.5) adopt newton-Gauss innovatory algorithm BP ANN, update hidden layer and output interlayer according to formula (9)
Weights, by formula (11) calculate each hidden layer node error, according to formula (12) update output layer threshold values;
In formula (9), ω (k+1) is new weights, and ω (k) is former weights, and Δ ω (k) is right value update amount, HkAnd gkFor intermediate variable
Matrix, JT(ω) it is Jacobian matrix, e (ω) is the vector set of single neuron node error composition, single neuron node
The set value of error composition is output layer output signal error Yl, μ is controlling elements, and I is unit matrix;Wherein Jacobean matrix
Battle array JT(ω) shown in function expression such as formula (10);
In formula (10), J represents Jacobian matrix JT(ω), eiRepresent the node error of i-th neuron in this hidden layer, i ∈
[1, M], M is the quantity of this layer of hidden layer neuron, ωijRepresent that i-th neuron of this hidden layer connects next layer of output layer jth
The weights of individual neuron, j ∈ [1, O], O represent the neuronal quantity of next layer of output layer;In hidden layer, the node of neuron misses
Difference is represented with formula (11) with the weights of corresponding output layer neuron:
In formula (11), eiFor the node error of i-th neuron in this hidden layer, YiSection for i-th neuron in this hidden layer
Point output signal δjFor the node error of j-th neuron in next layer of output layer, ωijUnder connecting for i-th in this hidden layer
The weights of j-th neuron of one layer of output layer;
In formula (12),For m-th neuron of output layer the t+1 moment threshold values,For m-th neuron of output layer
In the threshold values of t, β is correction factor,Node error for m-th neuron in l layer hidden layer;
6.6) judge whether all of training sample set and teacher's sample set is all trained finishes, if not yet training finishes,
Then redirect execution step 6.2), otherwise redirect execution step 6.7);
6.7) error and the study number of times of BP artificial neural network are checked, if the error of artificial neural network is more than or equal to set
When being worth and learning number of times and be not reaching to pre-determined number, redirect execution step 6.2);Otherwise, terminate to train and store current BP people
The network parameter of artificial neural networks;
6.8) N × (N-1) individual virtual load model is sent into the BP artificial neural network training to be deduced, obtain virtual negative
The individual virtual bus Power unbalance rate of the corresponding N of lotus model × (N-1);Individual for N × (N-1) virtual bus Power unbalance rate is entered
Row is sorted out, and sorting out principle is that corresponding N × (N-1) individual virtual load model has N number of mete-wand, and each mete-wand is to should have
There is N-1 comparison other, corresponding for the model of identical mete-wand bus Power unbalance rate is included into same set, formed
N number of set, N-1 element of each set, the element in each set corresponds to according to " line arrangement order " arrangement, mete-wand
Position be empty, using the set obtaining as the overall error in dipping initial value collection of the electric power meter of corresponding mete-wand.
7. the electric power meter error detection method based on big data deduction according to claim 1 it is characterised in that
Step 7) detailed step include:
7.1) concentrate in N number of electric power meter entirety error in dipping initial value and all add zero error, fill up mete-wand and correspond to
The evolution initial value collection being empty position, forming as shown in formula (13);
In formula (13), ε represents evolution initial value collection, εi,jRepresent that, on the basis of i-th set of metering device, jth covers metering device
Initial value, εi,xRepresent on the basis of i-th set of metering device, the Initial value of all metering devices (includes first set metering dress
Put) vector, i ∈ [1, N], j ∈ [1, N-1], N represent metering device quantity;
7.2) traversal chooses one of evolution initial value collection, takes the median of each element numerical value in its set;
7.3) by this evolution initial value integrate according to median as translation distance carry out translation obtain shift evolution initial value collection, according to " line
Road puts in order " obtain the data that displacement evolution initial value concentration is associated with each metering device, as overall error in dipping;
7.4) judge whether that all evolution initial value collection have stepped through selection and finish, if not yet traversal selection finishes, redirect and hold
Row step 7.2);Otherwise, it is determined that all evolution initial value collection are all translated, and often set electric quantity metering device all obtain N number of whole
Shown in body error in dipping formula (14);
In formula (14), ε ' represents error reconstruct collection, median (εi,x) represent εi,xMedian, ε 'i,jRepresent Initial value, εi,j
Reconstruction value after translation, translational movement is median (εi,x), i ∈ [1, N], j ∈ [1, N], N represent metering device quantity;
7.5) meansigma methodss of every set N number of overall error in dipping of metering device are asked for as this according to function expression shown in formula (15)
The overall error in dipping obtaining finally deduced by set metering device;
εn=[avg (ε 'x,1),avg(ε′x,2),…avg(ε′x,N)] (15)
In formula (14), εnRepresent that n-th set of metering device finally deduces the overall error in dipping obtaining, avg (ε 'x,i) represent set
ε′x,iThe meansigma methodss of middle all elements.
8. a kind of electric power meter error detecting system based on big data deduction is it is characterised in that include:
Electric energy data data capture unit, for obtaining multigroup electricity of each bar circuit in transformer station's given voltage region to be assessed
Energy datum;
Bus Power unbalance rate computing unit is uneven for being calculated corresponding bus electricity according to every group of electric energy data
Weighing apparatus rate, sets up electric flux and bus Power unbalance rate according to multigroup electric energy data and corresponding bus Power unbalance rate
Mapping model;
Virtual load model sets up unit, and for creating N bar virtual circuit, N is transformer station's given voltage region to be assessed interior lines
Way, takes any two virtual circuits to constitute circuit pair, by one of circuit to being set to current line pair, sets wherein one
The metering charge value of virtual circuit is mete-wand, another virtual circuit is comparison other;Dummy line by current line pair
The electric energy value on road is set to 1 and -1, and the electric energy value of the wherein corresponding virtual circuit of mete-wand is 1, and comparison other corresponds to
The electric energy value of virtual circuit is -1, and the electric energy value of other virtual circuits is set to 0, forms virtual load model, be obtained N ×
(N-1) individual virtual load model set;In each virtual load model, same sequence is pressed in the arrangement of virtual electric energy value, this order
For the ascending order of each bar virtual circuit numerical digit, it is defined as " line arrangement order ";
Approach degree computing unit, for by all of electric energy data according to " line arrangement order " arrangement, and one by one with all
Virtual load model compares, and obtains the approach degree of electric energy data and virtual load model;
Sample set signal generating unit, some groups of electric energy datas for selecting approach degree to exceed specified threshold enter as typical sample
Row replicates, and the sample after duplication forms training sample set with original sample, and corresponding bus Power unbalance rate is as teacher's sample
This collection;
Artificial neural network training unit, for creating artificial neural network, is trained artificial with training sample set, teacher's sample set
Neutral net, the artificial neural network after virtual load model is trained is deduced and is obtained virtual bus Power unbalance rate;
Global error computing unit, deduces, for arranging, all virtual bus Power unbalance rate obtaining, and forms electric energy metrical
Device entirety error in dipping initial value collection, the initial value collection reconstruct that overall error in dipping initial value is concentrated, obtain corresponding electric energy metrical dress
The global error put.
9. the electric power meter error detecting system based on big data deduction according to claim 8 it is characterised in that
Described electric energy data data capture unit includes:
Electric energy information acquisition module, for obtaining transformer station's given voltage region to be assessed in electric energy information acquisition system
Interior each bar circuit starts to be spaced the electric energy information of the Each point in time of specified time period Δ T from time point t (1) moment;
Electric energy magnitude calculation acquisition module, for calculating the electricity of each bar circuit in the time period between adjacent moment point according to formula (1)
Energy value Δ WTm(n);
ΔWTm(n)=Wt(m+1)(n)-Wt(m)(n) (1)
In formula (1), Δ WTmN () represents that nth bar circuit arrives the electric energy value between t (m+1) time period Tm, W in t (m)t(m)(n) table
Show the electric energy information in time point t (m) moment for the nth bar circuit, Wt(m+1)N () represents nth bar circuit when time point t (m+1)
The electric energy information carved, Δ WTmN () is just to represent that electric flux is sent bus, sent into bus for negative indication electric flux;
Electric energy data group builds module, for by the electric energy value Δ W of each bar circuit in the time period between adjacent moment pointTm
N () constitutes electric energy data group W to be analyzedTmAs shown in formula (2), by electric energy data group WTmConstitute electric energy data set W
As shown in formula (3);
WTm=[Δ WTm(1),ΔWTm(2),...,ΔWTm(N)] (2)
W=[WT1,WT2,...,WTm]T(3).
10. the electric power meter error detecting system based on big data deduction according to claim 8, its feature exists
In described bus Power unbalance rate computing unit is calculated the function expression such as formula of corresponding bus Power unbalance rate
(4) shown in;Described bus Power unbalance rate computing unit sets up the mapping model tool of electric flux and bus Power unbalance rate
Body refers to:Electric energy data set W is obtained corresponding bus Power unbalance rate set ε=[εT1,εT2,...,εTm]TConstitute
Association, forms the mapping model of electric flux and bus Power unbalance rate;
In formula (4), ε represents bus Power unbalance rate, and N represents the quantity of circuit in transformer station's given voltage region to be assessed, Wn
Represent nth bar line transmission electricity, WnIt is worth for just representing that electricity sends bus, negative indication electricity sends into bus;enRepresent nth bar
The global error of the corresponding electric quantity metering device of circuit;Wn' it is that electric quantity metering device shows the electric flux with control information, W 'nCome
Come from electric energy data group WTm.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611092500.8A CN106443566B (en) | 2016-12-01 | 2016-12-01 | The electric energy metering device error detection method and system deduced based on big data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611092500.8A CN106443566B (en) | 2016-12-01 | 2016-12-01 | The electric energy metering device error detection method and system deduced based on big data |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106443566A true CN106443566A (en) | 2017-02-22 |
CN106443566B CN106443566B (en) | 2019-03-12 |
Family
ID=58223657
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611092500.8A Active CN106443566B (en) | 2016-12-01 | 2016-12-01 | The electric energy metering device error detection method and system deduced based on big data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106443566B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107589391A (en) * | 2017-07-27 | 2018-01-16 | 武汉尤瑞卡节能科技有限公司 | A kind of methods, devices and systems for detecting electric power meter global error |
CN111222364A (en) * | 2018-11-23 | 2020-06-02 | 中移物联网有限公司 | Meter reading method and device |
CN111273212A (en) * | 2020-02-24 | 2020-06-12 | 国网湖南省电力有限公司 | Data-driven electric quantity sensor error online evaluation closed-loop improvement method, system and medium |
CN112858980A (en) * | 2021-01-13 | 2021-05-28 | 国家电网有限公司华东分部 | Gateway metering abnormity diagnosis method combining sampling and big data |
CN113468729A (en) * | 2021-06-15 | 2021-10-01 | 国网湖南省电力有限公司 | Method and system for measuring and calculating operation errors of electric power station metering device without calibration mode |
CN115840185A (en) * | 2023-02-20 | 2023-03-24 | 威胜集团有限公司 | Voltage transformer error online monitoring and analyzing method, medium and terminal |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11899516B1 (en) | 2023-07-13 | 2024-02-13 | T-Mobile Usa, Inc. | Creation of a digital twin for auto-discovery of hierarchy in power monitoring |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6671635B1 (en) * | 2001-02-23 | 2003-12-30 | Power Measurement Ltd. | Systems for improved monitoring accuracy of intelligent electronic devices |
CN1782730A (en) * | 2005-09-06 | 2006-06-07 | 淄博计保互感器研究所 | Integrate detecting and measuring method for metering property of high voltage power meter |
CN102749541A (en) * | 2012-07-09 | 2012-10-24 | 云南电力试验研究院(集团)有限公司电力研究院 | Energy conservation-based real-time bus and metering equipment state on-line monitoring system |
JP2014180089A (en) * | 2013-03-13 | 2014-09-25 | Ricoh Co Ltd | Power feed device |
CN104218570A (en) * | 2014-08-21 | 2014-12-17 | 国家电网公司 | Method and system for online evaluating overall measuring errors of electric energy measuring device |
-
2016
- 2016-12-01 CN CN201611092500.8A patent/CN106443566B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6671635B1 (en) * | 2001-02-23 | 2003-12-30 | Power Measurement Ltd. | Systems for improved monitoring accuracy of intelligent electronic devices |
CN1782730A (en) * | 2005-09-06 | 2006-06-07 | 淄博计保互感器研究所 | Integrate detecting and measuring method for metering property of high voltage power meter |
CN102749541A (en) * | 2012-07-09 | 2012-10-24 | 云南电力试验研究院(集团)有限公司电力研究院 | Energy conservation-based real-time bus and metering equipment state on-line monitoring system |
JP2014180089A (en) * | 2013-03-13 | 2014-09-25 | Ricoh Co Ltd | Power feed device |
CN104218570A (en) * | 2014-08-21 | 2014-12-17 | 国家电网公司 | Method and system for online evaluating overall measuring errors of electric energy measuring device |
Non-Patent Citations (1)
Title |
---|
陈丽红等: "浅探变电站母线电量不平衡率对电能计量的影响", 《甘肃科技》 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107589391A (en) * | 2017-07-27 | 2018-01-16 | 武汉尤瑞卡节能科技有限公司 | A kind of methods, devices and systems for detecting electric power meter global error |
CN107589391B (en) * | 2017-07-27 | 2020-09-15 | 武汉国测数据技术有限公司 | Method, device and system for detecting integral error of electric energy metering device |
CN111222364A (en) * | 2018-11-23 | 2020-06-02 | 中移物联网有限公司 | Meter reading method and device |
CN111222364B (en) * | 2018-11-23 | 2023-08-18 | 中移物联网有限公司 | Meter reading method and meter reading device |
CN111273212A (en) * | 2020-02-24 | 2020-06-12 | 国网湖南省电力有限公司 | Data-driven electric quantity sensor error online evaluation closed-loop improvement method, system and medium |
CN111273212B (en) * | 2020-02-24 | 2022-03-11 | 国网湖南省电力有限公司 | Data-driven electric quantity sensor error online evaluation closed-loop improvement method, system and medium |
CN112858980A (en) * | 2021-01-13 | 2021-05-28 | 国家电网有限公司华东分部 | Gateway metering abnormity diagnosis method combining sampling and big data |
CN113468729A (en) * | 2021-06-15 | 2021-10-01 | 国网湖南省电力有限公司 | Method and system for measuring and calculating operation errors of electric power station metering device without calibration mode |
CN115840185A (en) * | 2023-02-20 | 2023-03-24 | 威胜集团有限公司 | Voltage transformer error online monitoring and analyzing method, medium and terminal |
Also Published As
Publication number | Publication date |
---|---|
CN106443566B (en) | 2019-03-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106443566A (en) | Big data deduction based electric energy metering device error detection method and system | |
CN106406229B (en) | A kind of diagnosing faults of numerical control machine method | |
CN102496069B (en) | Cable multimode safe operation evaluation method based on fuzzy analytic hierarchy process (FAHP) | |
CN109857835A (en) | A kind of adaptive network security knowledge assessment method based on cognitive diagnosis theory | |
CN102622510A (en) | System and method for quantitative management of software defects | |
CN105590175A (en) | Skilled talent evaluation method based on factor analysis and BP neural networks | |
CN107576920A (en) | A kind of satellite power supply health degree computational methods based on fuzzy multi-attribute evaluation | |
CN104700153A (en) | PH (potential of hydrogen) value predicting method of BP (back propagation) neutral network based on simulated annealing optimization | |
CN107505837A (en) | A kind of semi-supervised neural network model and the soft-measuring modeling method based on the model | |
CN105938609A (en) | Power grid operation assessment method for realizing multilayer indicator system | |
CN102520342B (en) | Analog circuit test node selecting method based on dynamic feedback neural network modeling | |
CN110417011A (en) | A kind of online dynamic secure estimation method based on mutual information Yu iteration random forest | |
CN104155574A (en) | Power distribution network fault classification method based on adaptive neuro-fuzzy inference system | |
CN102467684A (en) | System and method for evaluating comprehensive performance based on improved radar chart | |
CN108921230A (en) | Method for diagnosing faults based on class mean value core pivot element analysis and BP neural network | |
CN110333414A (en) | The multi-level state evaluating method of power transformer | |
CN105320987B (en) | A kind of satellite telemetering data intelligent interpretation method based on BP neural network | |
CN104218570A (en) | Method and system for online evaluating overall measuring errors of electric energy measuring device | |
CN106777997A (en) | A kind of heath state of spacecraft appraisal procedure | |
CN111179576B (en) | Power utilization information acquisition fault diagnosis method and system with inductive learning function | |
CN105842607A (en) | Test point quantitative selection method and device in testing design | |
CN104077493A (en) | Method for constructing state evaluation index system of electric relaying protection system | |
CN107832789A (en) | Characteristic weighing k nearest neighbor method for diagnosing faults based on the conversion of average influence Value Data | |
CN113470475A (en) | Real-operation learning assessment method and system based on scene simulation and Internet of things | |
CN105741184A (en) | Transformer state evaluation method and apparatus |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |