CN106980905A - Distribution network reliability Forecasting Methodology and system - Google Patents
Distribution network reliability Forecasting Methodology and system Download PDFInfo
- Publication number
- CN106980905A CN106980905A CN201710154078.2A CN201710154078A CN106980905A CN 106980905 A CN106980905 A CN 106980905A CN 201710154078 A CN201710154078 A CN 201710154078A CN 106980905 A CN106980905 A CN 106980905A
- Authority
- CN
- China
- Prior art keywords
- reliability
- index
- factor
- power supply
- distribution network
- 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
- 238000000034 method Methods 0.000 title claims abstract description 34
- 238000011156 evaluation Methods 0.000 claims abstract description 43
- 238000013528 artificial neural network Methods 0.000 claims abstract description 18
- 239000002245 particle Substances 0.000 claims description 51
- 239000010410 layer Substances 0.000 claims description 32
- 238000012360 testing method Methods 0.000 claims description 26
- 238000012549 training Methods 0.000 claims description 25
- 238000004422 calculation algorithm Methods 0.000 claims description 16
- 238000003646 Spearman's rank correlation coefficient Methods 0.000 claims description 15
- 238000012423 maintenance Methods 0.000 claims description 15
- 230000001537 neural effect Effects 0.000 claims description 13
- 230000007935 neutral effect Effects 0.000 claims description 10
- 230000006835 compression Effects 0.000 claims description 7
- 238000007906 compression Methods 0.000 claims description 7
- 210000002569 neuron Anatomy 0.000 claims description 7
- 239000011229 interlayer Substances 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000010219 correlation analysis Methods 0.000 description 18
- 238000004458 analytical method Methods 0.000 description 14
- 230000000875 corresponding effect Effects 0.000 description 11
- 238000005070 sampling Methods 0.000 description 10
- 238000005457 optimization Methods 0.000 description 9
- 238000007689 inspection Methods 0.000 description 8
- 238000009413 insulation Methods 0.000 description 8
- 238000004364 calculation method Methods 0.000 description 6
- 230000002596 correlated effect Effects 0.000 description 6
- 238000012545 processing Methods 0.000 description 5
- 238000010276 construction Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 4
- 244000144992 flock Species 0.000 description 4
- 230000009467 reduction Effects 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000012352 Spearman correlation analysis Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 2
- 230000000052 comparative effect Effects 0.000 description 2
- 238000000205 computational method Methods 0.000 description 2
- 239000004020 conductor Substances 0.000 description 2
- 230000008094 contradictory effect Effects 0.000 description 2
- 238000013523 data management Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 235000013305 food Nutrition 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 210000005036 nerve Anatomy 0.000 description 2
- 238000002203 pretreatment Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000008439 repair process Effects 0.000 description 2
- 238000010206 sensitivity analysis Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 238000001521 two-tailed test Methods 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 description 1
- 239000010931 gold Substances 0.000 description 1
- 229910052737 gold Inorganic materials 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000004451 qualitative analysis Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000013024 troubleshooting Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Supply And Distribution Of Alternating Current (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to a kind of distribution network reliability Forecasting Methodology, comprise the following steps：Choose the original index factor of a variety of influence power supply reliabilities；The original index factor is standardized, obtain standardized index factor, the coefficient correlation between each standardized index factor and default reliability evaluation index is calculated, filtered out according to the coefficient correlation from the index factor influences more obvious core index factor to power supply reliability；Radial basis function neural network is built according to the core index factor, obtain the optimal center vector, optimal base fat vector and optimal output weight of the hidden layer of the radial basis function neural network, according to the optimal center vector, optimal base fat vector and optimal output Weight Acquisition power supply reliability forecast model, distribution network reliability is predicted according to the power supply reliability forecast model.
Description
Technical field
The present invention relates to technical field of electric power, more particularly to a kind of distribution network reliability Forecasting Methodology and system.
Background technology
The life of the reliable degree that power distribution network is powered and national product have it is close associate, distribution network reliability one
Determine to embody integration capability of the power supply enterprise in terms of the construction, transformation and operation maintenance of power network in degree.Further carry
The power supply reliability of power distribution network is risen, not only the need for electricity of power consumer can be met to greatest extent, while being also beneficial to promote
The further perfect and development of power grid construction.
Traditional distribution network reliability focus is gradually converted by passive management in recent years in the management of power distribution network
Actively to prevent.Existing related technology proposes to pass through artificial neural network, SVMs (Support Vector
Machine, SVM) etc. method the factor for influenceing urban distribution network reliability is analyzed, and then distribution network reliability is done
Go out prediction.
Although prediction algorithm gradually innovates optimization in these technologies, the accuracy predicted the outcome is poor.
The content of the invention
Based on this, it is necessary to for the accuracy that predicts the outcome it is poor the problem of there is provided a kind of distribution network reliability
Forecasting Methodology and system.
A kind of distribution network reliability Forecasting Methodology, comprises the following steps：
Joined according to the grid structure of power distribution network, technical equipment parameter, equipment quality parameter, fault compression and operation maintenance
Number chooses the original index factor of a variety of influence power supply reliabilities；
The original index factor is standardized, standardized index factor is obtained, each standardization is calculated and refers to
Coefficient correlation between mark factor and default reliability evaluation index, is sieved according to the coefficient correlation from the index factor
Select influences more obvious core index factor to power supply reliability；
Radial basis function neural network is built according to the core index factor, the radial basis function neural network is obtained
Hidden layer optimal center vector, optimal base fat vector and optimal output weight, according to the optimal center vector, optimal base
Fat vector and optimal output Weight Acquisition power supply reliability forecast model, according to the power supply reliability forecast model to power distribution network
Power supply reliability is predicted.
A kind of distribution network reliability forecasting system, including：
Index factor choose module, for the grid structure according to power distribution network, technical equipment parameter, equipment quality parameter,
Fault compression and operation maintenance parameter choose the original index factor of a variety of influence power supply reliabilities；
Initial data screening module, for being standardized to the original index factor, obtains standardized index
Factor, calculates the coefficient correlation between each standardized index factor and default reliability evaluation index, according to the correlation
Coefficient is filtered out from the index factor influences more obvious core index factor to power supply reliability；
Forecast model training module, for building radial basis function neural network according to the core index factor, is obtained
Optimal center vector, optimal base fat vector and the optimal output weight of the hidden layer of the radial basis function neural network, according to
The optimal center vector, optimal base fat vector and optimal output Weight Acquisition power supply reliability forecast model, are supplied according to described
Electric reliability prediction model is predicted to distribution network reliability.
Above-mentioned distribution network reliability Forecasting Methodology and system, choose the index factor of a variety of influence power supply reliabilities；
Using correlation analysis, initial data is standardized, the index and reliability of a variety of influence power supply reliabilities is calculated
Coefficient correlation between evaluation index, qualitative analysis goes out influences more obvious index to power supply reliability；Build radial direction base letter
Number neutral net simultaneously optimizes the center vector of hidden layer, sound stage width vector sum output weight, completes to power supply reliability forecast model
Training.During prediction, many power supply reliability influence factors have been taken into full account, it is adaptable to the feelings of multi input variable
Condition.Prescreening processing is carried out to characteristic index using correlation analysis is theoretical, radial basis function neural network instruction is effectively improved
Practice precision of prediction, stability and the Generalization Capability of model.In addition, optimizing radial basis function neural network, its convergence is overcome
Speed is relatively slow, the drawbacks of easily fall into local best points, and arrange parameter is few, effectively reduces the training time of forecast model, improves
Precision of prediction.By above-mentioned Forecasting Methodology and system, in the case of sample data is accurate, power supply reliability can be carried out more
Plus comprehensively analysis and predict, for guiding power supply enterprise formulate reliability Promotion Strategy provide it is scientific and effective with reference to according to
According to.
Brief description of the drawings
Fig. 1 is the flow chart of distribution network reliability Forecasting Methodology in an embodiment；
Fig. 2 is influence power supply reliability index classification figure in an embodiment；
Fig. 3 is influence power supply reliability RS-1 leading indicator comparison diagrams in an embodiment；
Fig. 4 is radial basis function neural network training flow chart in an embodiment；
Fig. 5 is distribution network reliability forecasting system structure chart in an embodiment.
Embodiment
Technical scheme is illustrated below in conjunction with the accompanying drawings.
As shown in figure 1, the present invention provides a kind of distribution network reliability Forecasting Methodology, it may include following steps：
Step S1：According to the grid structure of power distribution network, the technical equipment parameter of power distribution network, Distribution Network Equipment mass parameter,
The distribution network failure factor and power distribution network operation maintenance parameter choose the original index factor of a variety of influence power supply reliabilities.
The power supply reliability of power distribution network and power network itself deliverability, operation and maintenance level and the close phase of extraneous factor
Close.It is comprehensive for ensure to analyze power supply reliability, the technical equipment parameter of grid structure, power distribution network from power distribution network, set
Five aspects such as standby mass parameter, fault compression and operation maintenance parameter choose the index of 30 influence power supply reliabilities, make
For power supply reliability forecast model input variable.The grid structure of wherein power distribution network is related to power supply capacity and electrical stability,
That is ability to bear and capability to resist destruction of the power distribution network to load variations；Technical equipment parameter and equipment quality parameter, fault compression are closed
Fault rate and the troubleshooting time of equipment are tied to, evaluation is quick after fail-safe ability and failure with network operation occur
Examination, positioning, isolating power；Operation and maintenance level relates generally to livewire work level and power distribution network fast power restoration ability.
Power supply reliability index classification is as shown in Figure 2.The science and reasonability of the grid structure of power network are influence power supplies
The key factor of reliability, rational grid structure can realize effectively turning for electric load in any element failure
Move, and reduce the power off time of user as far as possible；Distribution network technology equipment mainly includes the insulating and cable journey of circuit
Degree, distribution network automated construction and maintenance levels；Distribution Network Equipment include distribution transformer, switch cubicle circuit, shaft tower, gold utensil,
The plurality of devices such as arrester, the service life of equipment and the number of times broken down have larger shadow for power supply reliability
Ring；Distribution network failure is mainly influenceed by natural cause, external force factor and operation maintenance construction；Power distribution network operation and maintenance level
Relate generally to livewire work level and power distribution network fast power restoration ability etc..The specific targets of five aspects are as shown in table 1.
Table 1 influences power supply reliability index
Step S2：The original index factor is standardized, standardized index factor is obtained, each mark is calculated
Coefficient correlation between standardization index factor and default reliability evaluation index, according to the coefficient correlation from the index because
Filtered out in element influences more obvious core index factor to power supply reliability.
Due to influenceing the factor of distribution network reliability varied, it is therefore desirable to many factors are screened, chosen
Go out the factor of core.Core index factor can be screened by correlation analysis (Correlation Analysis, CA).Correlation
Analysis is to weigh and evaluate the strong or weak relation between variable by calculating coefficient correlation.Power supply reliability evaluation index includes supplying
A variety of evaluation indexes such as electric reliability, average power off time of user.In one embodiment, can choose power supply reliability RS-1 is
Evaluation index, using SPSS, (Statistical Product and Service Solutions, statistical product is solved with service
Certainly scheme) instrument carries out correlation analysis calculating.Power supply reliability RS-1 is interior during counting, to the effective power-on time of user
The ratio of total hourage and statistics period hours number, its formula is
In formula, RS-1 is reliability evaluation index, T1For average power off time of user, T2For the time of a measurement period
Length.
In one embodiment, z-score standardizations can be carried out to initial data.By being carried out to initial data
Standardization, can eliminate the different influence of original variable dimension.Average and standard deviation based on initial data, by following
Formula standardizes original index factor using z-score, obtains standardized index factor.Below equation can specifically be used：
In formula, D' is standardized index factor, and D is original index factor,For the average of original index factor, δ is original
The standard deviation of beginning index factor.
The coefficient correlation can be Pearson correlation coefficient or Spearman rank correlation coefficient.Pearson correlation coefficient is used
The power of linear dependence between two variables are measured, it has certain application conditions：1) it is paired to necessarily assume that data
What ground was obtained from normal distribution；2) data must be equidistant at least in logical categories.Its mathematical definition is：
Wherein：N is sample size；xiAnd yiFor the variate-value of two variables；R is Pearson correlation coefficient.Pearson came is related
Coefficient value scope is [- 1,1], and r=1 is expressed as perfect positive correlation, and r=-1 is perfect negative correlation, and r=0 represents that line is not present
Property dependency relation.| r | it is better closer to 1 linear correlation degree, | r | it is weaker closer to 0 linear correlation degree.It should illustrate
, sample here refers to the sampling result to above-mentioned standard index factor, i.e. to each standardized index factor,
Several sample values, x are taken respectivelyiCorresponding i-th of the sample of currently processed standardized index factor is represented,Represent pre-treatment
Standardized index factor sample average, yiThe reliability evaluation index of i-th of sample is represented,Each reliability is represented to comment
The average of valency index.
Because coefficient correlation is that overall part sample is carried out calculating what is obtained, therefore it can not directly conclude samples sources
Totality meet significant dependency relation, it is necessary to by assuming that examine mode inferred, that is, utilize the mode of hypothesis testing
Mathematically illustrate the correctness of conclusion.Its basic step is：
(1) sample is built according to each standardized index factor, the sample and reliability evaluation index proposed former false
If H0With alternative hypothesis H1；Wherein, null hypothesis thinks the sample and reliability evaluation index without clear-cut correlation, alternative vacation
If thinking that the sample and reliability evaluation index have clear-cut correlation；
(2) according to sample size selection check statistic, such as t statistics, Z statistics；
(3) observation and corresponding Probability p of test statistics are calculated；
(4) determine significance decision-making α and make decisions.If the Probability p of test statistics is less than significance
α, then refuse null hypothesis, it was confirmed that alternative hypothesis, it is believed that two totality have significant dependency relation, can calculate each in sample
Pearson correlation coefficient between standardized index factor and default reliability evaluation index；, whereas if test statistics
Probability p be more than or equal to level of significance α, then can not refuse null hypothesis, it is believed that dependency relation is not present in two totality, can count
Calculate the Spearman rank correlation coefficient between each standardized index factor and default reliability evaluation index in sample.
Wherein, the statistic of step (2) depends on sample size size and selected.
Z statistics are applied to the large sample mean difference inspection that sample size n sample range is more than preset value (for example, n ＞ 30)
Method.It is that the probability that difference occurs is inferred with the theory of standardized normal distribution, so as to compare the difference of two average
It is whether notable.The calculation formula of its value is：
Wherein,It is the average of sample；μ0It is known overall average；S is the standard deviation of sample；N is that sample holds
Amount.Z dividing value tables are looked into, Probability p value are determined, then significance of difference relation table is judged according to table 2 below.Wherein, sample is pair
The sampling result of above-mentioned standard index factor, it is known that the overall statistical result i.e. to above-mentioned standard index factor, is with " taking out
The corresponding concept of sample ", sampling is produced from known totality.
The Z values of table 2 and significance of difference relation table
T statistics are applied to the small sample that sample size n sample range is less than or equal to preset value (for example, n≤30), population standard deviation
Normal distribution unknown σ.When overall distribution is normal distribution, such as population standard deviation is unknown and sample size is less than 30, then sample
The deviation statistic of this average and population average is distributed in t.T normalized set formula are：
Wherein,It is the average of sample；μ0It is known overall average；S is the standard deviation of sample；N is that sample holds
Amount.Wherein, sample is the sampling result to above-mentioned standard index factor, it is known that overall i.e. to above-mentioned standard index factor
Statistical result, is a concept corresponding with " sampling ", and sampling is produced from known totality.
It is the t statistics obtained by calculating that t, which is examined, according to free degree df=n-1, looks into t dividing value tables, determines Probability p value,
Significance of difference relation table is judged according to table 3 below again.
The t values of table 3 and significance of difference relation table
It is noted that numerical value in table 2 and table 3 can use above-mentioned value, also according to being actually needed sets itself.
In one embodiment, index and the power supply that 30 influence power supply reliabilities of selection can be calculated using SPSS are reliable
Rate RS-1 Pearson correlation coefficient.
In practical operation, the double tail significance tests of selection carry out SPSS correlation calculations inspections.Double tail significance tests
Refer to that alternative hypothesis does not have specific directionality, and the hypothesis testing containing symbol " ≠ ".Null hypothesis and alternative hypothesis are one
Self-contained mode, and it is mutually contradictory.For example：Null hypothesis is H0：M=m0, alternative hypothesis is H1：m≠m0, as two-tailed test.
Using SPSS data processings and examine post analysis, looped network rate can be drawn, stand between contact rate, insulate rate, overhead line
Insulation rate, cableization have significant correlation with power supply reliability RS-1.
Being met in coefficient correlation on the basis that conspicuousness is tested has following experience conclusion：
When the absolute value of correlation coefficient r is more than or equal to first threshold (for example, | r | >=0.8), two variables are can be considered
Between height correlation；
When the absolute value of correlation coefficient r be more than or equal to Second Threshold and less than first threshold (for example, 0.5≤| r |<
0.8) when, it can be considered that moderate is related；
When the absolute value of correlation coefficient r be more than or equal to the 3rd threshold value and less than Second Threshold (for example, 0.3≤| r |<
0.5) when, it is considered as lower correlation；
When correlation coefficient r absolute value be less than the 3rd threshold value (for example, | r |<0.3) when, the phase between two variables is illustrated
Pass degree is extremely weak, can be considered uncorrelated.
The index extraction for meeting significance test is come out, corresponding analysis result is as shown in table 4：
The Pearson came correlation analysis of table 4 and the significantly correlated indexs of power supply reliability RS-1
Because Pearson correlation coefficient necessarily assumes that data are obtained from normal distribution in couples, and data are being patrolled
It must be equidistant to collect in category.If this two condition is not met, a kind of possible solution is exactly to use Spearman
Rank correlation coefficient replaces Pearson correlation coefficient.Spearman rank correlation coefficient is used for evaluating working as characterizes two with monotonic function
How is effect during relation between variable, does not require that variable meets normal distribution.Spearman rank correlation coefficient is generally recognized
To be the Pearson correlation coefficient between the variable after arrangement, its computational methods is as follows：Assuming that original index factor xiAnd yi
By order arrangement from big to small, x is rememberedi' and yi' it is original index factor xiAnd yiPosition after arrangement where data, then xi′
And yi' it is referred to as variable xiAnd yiRank.
di=x 'i-y′i
Wherein：N is sample size；xiAnd yiFor the variate-value of two variables；R is Pearson correlation coefficient；diFor xiWith yi
Rank difference.The symbol of Spearman rank correlation coefficient represents the direction between x and y.When x increases with y increase, then phase
Relation number is positive number；When x reduces with y increase, then coefficient correlation is negative.When x and the absolute value of y coefficient correlations are 1
Represent to meet strict monotonic functional relationship between variable.
In one embodiment, index and the power supply that 30 influence power supply reliabilities of selection can be calculated using SPSS are reliable
Rate RS-1 Spearman rank correlation coefficient.The same double tail significance tests of selection, Data Management Analysis is obtained, and looped network rate, is stood
Between contact rate, the average segments of circuit, insulation rate, overhead line insulation rate, cable rate, bare conductor medium-voltage line failure
Rate, middle pressure fault outage suddenly repair average time, prerun regular inspection total degree, reduction average power off time of user, natural cause in place
Caused by the caused number of stoppages, external force factor there is clear-cut correlation in the number of stoppages and power supply reliability.
The index extraction for having clear-cut correlation is come out, corresponding analysis result is as shown in table 5：
The Spearman correlation analysis of table 5 and the significantly correlated indexs of power supply reliability RS-1
The analysis result of two kinds of correlation methods can be analyzed, influence power supply be filtered out after being weighed reliable
The leading indicator of property.
Due to thinking in correlation power analysis | r |<Degree of correlation when 0.3 between two variables is extremely weak, can be considered not
Correlation, therefore correlation is shown as into extremely weak index screened out.
The result obtained using power supply reliability RS-1 as power supply reliability evaluation index, Pearson came correlation analysis and this
The Comparative result that Joseph Pearman correlation analysis is obtained is as shown in Figure 3.Two methods of the result that contrast are obtained, the main shadow filtered out
Ring power supply reliability index for looped network rate, stand between contact rate, insulation rate, overhead line insulate rate, cable rate, order for a trial in advance
8 indexs such as the number of stoppages caused by the number of times of inspection, natural cause, the number of stoppages caused by external force factor.
Step S3：RBF (Radial Basis Function, RBF) is built according to the core index factor
Neutral net, obtains the optimal center vector of the hidden layer of the radial basis function neural network, optimal base fat vector and optimal
Weight is exported, mould is predicted according to the optimal center vector, optimal base fat vector and optimal output Weight Acquisition power supply reliability
Type, is predicted according to the power supply reliability forecast model to distribution network reliability.
RBF neural is a kind of three_layer planar waveguide being made up of input layer, hidden layer and output layer, hidden layer
The operation function of neuron uses the RBF with good local characteristicses so that input is only in the interval of a part
Certain output is produced by hidden layer, and interval outer output is substantially zeroed, therefore it has many advantages, such as, such as None-linear approximation energy
Power is strong, network structure is simple etc..RBF neural can avoid Local Minimum problem, but the requirement chosen to sample data is higher.
The action function of RBF neural hidden layer often takes Gaussian bases, and this layer of neuron j is output as：
X=[x1 x2 L xn]T∈Rn
Cj=[c1j c2j L cij L cnj]T
In formula：J is radial direction basic unit neuron number, and X is the input vector of neutral net, CjCentered on vector, b=(b1,b2,
L,bj) it is sound stage width vector.
Whole network is output as：
Y=WHT
W=[w1 w2 L wj]
H=[h1 h2 L hj]
In above-mentioned three formula：Y is the output vector of network, and W is radial direction basic unit and the weight vector of output interlayer, and H is hidden layer
Output, i.e. output layer input vector, wiFor output weight.
In actual applications, RBF keys are it needs to be determined that center vector Cj, sound stage width vector bjWith output weight wj.RBF nerves
Influence of the quantity at network hidden layer center to calculating speed is especially pronounced, and centric quantity is excessively obvious by the amount of calculation for making network
Increase.Traditional RBF neural typically asks for three above vector parameter, network instruction using clustering algorithm and least square method
Experienced accuracy and speed is all poor, therefore can be right using particle cluster algorithm (Particle Swarm Optimization, PSO)
RBF implicit layer parameter is optimized, to improve the training effectiveness of neutral net.
The search space of problem is compared to flock of birds flight space by PSO algorithms, by every bird it is abstract be a particle, for table
Show a solution to be selected of optimization problem, optimize the food that obtained optimal solution is equal to flock of birds searching.All particles have itself
Position and speed with determine its flight distance and direction, speed according to its own experience and colony's experience enter Mobile state adjust
It is whole.The speed of i-th particle and position are respectively V in D dimension spacesi=[vi,1,vi,2,L vi,D] and Xi=[xi,1,xi,2,L
xi,D], in each iteration, determine the optimum position p that t times each particle is passed throughbAnd the optimum bit that colony is found
Put gb, the two optimum positions, speed and position according to lower two formulas more new particle are updated by tracking.
vi,j(t+1)=vi,j(t)+c1r1[pi,j-xi,j(t)]+c2r2[pg,j-xi,j(t)]
xi,j(t+1)=xi,j(t)+vi,j(t+1), j=1,2, L D
In above-mentioned two formula, vi,j(t+1) be iteration j in i-th of particle in the speed at t+1 moment, vi,j(t) it is jth
I-th of particle is in the speed of t, c in secondary iteration1And c2For Studying factors, r1And r2For the random number between 0 to 1, D is mesh
The variable number of optimization, p are needed in scalar functionsi,jFor the individual history optimum position in iteration j, pg,jFor iteration j
In global optimum position, xi,j(t) be iteration j in i-th of particle in the position of t, xi,j(t+1) for jth time repeatedly
I-th of particle is in the position at t+1 moment in generation.
Such as Fig. 4, PSO-RBF training forecast models comprise the following steps：
(1) initialization particle colony, that is, determine the size and search dimension of population.And PSO relevant parameters are set, including
Inertia coeffeicent ω, Studying factors c1And c2, maximum iteration and particle maximal rate vmaxDeng.
(2) the corresponding center vector of each particle, sound stage width vector sum are exported according to the size of population and search dimension
Weight is substituted into RBF neural training algorithm, the predicted value exported using training sample.
(3) according to training sample calculate initialization population single particle MSE (Mean Squared Error, just
Error), as particle fitness fi。
(4) to each particle, by its current fitness fiWith each particle is passed through in an iteration individual history most
Best placement pbContrast, if fi＞ pbIllustrate that current fitness is higher, then will use current location more new individual history optimum position
pb, otherwise keep pbIt is constant.Similarly compare current adaptive value fiThe global optimum position g found with colony in an iterationb,
Work as fi＞ gbUpdate overall situation optimum position gb, otherwise keep gbIt is constant.
(5) according to the individual history optimum position and global optimum position, the speed of each particle is updated according to formula
With position.When iterations reaches that maximum iteration limitation or optimal adaptation degree reach that given threshold stops searching process.
(6) the optimal center vector C obtained by PSO algorithmsj, sound stage width vector bjWith output weight wjAfterwards, RBF god is utilized
Through algorithm for training network, formula Y=WH is substituted intoT, you can calculate model predication value.
Above-mentioned distribution network reliability Forecasting Methodology, the power distribution network based on correlation analysis and PSO-RBF is powered reliably
Property forecast model has good precision of prediction, stability and Generalization Capability.It is embodied in：
(1) the power supply reliability influence factor of many-sided multi-angle can be taken into full account, it is adaptable to the feelings of multi input variable
Condition, is conducive to more comprehensively analyzing prediction to power supply reliability.
(2) initial data to input carries out the processing of correlation analysis, and filter out influences more bright to power supply reliability
Aobvious index, effectively realizes the dimensionality reduction of input data, improves the precision of prediction and stably of RBF neural training pattern
Property.
(3) are exported by weight and is optimized for center vector, the sound stage width vector sum of neutral net using particle swarm optimization algorithm,
Effectively reduce training time and the precision of prediction of forecast model.
(4) it can be obtained according to the model trained to influenceing the correlative factor of power supply reliability index to carry out sensitivity analysis
Must be to the more sensitive correlated characteristic amount of power supply reliability index.
In one embodiment, the present invention also provides a kind of distribution network reliability forecasting system, as shown in figure 5, can
Including：
Index factor chooses module 10, for the grid structure according to power distribution network, technical equipment parameter, equipment quality ginseng
Number, fault compression and operation maintenance parameter choose the original index factor of a variety of influence power supply reliabilities；
The power supply reliability of power distribution network and power network itself deliverability, operation and maintenance level and the close phase of extraneous factor
Close.It is comprehensive for ensure to analyze power supply reliability, the technical equipment parameter of grid structure, power distribution network from power distribution network, set
Five aspects such as standby mass parameter, fault compression and operation maintenance parameter choose 30 influence power supply reliability indexs, as
Power supply reliability forecast model input variable.
Initial data screening module 20, for being standardized to the original index factor, obtains standardization and refers to
Mark factor, calculates the coefficient correlation between each standardized index factor and default reliability evaluation index, according to the phase
Relation number is filtered out from the index factor influences more obvious core index factor to power supply reliability；
Due to influenceing the factor of distribution network reliability varied, it is therefore desirable to many factors are screened, chosen
Go out the factor of core.Core index factor can be screened by correlation analysis.Correlation analysis be by calculate coefficient correlation come
Weigh and evaluate the strong or weak relation between variable.When power supply reliability evaluation index averagely has a power failure including power supply reliability, user
Between etc. a variety of evaluation indexes.In one embodiment, power supply reliability RS-1 can be chosen for evaluation index, SPSS is utilized
(Statistical Product and Service Solutions, statistical product and service solution) instrument carries out phase
The analysis of closing property is calculated.Power supply reliability RS-1 is interior during counting, to the total hourage of the effective power-on time of user and statistics phase
Between hourage ratio, its formula is
In formula, RS-1 is reliability evaluation index, T1For average power off time of user, T2For the time of a measurement period
Length.
In one embodiment, z-score standardizations can be carried out to initial data.By being carried out to initial data
Standardization, can eliminate the different influence of original variable dimension.Average and standard deviation based on initial data, by following
Formula standardizes original index factor using z-score, obtains standardized index factor.Below equation can specifically be used：
In formula, D' is standardized index factor, and D is original index factor,For the average of original index factor, δ is original
The standard deviation of beginning index factor.
The coefficient correlation can be Pearson correlation coefficient or Spearman rank correlation coefficient.Pearson correlation coefficient is used
The power of linear dependence between two variables are measured, it has certain application conditions：1) it is paired to necessarily assume that data
What ground was obtained from normal distribution；2) data must be equidistant at least in logical categories.Its mathematical definition is：
Wherein：N is sample size；xiAnd yiFor the variate-value of two variables；R is Pearson correlation coefficient.Pearson came is related
Coefficient value scope is [- 1,1], and r=1 is expressed as perfect positive correlation, and r=-1 is perfect negative correlation, and r=0 represents that line is not present
Property dependency relation.| r | it is better closer to 1 linear correlation degree, | r | it is weaker closer to 0 linear correlation degree.It should illustrate
, sample here refers to the sampling result to above-mentioned standard index factor, i.e. to each standardized index factor,
Several sample values, x are taken respectivelyiCorresponding i-th of the sample of currently processed standardized index factor is represented,Represent pre-treatment
Standardized index factor sample average, yiThe reliability evaluation index of i-th of sample is represented,Each reliability is represented to comment
The average of valency index.
Because coefficient correlation is that overall part sample is carried out calculating what is obtained, therefore it can not directly conclude samples sources
Totality meet significant dependency relation, it is necessary to by assuming that examine mode inferred, that is, utilize the mode of hypothesis testing
Mathematically illustrate the correctness of conclusion.Its basic step is：
(1) sample is built according to each standardized index factor, the sample and reliability evaluation index proposed former false
If H0With alternative hypothesis H1；Wherein, null hypothesis thinks the sample and reliability evaluation index without clear-cut correlation, alternative vacation
If thinking that the sample and reliability evaluation index have clear-cut correlation；
(2) according to sample size selection check statistic, such as t statistics, Z statistics；
(3) observation and corresponding Probability p of test statistics are calculated；
(4) determine significance decision-making α and make decisions.If the Probability p of test statistics is less than significance
α, then refuse null hypothesis, it was confirmed that alternative hypothesis, it is believed that two totality have significant dependency relation, can calculate each in sample
Pearson correlation coefficient between standardized index factor and default reliability evaluation index；, whereas if test statistics
Probability p be more than or equal to level of significance α, then can not refuse null hypothesis, it is believed that dependency relation is not present in two totality, can count
Calculate the Spearman rank correlation coefficient between each standardized index factor and default reliability evaluation index in sample.
Wherein, the statistic of step (2) depends on sample size size and selected.
Z statistics are applied to the large sample mean difference inspection that sample size n sample range is more than preset value (for example, n ＞ 30)
Method.It is that the probability that difference occurs is inferred with the theory of standardized normal distribution, so as to compare the difference of two average
It is whether notable.The calculation formula of its value is：
Wherein,It is the average of test samples；μ0It is known overall average；S is the standard deviation of sample；N is sample
This capacity.Z dividing value tables are looked into, Probability p value are determined, then significance of difference relation table is judged according to table 2.Wherein, sample is
To the sampling result of above-mentioned standard index factor, it is known that the overall statistical result i.e. to above-mentioned standard index factor, be with
" sampling " corresponding concept, sampling is produced from known totality.
The Z values of table 2 and significance of difference relation table
T statistics are applied to the small sample that sample size n sample range is less than or equal to preset value (for example, n≤30), population standard deviation
Normal distribution unknown σ.When overall distribution is normal distribution, such as population standard deviation is unknown and sample size is less than 30, then sample
The deviation statistic of this average and population average is distributed in t.T normalized set formula are：
Wherein,It is the average of test samples；μ0It is known overall average；S is the standard deviation of sample；N is sample
This capacity.
It is the t statistics obtained by calculating that t, which is examined, according to free degree df=n-1, looks into t dividing value tables, determines Probability p value,
Significance of difference relation table is judged according to table 3 below again.
The t values of table 3 and significance of difference relation table
It is noted that numerical value in table 2 and table 3 can use above-mentioned value, also according to being actually needed sets itself.
In one embodiment, index and the power supply that 30 influence power supply reliabilities of selection can be calculated using SPSS are reliable
Rate RS-1 Pearson correlation coefficient.
In practical operation, the double tail significance tests of selection carry out SPSS correlation calculations inspections.Double tail significance tests
Refer to that alternative hypothesis does not have specific directionality, and the hypothesis testing containing symbol " ≠ ".Null hypothesis and alternative hypothesis are one
Self-contained mode, and it is mutually contradictory.For example：Null hypothesis is H0：M=m0, alternative hypothesis is H1：m≠m0, as two-tailed test.
Using SPSS data processings and examine post analysis, looped network rate can be drawn, stand between contact rate, insulate rate, overhead line
Insulation rate, cableization have significant correlation with power supply reliability RS-1.
Being met in coefficient correlation on the basis that conspicuousness is tested has following experience conclusion：
When the absolute value of correlation coefficient r is more than or equal to first threshold (for example, | r | >=0.8), two variables are can be considered
Between height correlation；
When the absolute value of correlation coefficient r be more than or equal to Second Threshold and less than first threshold (for example, 0.5≤| r |<
0.8) when, it can be considered that moderate is related；
When the absolute value of correlation coefficient r be more than or equal to the 3rd threshold value and less than Second Threshold (for example, 0.3≤| r |<
0.5) when, it is considered as lower correlation；
When correlation coefficient r absolute value be less than the 3rd threshold value (for example, | r |<0.3) when, the phase between two variables is illustrated
Pass degree is extremely weak, can be considered uncorrelated.
The index extraction for meeting significance test is come out, corresponding analysis result is as shown in table 4：
The Pearson came correlation analysis of table 4 and the significantly correlated indexs of power supply reliability RS-1
Because Pearson correlation coefficient necessarily assumes that data are obtained from normal distribution in couples, and data are being patrolled
It must be equidistant to collect in category.If this two condition is not met, a kind of possible solution is exactly to use Spearman
Rank correlation coefficient replaces Pearson correlation coefficient.Spearman rank correlation coefficient is used for evaluating working as characterizes two with monotonic function
How is effect during relation between variable, does not require that variable meets normal distribution.Spearman rank correlation coefficient is generally recognized
To be the Pearson correlation coefficient between the variable after arrangement, its computational methods is as follows：Assuming that original index factor xiAnd yi
By order arrangement from big to small, x ' is rememberediWith y 'iFor original index factor xiAnd yiPosition after arrangement where data, then x 'i
With y 'iReferred to as variable xiAnd yiRank.
di=x 'i-y′i
Wherein：N is sample size；xiAnd yiFor the variate-value of two variables；R is Pearson correlation coefficient；diFor xiWith yi
Rank difference.The symbol of Spearman rank correlation coefficient represents the direction between x and y.When x increases with y increase, then phase
Relation number is positive number；When x reduces with y increase, then coefficient correlation is negative.When x and the absolute value of y coefficient correlations are 1
Represent to meet strict monotonic functional relationship between variable.
In one embodiment, index and the power supply that 30 influence power supply reliabilities of selection can be calculated using SPSS are reliable
Rate RS-1 Spearman rank correlation coefficient.The same double tail significance tests of selection, Data Management Analysis is obtained, and looped network rate, is stood
Between contact rate, the average segments of circuit, insulation rate, overhead line insulation rate, cable rate, bare conductor medium-voltage line failure
Rate, middle pressure fault outage suddenly repair average time, prerun regular inspection total degree, reduction average power off time of user, natural cause in place
Caused by the caused number of stoppages, external force factor there is clear-cut correlation in the number of stoppages and power supply reliability.
The index extraction for having clear-cut correlation is come out, corresponding analysis result is as shown in table 5：
The Spearman correlation analysis of table 5 and the significantly correlated indexs of power supply reliability RS-1
The analysis result of two kinds of correlation methods can be analyzed, influence power supply be filtered out after being weighed reliable
The leading indicator of property.
Due to thinking in correlation power analysis | r |<Degree of correlation when 0.3 between two variables is extremely weak, can be considered not
Correlation, therefore correlation is shown as into extremely weak index screened out.
The result obtained using power supply reliability RS-1 as power supply reliability evaluation index, Pearson came correlation analysis and this
The Comparative result that Joseph Pearman correlation analysis is obtained is as shown in Figure 3.Two methods of the result that contrast are obtained, the main shadow filtered out
Ring power supply reliability index for looped network rate, stand between contact rate, insulation rate, overhead line insulate rate, cable rate, order for a trial in advance
8 indexs such as the number of stoppages caused by the number of times of inspection, natural cause, the number of stoppages caused by external force factor.
Forecast model training module 30, for building radial basis function neural network according to the core index factor, is obtained
Take the optimal center vector, optimal base fat vector and optimal output weight of the hidden layer of the radial basis function neural network, root
According to the optimal center vector, optimal base fat vector and optimal output Weight Acquisition power supply reliability forecast model, according to described
Power supply reliability forecast model is predicted to distribution network reliability.
RBF neural is a kind of three_layer planar waveguide being made up of input layer, hidden layer and output layer, hidden layer
The operation function of neuron uses the RBF with good local characteristicses so that input is only in the interval of a part
Certain output is produced by hidden layer, and interval outer output is substantially zeroed, therefore it has many advantages, such as, such as None-linear approximation energy
Power is strong, network structure is simple etc..RBF neural can avoid Local Minimum problem, but the requirement chosen to sample data is higher.
The action function of RBF neural hidden layer often takes Gaussian bases, and this layer of neuron j is output as：
X=[x1 x2 L xn]T∈Rn
Cj=[c1j c2j L cij L cnj]T
In formula：J is radial direction basic unit neuron number, and X is the input vector of neutral net, CjCentered on vector, b=(b1,b2,
L,bj) it is sound stage width vector.
Whole network is output as：
Y=WHT
W=[w1 w2 L wj]
H=[h1 h2 L hj]
In above-mentioned three formula：Y is the output vector of network, and W is radial direction basic unit and the weight vector of output interlayer, and H is hidden layer
Output, i.e. output layer input vector, wiFor output weight.
In actual applications, RBF keys are it needs to be determined that center vector Cj, sound stage width vector bjWith output weight wj.RBF nerves
Influence of the quantity at network hidden layer center to calculating speed is especially pronounced, and centric quantity is excessively obvious by the amount of calculation for making network
Increase.Traditional RBF neural typically asks for three above vector parameter, network instruction using clustering algorithm and least square method
Experienced accuracy and speed is all poor, therefore can be right using particle cluster algorithm (Particle Swarm Optimization, PSO)
RBF implicit layer parameter is optimized, to improve the training effectiveness of neutral net.
The search space of problem is compared to flock of birds flight space by PSO algorithms, by every bird it is abstract be a particle, for table
Show a solution to be selected of optimization problem, optimize the food that obtained optimal solution is equal to flock of birds searching.All particles have itself
Position and speed with determine its flight distance and direction, speed according to its own experience and colony's experience enter Mobile state adjust
It is whole.The speed of i-th particle and position are respectively V in D dimension spacesi=[vi,1,vi,2,L vi,D] and Xi=[xi,1,xi,2,L
xi,D], in each iteration, determine the optimum position p that t times each particle is passed throughbAnd the optimum bit that colony is found
Put gb, the two optimum positions, speed and position according to lower two formulas more new particle are updated by tracking.
vi,j(t+1)=vi,j(t)+c1r1[pi,j-xi,j(t)]+c2r2[pg,j-xi,j(t)]
xi,j(t+1)=xi,j(t)+vi,j(t+1), j=1,2, L D
In above-mentioned two formula, vi,j(t+1) be iteration j in i-th of particle in the speed at t+1 moment, vi,j(t) it is jth
I-th of particle is in the speed of t, c in secondary iteration1And c2For Studying factors, r1And r2For the random number between 0 to 1, D is mesh
The variable number of optimization, p are needed in scalar functionsi,jFor the individual history optimum position in iteration j, pg,jFor iteration j
In global optimum position, xi,j(t) be iteration j in i-th of particle in the position of t, xi,j(t+1) for jth time repeatedly
I-th of particle is in the position at t+1 moment in generation.
Such as Fig. 4, PSO-RBF training forecast models comprise the following steps：
(1) initialization particle colony, that is, determine the size and search dimension of population.And PSO relevant parameters are set, including
Inertia coeffeicent ω, Studying factors c1And c2, maximum iteration and particle maximal rate vmaxDeng.
(2) the corresponding center vector of each particle, sound stage width vector sum are exported according to the size of population and search dimension
Weight is substituted into RBF neural training algorithm, the predicted value exported using training sample.
(3) according to training sample calculate initialization population single particle MSE (Mean Squared Error, just
Error), as particle fitness fi。
(4) to each particle, by its current fitness fiWith each particle is passed through in an iteration individual history most
Best placement pbContrast, if fi＞ pbIllustrate that current fitness is higher, then will use current location more new individual history optimum position
pb, otherwise keep pbIt is constant.Similarly compare current adaptive value fiThe global optimum position g found with colony in an iterationb,
Work as fi＞ gbUpdate overall situation optimum position gb, otherwise keep gbIt is constant.
(5) according to the individual history optimum position and global optimum position, the speed of each particle is updated according to formula
With position.When iterations reaches that maximum iteration limitation or optimal adaptation degree reach that given threshold stops searching process.
(6) the optimal center vector C obtained by PSO algorithmsj, sound stage width vector bjWith output weight wjAfterwards, RBF god is utilized
Through algorithm for training network, formula Y=WH is substituted intoT, you can calculate model predication value.
Above-mentioned distribution network reliability Forecasting Methodology, the power distribution network based on correlation analysis and PSO-RBF is powered reliably
Property forecast model has good precision of prediction, stability and Generalization Capability.It is embodied in：
(1) the power supply reliability influence factor of many-sided multi-angle can be taken into full account, it is adaptable to the feelings of multi input variable
Condition, is conducive to more comprehensively analyzing prediction to power supply reliability.
(2) initial data to input carries out the processing of correlation analysis, and filter out influences more bright to power supply reliability
Aobvious index, effectively realizes the dimensionality reduction of input data, improves the precision of prediction and stably of RBF neural training pattern
Property.
(3) are exported by weight and is optimized for center vector, the sound stage width vector sum of neutral net using particle swarm optimization algorithm,
Effectively reduce training time and the precision of prediction of forecast model.
(4) it can be obtained according to the model trained to influenceing the correlative factor of power supply reliability index to carry out sensitivity analysis
Must be to the more sensitive correlated characteristic amount of power supply reliability index.
The distribution network reliability Forecasting Methodology one of the distribution network reliability forecasting system of the present invention and the present invention
One correspondence, the technical characteristic illustrated in the embodiment of above-mentioned distribution network reliability Forecasting Methodology and its advantage are applicable
In the embodiment of distribution network reliability forecasting system, hereby give notice that.
Each technical characteristic of embodiment described above can be combined arbitrarily, to make description succinct, not to above-mentioned reality
Apply all possible combination of each technical characteristic in example to be all described, as long as however, the combination of these technical characteristics is not deposited
In contradiction, the scope of this specification record is all considered to be.
Embodiment described above only expresses the several embodiments of the present invention, and it describes more specific and detailed, but simultaneously
Can not therefore it be construed as limiting the scope of the patent.It should be pointed out that coming for one of ordinary skill in the art
Say, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to the protection of the present invention
Scope.Therefore, the protection domain of patent of the present invention should be determined by the appended claims.
Claims (10)
1. a kind of distribution network reliability Forecasting Methodology, it is characterised in that comprise the following steps：
Selected according to the grid structure of power distribution network, technical equipment parameter, equipment quality parameter, fault compression and operation maintenance parameter
Take the original index factor of a variety of influence power supply reliabilities；
The original index factor is standardized, standardized index factor is obtained, calculate each standardized index because
The plain coefficient correlation between default reliability evaluation index, is filtered out according to the coefficient correlation from the index factor
More obvious core index factor is influenceed on power supply reliability；
Radial basis function neural network is built according to the core index factor, the hidden of the radial basis function neural network is obtained
Optimal center vector, optimal base fat vector and optimal output weight containing layer, according to the optimal center vector, optimal sound stage width to
Amount and optimal output Weight Acquisition power supply reliability forecast model, power according to the power supply reliability forecast model to power distribution network
Reliability is predicted.
2. distribution network reliability Forecasting Methodology according to claim 1, it is characterised in that enter to the index factor
The step of row standardization, includes：
The index factor is standardized according in the following manner：
In formula, D' is standardized index factor, and D is original index factor,For the average of original index factor, δ refers to be original
The standard deviation of mark factor.
3. distribution network reliability Forecasting Methodology according to claim 1, it is characterised in that calculate each index because
It is further comprising the steps of before the plain coefficient correlation between default reliability evaluation index：
Reliability evaluation index is calculated according in the following manner：
In formula, RS-1 is reliability evaluation index, T1For average power off time of user, T2For the time span of a measurement period.
4. distribution network reliability Forecasting Methodology according to claim 1, it is characterised in that calculate each standardization and refer to
The step of coefficient correlation between mark factor and default reliability evaluation index, includes：
Sample is built according to each standardized index factor, null hypothesis is proposed and alternative to the sample and reliability evaluation index
Assuming that；Wherein, null hypothesis thinks the sample and reliability evaluation index without clear-cut correlation, and alternative hypothesis thinks the sample
This has clear-cut correlation with reliability evaluation index；
According to sample size selection check statistic, the observation and corresponding probability of test statistics are calculated；
If the probability of test statistics is less than default significance decision factor, the sample and reliability evaluation are judged
Index has clear-cut correlation, calculates in sample between each standardized index factor and default reliability evaluation index
Pearson correlation coefficient.
5. distribution network reliability Forecasting Methodology according to claim 4, it is characterised in that calculate each standardization and refer to
The step of coefficient correlation between mark factor and default reliability evaluation index, also includes：
If the probability of test statistics is more than or equal to default significance decision factor, the sample is judged and reliable
Property evaluation index do not have clear-cut correlation；
Calculate the Spearman rank correlation coefficient between each standardized index factor and default reliability evaluation index.
6. distribution network reliability Forecasting Methodology according to claim 4, it is characterised in that selected according to sample size
The step of test statistics, includes：
When sample size is larger, selection Z statistics are used as test statistics；
When sample size is smaller, selection t statistics are used as test statistics.
7. distribution network reliability Forecasting Methodology according to claim 1, it is characterised in that obtain the radial direction base letter
The step of optimal center vector, optimal base fat vector and optimal output weight for counting the hidden layer of neutral net, includes：
Determine the size and search dimension of population；
The corresponding center vector of each particle, sound stage width vector sum output weight are substituted into according to the size of population and search dimension
In RBF neural training algorithm, the predicted value exported using training sample；
The MSE of initialization population single particle is calculated according to training sample, as particle fitness；
To each particle, the individual history optimum position that each particle is passed through in its current fitness and an iteration is carried out
Contrast, and the global optimum position that its current fitness is found with colony in an iteration is contrasted；
If current fitness is more than individual history optimum position, according to current location more new individual history optimum position, if currently
Fitness is more than global optimum position, and global optimum position is updated according to current location；
Speed and the position of each particle are updated according to the individual history optimum position and global optimum position, according to each grain
The speed of son and position calculate the optimal center vector of the hidden layer of the radial basis function neural network, optimal base fat vector and
Optimal output weight.
8. distribution network reliability Forecasting Methodology according to claim 7, it is characterised in that according to the individual history
The step of optimum position and global optimum position update speed and the position of each particle includes：
Speed and the position of each particle are updated according in the following manner：
vi,j(t+1)=vi,j(t)+c1r1[pi,j-xi,j(t)]+c2r2[pg,j-xi,j(t)]；
xi,j(t+1)=xi,j(t)+vi,j(t+1), j=1,2, L D；
In formula, vi,j(t+1) be iteration j in i-th of particle in the speed at t+1 moment, vi,j(t) it is i-th in iteration j
Individual particle is in the speed of t, c1And c2For Studying factors, r1And r2For the random number between 0 to 1, D is to need in object function
The variable number to be optimized, pi,jFor the individual history optimum position in iteration j, pg,jFor the overall situation in iteration j most
Best placement, xi,j(t) be iteration j in i-th of particle in the position of t, xi,j(t+1) it is i-th in iteration j
Son is in the position at t+1 moment.
9. distribution network reliability Forecasting Methodology according to claim 1, it is characterised in that according to the center to
The step of amount, sound stage width vector sum output Weight Acquisition power supply reliability forecast model, includes：
Power supply reliability forecast model is obtained according in the following manner：
Y=WHT
W=[w1 w2 L wj]；
H=[h1 h2 L hj]
In formula, Y is power supply reliability forecast model, and W is radial direction basic unit and the weight vector of output interlayer, and H is the output of hidden layer,
wiFor output weight, X is the input vector of neutral net, CjCentered on vector, b=(b1,b2,L,bj) vectorial for sound stage width, J is footpath
To basic unit's neuron number.
10. a kind of distribution network reliability forecasting system, it is characterised in that including：
Index factor chooses module, for the grid structure according to power distribution network, technical equipment parameter, equipment quality parameter, failure
The factor and operation maintenance parameter choose the original index factor of a variety of influence power supply reliabilities；
Initial data screening module, for being standardized to the original index factor, obtains standardized index factor,
Calculate the coefficient correlation between each standardized index factor and default reliability evaluation index, according to the coefficient correlation from
Filtered out in the index factor influences more obvious core index factor to power supply reliability；
Forecast model training module, for building radial basis function neural network according to the core index factor, obtains described
Optimal center vector, optimal base fat vector and the optimal output weight of the hidden layer of radial basis function neural network, according to described
Optimal center vector, optimal base fat vector and optimal output Weight Acquisition power supply reliability forecast model, can according to the power supply
Distribution network reliability is predicted by property forecast model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710154078.2A CN106980905B (en) | 2017-03-15 | 2017-03-15 | Power distribution network power supply reliability prediction method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710154078.2A CN106980905B (en) | 2017-03-15 | 2017-03-15 | Power distribution network power supply reliability prediction method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106980905A true CN106980905A (en) | 2017-07-25 |
| CN106980905B CN106980905B (en) | 2020-07-07 |
Family
ID=59338794
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710154078.2A Active CN106980905B (en) | 2017-03-15 | 2017-03-15 | Power distribution network power supply reliability prediction method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106980905B (en) |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107833170A (en) * | 2017-11-28 | 2018-03-23 | 东莞优闪电子科技有限公司 | A kind of method that individualized learning is realized in K12 education coupling |
| CN107886358A (en) * | 2017-11-08 | 2018-04-06 | 北京理工大学 | A kind of Agricultural Machinery Equipment service of goods needing forecasting method |
| CN108665181A (en) * | 2018-05-18 | 2018-10-16 | 中国电力科学研究院有限公司 | A kind of appraisal procedure and device of distribution network reliability |
| CN109165819A (en) * | 2018-08-03 | 2019-01-08 | 国网山东省电力公司聊城供电公司 | A kind of active power distribution network reliability fast evaluation method based on improvement AdaBoost.M1-SVM |
| CN109242340A (en) * | 2018-09-29 | 2019-01-18 | 国网辽宁省电力有限公司电力科学研究院 | A kind of thermal storage electric boiler consumption system monitoring data evaluation system and its evaluation method |
| CN109858663A (en) * | 2018-11-19 | 2019-06-07 | 中国农业大学 | Distribution network failure power failure INFLUENCING FACTORS analysis based on BP neural network |
| CN110059013A (en) * | 2019-04-24 | 2019-07-26 | 北京百度网讯科技有限公司 | The determination method and device operated normally after software upgrading |
| CN110210740A (en) * | 2019-05-22 | 2019-09-06 | 广西电网有限责任公司电力科学研究院 | A kind of distribution network reliability evaluation method considering power supply quality |
| CN110570104A (en) * | 2019-08-23 | 2019-12-13 | 国网山东省电力公司泗水县供电公司 | Power distribution network power supply reliability analysis method and device |
| CN110764938A (en) * | 2019-10-29 | 2020-02-07 | 广东电网有限责任公司 | Mutual inductor equipment reliability prediction method and device and storage medium |
| CN110909916A (en) * | 2019-10-24 | 2020-03-24 | 国网辽宁省电力有限公司 | Entropy weight method based wind power generation monthly electric quantity interval prediction method |
| CN111458661A (en) * | 2020-03-03 | 2020-07-28 | 国网江苏省电力有限公司电力科学研究院 | Power distribution network line variation relation diagnosis method, device and system |
| CN111784019A (en) * | 2019-12-26 | 2020-10-16 | 国网北京市电力公司 | Power load processing method and device |
| CN111931966A (en) * | 2019-05-13 | 2020-11-13 | 国家电网有限公司 | Power supply reliability prediction method based on decision tree regression |
| CN112100769A (en) * | 2020-09-08 | 2020-12-18 | 上海三菱电梯有限公司 | Method for constructing elevator component performance model |
| CN112330121A (en) * | 2020-10-28 | 2021-02-05 | 广东电网有限责任公司佛山供电局 | Power distribution network automation construction investment benefit evaluation method |
| CN112446599A (en) * | 2020-11-19 | 2021-03-05 | 广东电网有限责任公司 | Power supply reliability index prediction method, device, equipment and storage medium |
| CN112906251A (en) * | 2021-04-16 | 2021-06-04 | 云南电网有限责任公司 | Analysis method and system for reliability influence factors of power distribution network |
| CN113436737A (en) * | 2021-06-24 | 2021-09-24 | 杭州师范大学 | Prediction evaluation method and device for depression level of large population |
| CN117455269A (en) * | 2023-12-21 | 2024-01-26 | 国网天津市电力公司城南供电分公司 | Snowflake type power distribution network power supply safety prediction method, device, equipment and storage medium |
| CN117852229A (en) * | 2024-03-06 | 2024-04-09 | 山东大学 | New energy-containing power distribution network power supply reliability influence factor analysis method and system |
-
2017
- 2017-03-15 CN CN201710154078.2A patent/CN106980905B/en active Active
Non-Patent Citations (5)
| Title |
|---|
| 刘明华，等: "《配电网可靠性预测综述》", 《电网技术》 * |
| 崔海青，等: "《基于粒子群算法的RBF网络参数优化算法》", 《计算机技术与发展》 * |
| 朱震曙，等: "《基于QPSO_RBFNN的短期电力负荷预测模型》", 《南京理工大学学报》 * |
| 杨江，刘鹏飞: "《基于指标数据的供电可靠性预测方法》", 《云南电力技术》 * |
| 杨海波，等: "《假设检验在成绩分析中的应用》", 《内蒙古农业大学学报（自然科学版）》 * |
Cited By (31)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107886358A (en) * | 2017-11-08 | 2018-04-06 | 北京理工大学 | A kind of Agricultural Machinery Equipment service of goods needing forecasting method |
| CN107833170B (en) * | 2017-11-28 | 2021-08-24 | 广东墨痕教育科技有限公司 | Method for realizing personalized learning by K12 education coupling |
| CN107833170A (en) * | 2017-11-28 | 2018-03-23 | 东莞优闪电子科技有限公司 | A kind of method that individualized learning is realized in K12 education coupling |
| CN108665181A (en) * | 2018-05-18 | 2018-10-16 | 中国电力科学研究院有限公司 | A kind of appraisal procedure and device of distribution network reliability |
| CN109165819A (en) * | 2018-08-03 | 2019-01-08 | 国网山东省电力公司聊城供电公司 | A kind of active power distribution network reliability fast evaluation method based on improvement AdaBoost.M1-SVM |
| CN109165819B (en) * | 2018-08-03 | 2021-09-14 | 国网山东省电力公司聊城供电公司 | Active power distribution network reliability rapid evaluation method based on improved AdaBoost. M1-SVM |
| CN109242340B (en) * | 2018-09-29 | 2021-08-27 | 国网辽宁省电力有限公司电力科学研究院 | Monitoring data evaluation system and evaluation method for heat storage electric boiler digestion system |
| CN109242340A (en) * | 2018-09-29 | 2019-01-18 | 国网辽宁省电力有限公司电力科学研究院 | A kind of thermal storage electric boiler consumption system monitoring data evaluation system and its evaluation method |
| CN109858663A (en) * | 2018-11-19 | 2019-06-07 | 中国农业大学 | Distribution network failure power failure INFLUENCING FACTORS analysis based on BP neural network |
| CN110059013A (en) * | 2019-04-24 | 2019-07-26 | 北京百度网讯科技有限公司 | The determination method and device operated normally after software upgrading |
| CN111931966A (en) * | 2019-05-13 | 2020-11-13 | 国家电网有限公司 | Power supply reliability prediction method based on decision tree regression |
| CN110210740A (en) * | 2019-05-22 | 2019-09-06 | 广西电网有限责任公司电力科学研究院 | A kind of distribution network reliability evaluation method considering power supply quality |
| CN110210740B (en) * | 2019-05-22 | 2023-09-15 | 广西电网有限责任公司电力科学研究院 | Power distribution network reliability assessment method considering power supply quality |
| CN110570104A (en) * | 2019-08-23 | 2019-12-13 | 国网山东省电力公司泗水县供电公司 | Power distribution network power supply reliability analysis method and device |
| CN110909916B (en) * | 2019-10-24 | 2023-06-13 | 国网辽宁省电力有限公司 | Wind power generation month electric quantity interval prediction method based on entropy weight method |
| CN110909916A (en) * | 2019-10-24 | 2020-03-24 | 国网辽宁省电力有限公司 | Entropy weight method based wind power generation monthly electric quantity interval prediction method |
| CN110764938A (en) * | 2019-10-29 | 2020-02-07 | 广东电网有限责任公司 | Mutual inductor equipment reliability prediction method and device and storage medium |
| CN111784019A (en) * | 2019-12-26 | 2020-10-16 | 国网北京市电力公司 | Power load processing method and device |
| CN111458661A (en) * | 2020-03-03 | 2020-07-28 | 国网江苏省电力有限公司电力科学研究院 | Power distribution network line variation relation diagnosis method, device and system |
| CN112100769A (en) * | 2020-09-08 | 2020-12-18 | 上海三菱电梯有限公司 | Method for constructing elevator component performance model |
| CN112100769B (en) * | 2020-09-08 | 2022-08-02 | 上海三菱电梯有限公司 | Method for constructing elevator component performance model |
| CN112330121B (en) * | 2020-10-28 | 2022-04-19 | 广东电网有限责任公司佛山供电局 | Power distribution network automation construction investment benefit evaluation method |
| CN112330121A (en) * | 2020-10-28 | 2021-02-05 | 广东电网有限责任公司佛山供电局 | Power distribution network automation construction investment benefit evaluation method |
| CN112446599B (en) * | 2020-11-19 | 2023-01-24 | 广东电网有限责任公司 | Power supply reliability index prediction method, device, equipment and storage medium |
| CN112446599A (en) * | 2020-11-19 | 2021-03-05 | 广东电网有限责任公司 | Power supply reliability index prediction method, device, equipment and storage medium |
| CN112906251A (en) * | 2021-04-16 | 2021-06-04 | 云南电网有限责任公司 | Analysis method and system for reliability influence factors of power distribution network |
| CN112906251B (en) * | 2021-04-16 | 2023-05-02 | 云南电网有限责任公司 | Analysis method and system for reliability influence elements of power distribution network |
| CN113436737A (en) * | 2021-06-24 | 2021-09-24 | 杭州师范大学 | Prediction evaluation method and device for depression level of large population |
| CN117455269A (en) * | 2023-12-21 | 2024-01-26 | 国网天津市电力公司城南供电分公司 | Snowflake type power distribution network power supply safety prediction method, device, equipment and storage medium |
| CN117455269B (en) * | 2023-12-21 | 2024-04-16 | 国网天津市电力公司城南供电分公司 | Snowflake type power distribution network power supply safety prediction method, device, equipment and storage medium |
| CN117852229A (en) * | 2024-03-06 | 2024-04-09 | 山东大学 | New energy-containing power distribution network power supply reliability influence factor analysis method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106980905B (en) | 2020-07-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106980905A (en) | Distribution network reliability Forecasting Methodology and system | |
| CN105117602B (en) | A kind of metering device running status method for early warning | |
| CN105512448B (en) | A kind of appraisal procedure of power distribution network health index | |
| CN112149873B (en) | Low-voltage station line loss reasonable interval prediction method based on deep learning | |
| CN111429034A (en) | Method for predicting power distribution network fault | |
| CN109872003A (en) | Obj State prediction technique, system, computer equipment and storage medium | |
| CN106384186A (en) | Distributed new energy grid-connected power quality monitoring method and system | |
| CN103440410A (en) | Main variable individual defect probability forecasting method | |
| CN103617447B (en) | The evaluation system of intelligent substation and evaluation methodology | |
| CN112307963A (en) | Converter transformer running state identification method based on vibration signals | |
| CN111680712B (en) | Method, device and system for predicting oil temperature of transformer based on similar time in day | |
| CN106408016A (en) | Distribution network power outage time automatic identification model construction method | |
| Nouri et al. | Reconciliation of measured and forecast data for topology identification in distribution systems | |
| Ghaemi et al. | Situational awareness and deficiency warning system in a smart distribution network based on stacking ensemble learning | |
| CN110533213A (en) | Transmission line of electricity defect Risk Modeling and its prediction technique based on support vector machines | |
| CN113486291B (en) | Petroleum drilling machine micro-grid fault prediction method based on deep learning | |
| CN116415724A (en) | Photovoltaic power station operation maintenance time prediction method and device | |
| Chi et al. | A fault diagnosis method of double-layer LSTM for 10 kV single-core cable based on multiple observable electrical quantities | |
| CN110727912B (en) | Method for selecting differential transformation scheme of power secondary equipment | |
| Jin et al. | Deep Machine Learning-Based Asset Management Approach for Oil-Immersed Power Transformers Using Dissolved Gas Analysis | |
| Huang et al. | Data-driven fault risk warning method for distribution system | |
| López et al. | Phasor measurement unit placement for wide area instability detection and prediction | |
| Li et al. | Risk prediction of the SCADA communication network based on entropy-gray model | |
| Murali et al. | Forecasting and Classification Of Power Quality Disturbance In Smart Grid Using Hybrid Networks | |
| Bai et al. | Abnormal Detection Scheme of Substation Equipment based on Intelligent Fusion Terminal |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20210812 Address after: 510663 3 building, 3, 4, 5 and J1 building, 11 building, No. 11, Ke Xiang Road, Luogang District Science City, Guangzhou, Guangdong. Patentee after: China Southern Power Grid Research Institute Co.,Ltd. Patentee after: SOUTH CHINA University OF TECHNOLOGY Address before: 510080 water Donggang 8, Dongfeng East Road, Yuexiu District, Guangzhou, Guangdong. Patentee before: China Southern Power Grid Research Institute Co.,Ltd. Patentee before: CSG POWER GRID TECHNOLOGY RESEARCH CENTER Patentee before: SOUTH CHINA University OF TECHNOLOGY |