CN103605103B - Electric energy metrical intelligent fault diagnosis method based on S type curvilinear function - Google Patents
Electric energy metrical intelligent fault diagnosis method based on S type curvilinear function Download PDFInfo
- Publication number
- CN103605103B CN103605103B CN201310261411.1A CN201310261411A CN103605103B CN 103605103 B CN103605103 B CN 103605103B CN 201310261411 A CN201310261411 A CN 201310261411A CN 103605103 B CN103605103 B CN 103605103B
- Authority
- CN
- China
- Prior art keywords
- phase
- electricity
- fault
- metering
- data
- 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.)
- Active
Links
Abstract
A kind of electric energy metrical intelligent fault diagnosis method based on S type curvilinear function, comprises the following steps: S1: extract Real-time Load, day electricity, terminal alarms, the historical data of main website warning from metering automation system;S2: extract metering fault evaluation index from historical data;S3: build training sample set based on S type curvilinear function;S4: build measurement fault diagnostic model;S5: metering fault intelligent diagnostics;S6: optimize measurement fault diagnostic model parameter, and reconstruct measurement fault diagnostic model.This method, based on electric energy measurement automation system, automatically can carry out intelligent diagnostics analysis to the electric power data of magnanimity.Metering fault diagnostic system can carry out automatically adjusting and optimizing of model structure and parameters according to its environment applied, up-to-date real data and system feedback, reaches more preferable diagnosis effect.
Description
Technical field
The method that the metering fault that the present invention relates to occur in a kind of electric energy measurement automation system carries out intelligent diagnostics, especially
It relates to a kind of electric energy metrical intelligent fault diagnosis method based on S type curvilinear function.
Technical background
Along with developing rapidly of China market economy, the increase rapidly of business electrical amount, power supply complexity and electricity consumption are various
The electricity characteristics such as property become increasingly conspicuous, and multiplexing electric abnormality phenomenon is also occurring again and again towards more serious direction.The swift and violent of science and technology is sent out
Exhibition makes power supply enterprise can provide the electric service of high-quality more efficiently for power consumer, and multiplexing electric abnormality behavior simultaneously also exists
More complexity and crypticity under high-tech background, cause country and power supply enterprise to suffer huge loss.Due to the most right
Monitoring in power consumer multiplexing electric abnormality lacks advanced technological means, seriously constrain power utility check department work efficiency and
Work accuracy.
Electric energy measurement automation system be established as the supervision of electricity consumption behavior is provided the technical support of science, for set up
A set of effective multiplexing electric abnormality investigation mechanism provides solid data basis.When power supply enterprise is to certain power consumer
When electricity consumption behavior throws doubt upon, the electric quantity data of electric energy measurement automation system Real-time Collection power consumer, load number can be passed through
According to this and event information, then by technical backbone, substantial amounts of electric power data is collected, analyzes, adds up.Via electric energy metrical
Automatization's means of automated system advanced person combine with the industry experience of power supply enterprise technical backbone, it is possible to relatively accurately sentence
Break and whether suspicion user exists multiplexing electric abnormality behavior.The operational efficiency of system depends on the performance of server, hard by promoting
Part configuration or optimization software design can effectively strengthen system remote and gather the ability of electricity consumption data.But the personnel of power supply enterprise
Configuration but can become the bottleneck that restriction multiplexing electric abnormality is analyzed, and has the technical backbone of abundant industry experience after all still for counting not
Many, it is insufficient for mass data being carried out the personnel demand of empirical analysis.
Summary of the invention
The technical problem to be solved, it is simply that a kind of electric energy metrical fault intelligence based on S type curvilinear function is provided
Can diagnostic method, it can diagnose the metering fault problem occurred in electric energy measurement automation system, reaches really to navigate to different
The purpose of conventional electricity user.
Solving above-mentioned technical problem, the technical solution used in the present invention is:
A kind of electric energy metrical intelligent fault diagnosis method based on S type curvilinear function, comprises the following steps:
S1: extract Real-time Load, day electricity, terminal alarms, the history of main website warning from electric energy measurement automation system
Data;
S2: extract metering fault evaluation index from historical data;
S3: build training sample set based on S type curvilinear function;
S4: build measurement fault diagnostic model;
S5: metering fault intelligent diagnostics;
S6: optimize measurement fault diagnostic model parameter, and reconstruct measurement fault diagnostic model.
Described S1 includes following sub-step:
S1.1: determine sample coverage
Whole relevant continuous data in the sample data time range chosen;
S1.2: determine sample data scope
During sample data extraction, all large users of abnormal electricity consumption to be comprised, gather from electric energy measurement automation system
Data include:
Real-time Load: time point, stoichiometric point, A/B/C phase are meritorious, A/B/C phase current, A/B/C phase voltage and A/B/C phase merit
Rate factor;
Day electricity: time point, stoichiometric point, meritorious total, peak/flat/paddy and capacity of idle power;
Terminal alarms: electricity sampling open-phase, voltage phase shortage and voltage circuit negative phase sequence day alarm times;
Main website reports to the police: defluidization, overload and imbalance day alarm times.
Described S2 comprises following sub-step:
S2.1 extracts average daily decompression/overvoltage rate
Calculate the every phase voltage in three-phase and rated voltage difference respectively, count numCount simultaneously;
If difference is less than 0 and less than decompression dividing value, it is judged that for decompression and count numLower;
If difference is more than 0 and more than overvoltage dividing value, it is judged that for overvoltage and count numOver;
After the stoichiometric point of cycle calculations every day, draw decompression rate (numLower/numCount) and overvoltage rate (numOver/
NumCount), and import metering fault analytical data concentrate;
S2.2 extracts average daily defluidization/overload/current imbalance rate
In metering system data base, rated current sets, calculates starting current=rated current × 0.05;Calculate respectively
Three-phase current meansigma methods in each stoichiometric point in every day 96 stoichiometric points, and count numCount;
If three-phase current meansigma methods more than 0, judges three-phase current situation the most respectively:
If certain phase current is 0 peace, other phase current is more than starting current, it is judged that for defluidization and count numLower;
If current flow is more than 1.2 times of rated current, it is judged that for overload and count numOver;
(if the currently average three-phase current of certain phase current)/average three-phase current > 0.3, it is judged that for imbalance and count
numUnbalance;
Defluidization, overload and unbalanced ratio, i.e. (numLower+numOver is drawn after cycle calculations stoichiometric point every day
+ numUnbalance)/numCount, and ratio is imported metering fault analytical data concentration;
S2.3 extracts voltage circuit abnormal alarm number of times
By to associated data set field electricity sampling open-phase in operational database or business datum warehouse, voltage phase shortage, electricity
Push back road negative phase sequence statistics, obtain total voltage circuit abnormal alarm number of times, and be conducted into metering fault analytical data concentration;
S2.4 extracts current loop abnormal alarm number of times
By to associated data set field defluidization number of times, overload number of times in operational database or business datum warehouse,
Three-phase imbalance number of times is added up, and obtains current loop abnormal alarm number of times, and is conducted into metering fault analytical data concentration;
S2.5 extracts Pinggu, average daily peak electricity abnormal rate
Simple queries also calculates electricity=load × time, draws total electricity every day, peak electricity, ordinary telegram amount and paddy electricity,
And count numCount;Calculate | the total electricity of peak electricity |, | the total electricity of ordinary telegram amount | and | the total electricity of paddy electricity |, if wherein
Difference more than 3 degree, is then judged as Pinggu, average daily peak electricity exception and counts numException;
Pinggu, average daily peak electricity abnormal rate (numException/numCount) is drawn, by it after cycle calculations stoichiometric point
Import metering fault analytical data to concentrate;
S2.6 extracts average daily electricity ratio to abnormal rate
By associated data set in inquiry operational database or business datum depot data bank, draw just for three-phase three
The full AC sampling of terminal of the line mode of connection calculates electricity, and the total electricity calculated with step S2.5 compares, i.e. total electricity/terminal
Full AC sampling electricity;If comparison value is more than 0.05, it is judged that to exception and counts dlbdyc for electricity ratio, then draws average daily electricity
Comparison abnormal rate (dlbdyc/numCount);
S2.7 extracts average daily phase angle abnormal rate
By the Real-time Load of associated data set in inquiry operational database or business datum warehouse, calculate A, B, C tri-
The anticosine of phase voltage electric current, changes discernible integer and combines the biphase of correspondence, it is judged that the mode of connection is the most correct.If connecing
Line mode is incorrect, it is judged that for phase angle exception and count numException;Phase place is drawn after cycle calculations stoichiometric point every day
Angle abnormal rate (numException/numCount), is conducted into metering fault analytical data and concentrates.
Described S3 includes following sub-step:
S3.1 metering fault sample configuration
According to metering fault evaluation index trend analysis and combine electric power expertise instruct, configure each index relative to
The relevant flex point of S type curve, configures as follows
S3.2 builds evaluation index S type curvilinear function
Understand according to business and model is thought deeply, now take K=0.999, N=0.001;Take intermediate value in conjunction with S3.1 step y and y becomes
In stable point, draw the S type curve of different metering fault evaluation index:
S type rate of increase curve
Y=k/ (1+e-rt(k/n-1));
As it is shown in figure 1, S type rate of increase curve reflects biotic population growth trend, wherein r is natural growth, and k is one
Individual constant, n is population at individual radix, population quantity y t over time and present growth trend as depicted;But given birth to
Depositing space, the existence factor such as resource and Species Competition impact, population must be tended towards stability after reaching some;
Wherein natural growth r can be tried to achieve by Math.log ((K-N0)/N0)/mid, the value of t when mid is y=0.5;
S3.3 creates training sample set
The metering fault analytical data that S2 obtains is concentrated, and comprises extraction metering fault evaluation index from historical sample data
Stoichiometric point sequential value;The S type curvilinear function that different metering fault evaluation index market demands are corresponding in walking to S3.2, draws phase
Index result of calculation y (0.00~1.00 scope) answered;
If y < 0.50, index is normal and is tagged in corresponding data item;If y > 0.50, index is abnormal and labelling
In corresponding data item, thus create metering fault training sample set;
Described step S4, according to the required situation of reality application, now proposes two kinds of metering faults diagnosis to described S4
Model scheme:
Scheme 1
Owing to y value meets Probability Condition (nonnegativity, standardization and countable additivity), and each metering fault evaluation index
Probability of happening meets independence condition, can pass through yTotally=1-(1-yAverage daily decompression/overvoltage rate)(1-yAverage daily defluidization/overload/current imbalance rate)(1-
yAverage daily phase angle abnormal rate)(1-yAverage daily electricity ratio is to abnormal rate)(1-yAverage daily Pinggu, peak electricity abnormal rate)(1-yVoltage circuit abnormal alarm number of times)(1-yCurrent loop abnormal alarm number of times), draw meter
The amount abnormal probability of fault;
If probability is more than 0.5, then it is judged as that this user measures faulty;
If probability is less than 0.5, then it is judged as that this user metering does not has fault;
Scheme 2
Concentrate the value of each evaluation index y for input, support vector machine with the metering fault training sample that S3.3 step creates
For disaggregated model, obtain relative users metering the most faulty.
Described step S5 metering fault intelligent diagnostics, comprises following sub-step:
S5.1 extracts Real-time Load in real time from electric energy measurement automation system, day electricity, terminal alarms and main website warning number
According to;
Processing method corresponding in S5.2 invocation step 2, extracts the metering fault evaluation index of real time data;
S5.3 calling model, it is judged that the metering of each stoichiometric point user is the most faulty.
Described step S6 includes following sub-step:
S6.1 carries out manual examination and verification to the evaluation result of model, if diagnostic result is correct, then terminates, otherwise goes to S6.2;
S6.2 examines evaluation index and the training sample of measurement fault diagnostic model again closely, and adjusts metering fault diagnosis mould
The parameter of type, result is satisfied to be scheduled to last;
S6.3 preservation model.
Beneficial effect: the electric power data of magnanimity, based on electric energy measurement automation system, can automatically be entered by this method
Row intelligent diagnostics is analyzed.Metering fault diagnostic system can enter according to the environment that it is applied, up-to-date real data and system feedback
Automatically adjusting and optimizing of row model structure and parameters, reaches more preferable diagnosis effect.
This method constantly can learn, constantly with new knowledge according to the conditions for diagnostics of new data point reuse oneself
Tackle customer electricity behavior complicated and changeable.This method will become one can be greatly improved power utility check work efficiency with accurate
The sharp sword of property.
Accompanying drawing explanation
Fig. 1 is applied to the S type rate of increase curve of Real-time Load metering in a day;
Fig. 2 metering fault assessment indicator system;
Fig. 3 measurement fault diagnostic model builds;
Fig. 4 phase angle computational methods;
The electric energy metrical intelligent fault diagnosis method flow diagram based on S type curvilinear function of Fig. 5 present invention.
Detailed description of the invention
As it is shown in figure 5, the electric energy metrical intelligent fault diagnosis method based on S type curvilinear function of the present invention, including following
Step:
S1: extract Real-time Load, day electricity, terminal alarms, the historical data of main website warning from metering automation system
Specifically include following sub-step:
S1.1: determine sample coverage
December in 2009 31, whole relevant continuous datas of 31 to 2011 on Decembers;
S1.2: determine sample data scope
During sample data extraction, all large users of abnormal electricity consumption to be comprised, including big commercial power, commercial power, city
City's residential electricity consumption etc..The data attribute gathered from electric energy measurement automation system is mainly
Real-time Load: time point, stoichiometric point, A/B/C phase are meritorious, A/B/C phase current, A/B/C phase voltage and A/B/C phase merit
Rate factor;
Day electricity: time point, stoichiometric point, meritorious total, peak/flat/paddy and capacity of idle power;
Terminal alarms: electricity sampling open-phase, voltage phase shortage and voltage circuit negative phase sequence day alarm times;
Main website reports to the police: defluidization, overload and imbalance day alarm times.
S2: extract metering fault evaluation index from historical data
Specifically comprise following sub-step:
S2.1 extracts average daily decompression/overvoltage rate
Calculate the every phase voltage in three-phase and rated voltage difference respectively, count numCount simultaneously;
If difference is less than 0 and less than decompression dividing value, it is judged that for decompression and count numLower;
If difference is more than 0 and more than overvoltage dividing value, it is judged that for overvoltage and count numOver;
After cycle calculations stoichiometric point every day, draw decompression rate (numLower/numCount) and overvoltage rate (numOver/
NumCount), and import metering fault analytical data concentrate;
S2.2 extracts average daily defluidization/overload/current imbalance rate
In data base, rated current is set by associate power technical staff, and calculate starting current=rated current ×
0.05;Calculate three-phase current meansigma methods in each stoichiometric point in 96 stoichiometric points every day respectively, and count numCount;
If three-phase current meansigma methods more than 0, judges three-phase current situation the most respectively:
If certain phase current is 0 peace, other phase current is more than starting current, it is judged that for defluidization and count numLower;
If current flow is more than 1.2 times of rated current, it is judged that for overload and count numOver;
(if the currently average three-phase current of certain phase current)/average three-phase current > 0.3, it is judged that for imbalance and count
numUnbalance;
Defluidization, overload and unbalanced ratio, i.e. (numLower+numOver is drawn after cycle calculations stoichiometric point every day
+ numUnbalance)/numCount, and ratio is imported metering fault analytical data concentration;
S2.3 extracts voltage circuit abnormal alarm number of times
By to associated data set field electricity sampling open-phase in operational database or business datum warehouse, voltage phase shortage, electricity
Push back road negative phase sequence statistics, obtain total voltage circuit abnormal alarm number of times, and be conducted into metering fault analytical data concentration;
S2.4 extracts current loop abnormal alarm number of times
By to associated data set field defluidization number of times, overload number of times in operational database or business datum warehouse,
Three-phase imbalance number of times is added up, and obtains current loop abnormal alarm number of times, and is conducted into metering fault analytical data concentration;
S2.5 extracts Pinggu, average daily peak electricity abnormal rate
Simple queries also calculates electricity=load × time, draws total electricity every day, peak electricity, ordinary telegram amount and paddy electricity,
And count numCount;Calculate | the total electricity of peak electricity |, | the total electricity of ordinary telegram amount | and | the total electricity of paddy electricity |, if wherein
Difference more than 3 degree, is then judged as Pinggu, average daily peak electricity exception and counts numException;
Pinggu, average daily peak electricity abnormal rate (numException/numCount) is drawn, by it after cycle calculations stoichiometric point
Import metering fault analytical data to concentrate;
S2.6 extracts average daily electricity ratio to abnormal rate
By associated data set in inquiry operational database or business datum depot data bank, draw just for three-phase three
The full AC sampling of terminal of the line mode of connection calculates electricity, and the total electricity calculated with step S2.5 compares, i.e. total electricity/terminal
Full AC sampling electricity;If comparison value is more than 0.05, it is judged that to exception and counts dlbdyc for electricity ratio, then draws average daily electricity
Comparison abnormal rate (dlbdyc/numCount);
S2.7 extracts average daily phase angle abnormal rate
By the Real-time Load of associated data set in inquiry operational database or business datum warehouse, calculate A, B, C tri-
The anticosine of phase voltage electric current, changes discernible integer and combines the biphase of correspondence, it is judged that the mode of connection is the most correct.If
The mode of connection is incorrect, it is judged that for phase angle exception and count numException.Phase is drawn after cycle calculations stoichiometric point every day
Parallactic angle abnormal rate (numException/numCount), is conducted into metering fault analytical data and concentrates.
Seeing Fig. 2, metering fault evaluation index computational methods are summarized as follows table
S3: build training sample set based on S type curvilinear function
Specifically include following sub-step:
S3.1 metering fault sample configuration
According to metering fault evaluation index trend analysis and combine electric power expertise instruct, configure each index relative to
The relevant flex point of S type curve, configures as follows
S3.2 builds evaluation index S type curvilinear function
Understand according to business and model is thought deeply, now take K=0.999, N=0.001;Take intermediate value in conjunction with S3.1 step y and y becomes
In stable point, draw the S type curve of different metering fault evaluation index:
S type rate of increase curve
Y=k/ (1+e-rt(k/n-1));
As it is shown in figure 1, S type rate of increase curve reflects biotic population growth trend, wherein r is natural growth, and k is one
Individual constant, n is population at individual radix, population quantity y t over time and present growth trend as shown in Figure 1;But given birth to
Depositing space, the existence factor such as resource and Species Competition impact, population must be tended towards stability after reaching some;
Wherein natural growth r can be tried to achieve by Math.log ((K-N0)/N0)/mid, the value of t when mid is y=0.5;
S3.3 creates training sample set
The metering fault analytical data obtained from S2 is concentrated, and comprises extraction metering fault assessment from historical sample data and refers to
Target stoichiometric point sequential value;The S type curvilinear function that different metering fault evaluation index market demands are corresponding in walking to S3.2, draws
Corresponding index result of calculation y (0.00~1.00 scope);
If y < 0.50, index is normal and is tagged in corresponding data item;If y > 0.50, index is abnormal and labelling
In corresponding data item, thus create metering fault training sample set.
S4: build measurement fault diagnostic model
According to the required situation of reality application, it is proposed that two kinds of measurement fault diagnostic model schemes:
Scheme 1
Owing to y value meets Probability Condition (nonnegativity, standardization and countable additivity), and each metering fault evaluation index
Probability of happening meets independence condition, can pass through yTotally=1-(1-yAverage daily decompression/overvoltage rate)(1-yAverage daily defluidization/overload/current imbalance rate)(1-
yAverage daily phase angle abnormal rate)(1-yAverage daily electricity ratio is to abnormal rate)(1-yAverage daily Pinggu, peak electricity abnormal rate)(1-yVoltage circuit abnormal alarm number of times)(1-yCurrent loop abnormal alarm number of times), draw meter
The amount abnormal probability of fault;
If probability is more than 0.5, then it is judged as that this user measures faulty;
If probability is less than 0.5, then it is judged as that this user metering does not has fault;
Scheme 2
Concentrate the value of each evaluation index y for input, support vector machine with the metering fault training sample that S3.3 step creates
For disaggregated model, obtain relative users metering the most faulty;
SVM is to have vapnik to propose the earliest, is machine learning algorithm popular in current data excavation applications, has
The advantages such as structure risk is little, and generalization ability is excellent, can solve high-dimensional, nonlinear problem.Metering fault is had and well divides
Class effect.
SVM algorithm is that by a kind of nonlinear function, the sample in the input space is mapped as a high dimensional feature sky
In between, make sample linear separability in this high-dimensional feature space, and find out sample optimum linearity in this high-dimensional feature space
Optimal Separating Hyperplane.It is expressed as following formula
Wherein, yiIt is training example xiThe class value of (supporting vector);X is test case;k(x,xi) it is kernel function, αi(draw
Ge Lang multiplier) and b be the numerical parameter needing learning algorithm to determine.Find the support vector of example set and determine αiBelong to b
Limited double optimization problem in standard.The linear kernel function of kernel function used in SVM, Polynomial kernel function, radially
Base kernel function and Sigmoid kernel function, wherein Radial basis kernel function generally to be preferred over other kernel functions, so positioning last support
Vector machine discriminant is
Wherein σ is Radial basis kernel function width.
S5: metering fault intelligent diagnostics
After building measurement fault diagnostic model, it is possible to Real-time Collection continuous data also calls the model realization pair trained
User's metering fault diagnoses, and specifically comprises following sub-step:
S5.1 extracts Real-time Load in real time from electric energy measurement automation system, day electricity, terminal alarms and main website warning number
According to;
Processing method corresponding in S5.2 invocation step 2, extracts the metering fault evaluation index of real time data;
S5.3 calling model, it is judged that the metering of each stoichiometric point user is the most faulty.
S6: optimize measurement fault diagnostic model parameter, and reconstruct measurement fault diagnostic model
Specifically include following sub-step:
S6.1 carries out manual examination and verification to the evaluation result of model, if diagnostic result is correct, then terminates, otherwise goes to S6.2;
S6.2 examines evaluation index and the training sample of measurement fault diagnostic model again closely, and adjusts metering fault diagnosis mould
The parameter of type, result is satisfied to be scheduled to last;As shown in Figure 3.
S6.3 preservation model.
Claims (5)
1. an electric energy metrical intelligent fault diagnosis method based on S type curvilinear function, is characterized in that comprising the following steps:
S1: extract Real-time Load, day electricity, terminal alarms, the historical data of main website warning from electric energy measurement automation system;
S2: extract metering fault evaluation index from historical data;
S3: build training sample set based on S type curvilinear function;
S4: build measurement fault diagnostic model;
S5: metering fault intelligent diagnostics;
S6: optimize measurement fault diagnostic model parameter, and reconstruct measurement fault diagnostic model;
Described S1 includes following sub-step:
S1.1: determine sample coverage
Choose the whole relevant continuous data in sample data time range;
S1.2: determine sample data scope
Including:
Real-time Load: time point, stoichiometric point, A phase and B phase and C phase are meritorious, A phase and B phase and C phase current, A phase and B phase and C phase
Voltage, A phase and B phase and C phase power factor;
Day electricity: time point, stoichiometric point, meritorious total, peak/flat/paddy and capacity of idle power;
Terminal alarms: electricity sampling open-phase, voltage phase shortage and voltage circuit negative phase sequence day alarm times;
Main website reports to the police: defluidization, overload and imbalance day alarm times;
Described S2 comprises following sub-step:
S2.1 extracts average daily decompression/overvoltage rate
Calculate the every phase voltage in three-phase and rated voltage difference respectively, count numCount simultaneously;
If difference is less than 0 and less than decompression dividing value, it is judged that for decompression and count numLower;
If difference is more than 0 and more than overvoltage dividing value, it is judged that for overvoltage and count numOver;
After the stoichiometric point of cycle calculations every day, draw decompression rate and overvoltage rate, and import metering fault analytical data and concentrate;
S2.2 extracts average daily defluidization/overload/current imbalance rate
In metering system data base, rated current sets, calculates starting current=rated current × 0.05;Calculate every day respectively
Three-phase current meansigma methods in each stoichiometric point in 96 stoichiometric points, and count numCount;
If three-phase current meansigma methods more than 0, judges three-phase current situation the most respectively:
If certain phase current is 0, other phase current is more than starting current, it is judged that for defluidization and count numLower;
If current flow is more than 1.2 times of rated current, it is judged that for overload and count numOver;
(if currently certain phase current three-phase current meansigma methods)/three-phase current meansigma methods > 0.3, it is judged that for imbalance and count
numUnbalance;
Draw defluidization, overload and unbalanced ratio after cycle calculations stoichiometric point every day, and ratio importing metering fault is divided
In analysis data set;
S2.3 extracts voltage circuit abnormal alarm number of times
By to associated data set field electricity sampling open-phase, voltage phase shortage in operational database or business datum warehouse, voltage returns
Road negative phase sequence statistics, obtains total voltage circuit abnormal alarm number of times, and is conducted into metering fault analytical data concentration;
S2.4 extracts current loop abnormal alarm number of times
By to associated data set field defluidization number of times, overload number of times, three-phase in operational database or business datum warehouse
Uneven number of times statistics, obtains current loop abnormal alarm number of times, and is conducted into metering fault analytical data concentration;
S2.5 extracts Pinggu, average daily peak electricity abnormal rate
Simple queries also calculates electricity=load × time, draws total electricity every day, peak electricity, ordinary telegram amount and paddy electricity, and counts
Number numCount;Calculate | the total electricity of peak electricity |, | the total electricity of ordinary telegram amount | and | the total electricity of paddy electricity |, if wherein difference
More than 3 degree, then it is judged as Pinggu, average daily peak electricity exception and counts numException;
Draw Pinggu, average daily peak electricity abnormal rate after cycle calculations stoichiometric point, be conducted into metering fault analytical data and concentrate;
S2.6 extracts average daily electricity ratio to abnormal rate
By associated data set in inquiry operational database or business datum warehouse, draw just for the phase three-wire three mode of connection
The full AC sampling of terminal calculate electricity, the total electricity calculated with step S2.5 compares, i.e. total full AC sampling of electricity/terminal
Electricity;If comparison value is more than 0.05, it is judged that to exception and counts dlbdyc for electricity ratio, then show that average daily electricity ratio is to exception
Rate;
S2.7 extracts average daily phase angle abnormal rate
By the Real-time Load of associated data set in inquiry operational database or business datum warehouse, calculate A, B, C three-phase electricity
The anticosine of current voltage, changes discernible integer and combines the biphase of correspondence, it is judged that the mode of connection is the most correct;If wiring side
Formula is incorrect, it is judged that for phase angle exception and count numException;Show that phase angle is different after cycle calculations stoichiometric point every day
Often rate, is conducted into metering fault analytical data and concentrates.
Electric energy metrical intelligent fault diagnosis method based on S type curvilinear function the most according to claim 1, is characterized in that:
Described S3 includes following sub-step:
S3.1 metering fault sample configuration
Configure each index relevant flex point relative to S type curve, configure as follows
S3.2 builds evaluation index S type curvilinear function
Take K=0.999, N0=0.001;Take intermediate value in conjunction with S3.1 step y and point that y tends towards stability, show that different metering fault is commented
Estimate the S type curve of index:
S type curve
Y=K/ (1+e-rt(K/N0-1));
S type curve reflects biotic population growth trend, and wherein r is natural growth, and K is a constant, N0For population at individual base
Number, population quantity y t over time and present growth trend;But by vivosphere, existence resource and Species Competition factor shadow
Ringing, population must be tended towards stability after reaching some;
Wherein natural growth r passes through Math.log ((K-N0)/N0)/mid tries to achieve, the value of t when mid is y=0.5;
S3.3 creates training sample set
The metering fault analytical data that S2 obtains is concentrated, and comprises the meter extracting metering fault evaluation index from historical sample data
Amount point sequence value;The S type curvilinear function that different metering fault evaluation index market demands are corresponding in walking to S3.2, draws corresponding
Index result of calculation y;
If y < 0.50, index is normal and is tagged in corresponding data item;If y > 0.50, index is abnormal and is tagged to phase
In the data item answered, thus create metering fault training sample set.
Electric energy metrical intelligent fault diagnosis method based on S type curvilinear function the most according to claim 2, is characterized in that:
Described step S4 proposes two kinds of measurement fault diagnostic model:
Model 1
Owing to y value meets Probability Condition, and each metering fault evaluation index probability of happening meets independence condition, passes through yTotally
=1-(1-yAverage daily decompression/overvoltage rate)(1-yAverage daily defluidization/overload/current imbalance rate)(1-yAverage daily phase angle abnormal rate)(1-yAverage daily electricity ratio is to abnormal rate)(1-
yAverage daily Pinggu, peak electricity abnormal rate)(1-yVoltage circuit abnormal alarm number of times)(1-yCurrent loop abnormal alarm number of times), draw the abnormal probability of metering fault;
If probability is more than 0.5, then it is judged as that user measures faulty;
If probability less than 0.5, is then judged as user and measures and do not have fault;
Model 2
The value walking the metering fault training sample each evaluation index y of concentration created with S3.3 is input, and support vector machine is for dividing
Class model, obtains relative users metering the most faulty.
Electric energy metrical intelligent fault diagnosis method based on S type curvilinear function the most according to claim 3, is characterized in that:
Described step S5 comprises following sub-step:
S5.1 extracts Real-time Load in real time from electric energy measurement automation system, day electricity, terminal alarms and main website alert data;
Processing method corresponding in S5.2 invocation step 2, extracts the metering fault evaluation index of real time data;
S5.3 calling model, it is judged that the metering of each stoichiometric point user is the most faulty.
Electric energy metrical intelligent fault diagnosis method based on S type curvilinear function the most according to claim 4, is characterized in that:
Described step S6 includes following sub-step:
S6.1 carries out manual examination and verification to the evaluation result of model, if diagnostic result is correct, then terminates, otherwise goes to S6.2;
S6.2 examines evaluation index and the training sample of measurement fault diagnostic model again closely, and adjusts measurement fault diagnostic model
Parameter, result is satisfied to be scheduled to last;
S6.3 preservation model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310261411.1A CN103605103B (en) | 2013-06-26 | 2013-06-26 | Electric energy metrical intelligent fault diagnosis method based on S type curvilinear function |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310261411.1A CN103605103B (en) | 2013-06-26 | 2013-06-26 | Electric energy metrical intelligent fault diagnosis method based on S type curvilinear function |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103605103A CN103605103A (en) | 2014-02-26 |
CN103605103B true CN103605103B (en) | 2016-12-28 |
Family
ID=50123343
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310261411.1A Active CN103605103B (en) | 2013-06-26 | 2013-06-26 | Electric energy metrical intelligent fault diagnosis method based on S type curvilinear function |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103605103B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108227678A (en) * | 2018-01-04 | 2018-06-29 | 广东电网有限责任公司电力科学研究院 | A kind of power distribution network no-voltage fault diagnostic method and device based on metering automation system |
Families Citing this family (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104991222B (en) * | 2015-07-14 | 2018-01-16 | 广东电网有限责任公司佛山供电局 | metering automation terminal quality evaluation system |
CN105631762B (en) * | 2015-12-25 | 2019-10-15 | 长沙威胜信息技术有限公司 | Electric quantity compensating method based on load curve |
CN107045506A (en) * | 2016-02-05 | 2017-08-15 | 阿里巴巴集团控股有限公司 | Evaluation index acquisition methods and device |
CN107656232B (en) * | 2017-08-04 | 2020-11-10 | 国网浙江省电力公司 | On-site defect eliminating method for electric meter with no data fault under II-type concentrator |
CN107632280B (en) * | 2017-08-04 | 2020-11-10 | 国网浙江省电力公司 | On-site defect eliminating method for electric meter with no data fault under I-type concentrator |
CN107657375A (en) * | 2017-09-25 | 2018-02-02 | 国网上海市电力公司 | A kind of method for electric network fault judgement, verification and fault incidence analysis |
CN107944716A (en) * | 2017-11-29 | 2018-04-20 | 国网江苏省电力有限公司 | Based on the modified substation's electrical energy measurement cycle balance abnormality diagnostic method of learning outcome |
CN108108839B (en) * | 2017-12-18 | 2021-10-08 | 华北电力大学 | Power grid information system equipment state early warning method based on reverse fuzzy hierarchical analysis |
CN108802663B (en) * | 2018-04-18 | 2020-12-01 | 电子科技大学 | Intelligent electric energy meter function verification method based on source regulation parameter vector optimization |
CN109193563B (en) * | 2018-09-18 | 2020-09-11 | 深圳供电局有限公司 | Current loss fault monitoring method and device based on three-phase three-wire meter equipment |
CN110222991B (en) * | 2019-06-10 | 2022-08-30 | 国网江苏省电力有限公司常州供电分公司 | Metering device fault diagnosis method based on RF-GBDT |
CN110244254B (en) * | 2019-06-21 | 2021-07-20 | 上海市质子重离子医院有限公司 | Multi-stage electric energy meter error estimation and fault diagnosis method based on ratio |
CN110297207A (en) * | 2019-07-08 | 2019-10-01 | 国网上海市电力公司 | Method for diagnosing faults, system and the electronic device of intelligent electric meter |
CN110888101B (en) * | 2019-12-05 | 2022-11-22 | 新奥数能科技有限公司 | Method and device for diagnosing abnormity of electric energy meter |
CN111612019A (en) * | 2020-05-15 | 2020-09-01 | 国网河北省电力有限公司电力科学研究院 | Method for identifying and analyzing fault abnormality of intelligent electric meter based on big data model |
CN111985561B (en) * | 2020-08-19 | 2023-02-21 | 安徽蓝杰鑫信息科技有限公司 | Fault diagnosis method and system for intelligent electric meter and electronic device |
CN112098876A (en) * | 2020-08-27 | 2020-12-18 | 浙江省邮电工程建设有限公司 | Method for detecting abnormality of single battery in storage battery |
CN112798738B (en) * | 2020-12-28 | 2023-06-13 | 汉威科技集团股份有限公司 | Construction method and concentration compensation method of response model based on sensor characteristic curve |
CN113110389A (en) * | 2021-04-21 | 2021-07-13 | 东方电气自动控制工程有限公司 | Fault recording data processing method based on intelligent power plant monitoring system |
CN113538884B (en) * | 2021-05-31 | 2022-04-26 | 宁波三星医疗电气股份有限公司 | Method for improving acquisition success rate of power acquisition terminal and power acquisition terminal |
CN114527315B (en) * | 2022-02-17 | 2023-07-21 | 国网山东省电力公司营销服务中心(计量中心) | System and method for monitoring reliability of measuring equipment |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2007024825A (en) * | 2005-07-21 | 2007-02-01 | Nissan Motor Co Ltd | Fault diagnostic system for current sensor |
CN101872165A (en) * | 2010-06-13 | 2010-10-27 | 西安交通大学 | Method for fault diagnosis of wind turbines on basis of genetic neural network |
CN101975910A (en) * | 2010-09-07 | 2011-02-16 | 昆明理工大学 | Intelligent fault classification and location method for ultra-high voltage direct current transmission line |
CN102507230A (en) * | 2011-10-08 | 2012-06-20 | 中北大学 | Method for diagnosing fault of automatic ammunition supply and transportation device |
CN102866321A (en) * | 2012-08-13 | 2013-01-09 | 广东电网公司电力科学研究院 | Self-adaptive stealing-leakage prevention diagnosis method |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2447792A1 (en) * | 2005-09-19 | 2012-05-02 | Cleveland State University | Controllers, observer, and applications thereof |
-
2013
- 2013-06-26 CN CN201310261411.1A patent/CN103605103B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2007024825A (en) * | 2005-07-21 | 2007-02-01 | Nissan Motor Co Ltd | Fault diagnostic system for current sensor |
CN101872165A (en) * | 2010-06-13 | 2010-10-27 | 西安交通大学 | Method for fault diagnosis of wind turbines on basis of genetic neural network |
CN101975910A (en) * | 2010-09-07 | 2011-02-16 | 昆明理工大学 | Intelligent fault classification and location method for ultra-high voltage direct current transmission line |
CN102507230A (en) * | 2011-10-08 | 2012-06-20 | 中北大学 | Method for diagnosing fault of automatic ammunition supply and transportation device |
CN102866321A (en) * | 2012-08-13 | 2013-01-09 | 广东电网公司电力科学研究院 | Self-adaptive stealing-leakage prevention diagnosis method |
Non-Patent Citations (2)
Title |
---|
发酵过程中神经网络训练样本的选取;李运锋 等;《化工自动化及仪表》;20041231;第31卷(第6期);第21-24页 * |
训练样本的选取对网络性能的影响;孙功星 等;《核电子学与探测技术》;19961130;第16卷(第6期);第401-404页 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108227678A (en) * | 2018-01-04 | 2018-06-29 | 广东电网有限责任公司电力科学研究院 | A kind of power distribution network no-voltage fault diagnostic method and device based on metering automation system |
Also Published As
Publication number | Publication date |
---|---|
CN103605103A (en) | 2014-02-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103605103B (en) | Electric energy metrical intelligent fault diagnosis method based on S type curvilinear function | |
CN106154209B (en) | Electrical energy meter fault prediction technique based on decision Tree algorithms | |
CN103630869B (en) | Based on method of the clustering algorithm to anomalous event assay electric energy meter integrality | |
CN102866321B (en) | Self-adaptive stealing-leakage prevention diagnosis method | |
CN103135009B (en) | Electric appliance detection method and system based on user feedback information | |
CN107527114B (en) | A kind of route platform area exception analysis method based on big data | |
CN103839197A (en) | Method for judging abnormal electricity consumption behaviors of users based on EEMD method | |
CN101464964A (en) | Pattern recognition method capable of holding vectorial machine for equipment fault diagnosis | |
CN102282552A (en) | System, method and computer program for pattern based intelligent control, monitoring and automation | |
CN108052665A (en) | A kind of data cleaning method and device based on distributed platform | |
CN113032454A (en) | Interactive user power consumption abnormity monitoring and early warning management cloud platform based on cloud computing | |
CN107085653A (en) | A kind of anti-electricity-theft real-time diagnosis method of data-driven | |
CN110276511A (en) | A kind of line change relationship anomalous discrimination method based on electricity and line loss relevance | |
CN110311709A (en) | Power information acquisition system fault distinguishing method | |
CN112136161A (en) | System and method for intelligent alarm grouping | |
CN107301471A (en) | The accurate Forecasting Methodology of industrial trend and its system based on big data | |
CN103869192A (en) | Smart power grid line loss detection method and system | |
CN108037343A (en) | A kind of voltage monitoring Management System Data deepens analysis method | |
CN115730749B (en) | Power dispatching risk early warning method and device based on fusion power data | |
CN111861786A (en) | Special transformer electricity stealing identification method based on feature selection and isolated random forest | |
CN109257383A (en) | A kind of BGP method for detecting abnormality and system | |
CN106980055A (en) | A kind of student dormitory based on data mining electrical equipment violating the regulations uses monitoring system | |
CN111967618A (en) | Online diagnosis method for voltage regulator based on deep learning | |
CN109543083A (en) | The detection method of abnormal data in a kind of polynary real-time data of power grid | |
CN105354622A (en) | Fuzzy comprehensive evaluation based enterprise production management evaluation method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant |