CN116522241A - Intelligent ammeter reliability assessment method based on survival analysis - Google Patents
Intelligent ammeter reliability assessment method based on survival analysis Download PDFInfo
- Publication number
- CN116522241A CN116522241A CN202310485924.4A CN202310485924A CN116522241A CN 116522241 A CN116522241 A CN 116522241A CN 202310485924 A CN202310485924 A CN 202310485924A CN 116522241 A CN116522241 A CN 116522241A
- Authority
- CN
- China
- Prior art keywords
- electric energy
- intelligent electric
- reliability
- energy meter
- intelligent
- 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.)
- Pending
Links
- 230000004083 survival effect Effects 0.000 title claims abstract description 68
- 238000000034 method Methods 0.000 title claims abstract description 63
- 238000004458 analytical method Methods 0.000 title claims abstract description 32
- 238000012545 processing Methods 0.000 claims abstract description 19
- 230000009467 reduction Effects 0.000 claims abstract description 17
- 230000002159 abnormal effect Effects 0.000 claims abstract description 14
- 238000004364 calculation method Methods 0.000 claims description 19
- 238000010586 diagram Methods 0.000 claims description 15
- 238000004422 calculation algorithm Methods 0.000 claims description 13
- 238000011156 evaluation Methods 0.000 claims description 13
- 238000012216 screening Methods 0.000 claims description 13
- 230000001419 dependent effect Effects 0.000 claims description 9
- 238000004519 manufacturing process Methods 0.000 claims description 9
- 238000009499 grossing Methods 0.000 claims description 8
- 238000005457 optimization Methods 0.000 claims description 8
- 238000009826 distribution Methods 0.000 claims description 7
- 238000004891 communication Methods 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 6
- 238000013210 evaluation model Methods 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 3
- 238000002790 cross-validation Methods 0.000 claims description 3
- 239000006185 dispersion Substances 0.000 claims description 3
- 238000001514 detection method Methods 0.000 claims description 2
- 238000012423 maintenance Methods 0.000 abstract description 2
- 230000007704 transition Effects 0.000 abstract description 2
- 230000001186 cumulative effect Effects 0.000 description 6
- 230000000737 periodic effect Effects 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 238000012360 testing method Methods 0.000 description 6
- 238000006243 chemical reaction Methods 0.000 description 5
- 238000010276 construction Methods 0.000 description 4
- 238000007689 inspection Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000013398 bayesian method Methods 0.000 description 2
- 238000003066 decision tree Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000011425 standardization method Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005315 distribution function Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R35/00—Testing or calibrating of apparatus covered by the other groups of this subclass
- G01R35/04—Testing or calibrating of apparatus covered by the other groups of this subclass of instruments for measuring time integral of power or current
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06F18/24155—Bayesian classification
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/80—Management or planning
- Y02P90/82—Energy audits or management systems therefor
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Probability & Statistics with Applications (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an intelligent ammeter reliability assessment method based on survival analysis based on a cloud model and optimal combination weighting wind turbine generator power characteristic assessment method, which comprises the following steps of S101, ammeter original data acquisition and category characteristic numerical treatment; step S102, performing dimension reduction processing on the category characteristics; step S103, evaluating the reliability of the batch intelligent ammeter; step S104, evaluating the reliability of the intelligent ammeter individuals; the method can realize reliability analysis of the batch intelligent electric energy meters, evaluate the survival rate of each intelligent electric energy meter, help the electric power department to find abnormal conditions of the intelligent electric energy meters in time, provide theoretical support for the intelligent electric energy meters to realize transition from 'due rotation' to 'state replacement', and improve the lean operation and maintenance management level of the intelligent electric energy meters.
Description
Technical Field
The invention relates to an intelligent ammeter reliability assessment method based on survival analysis.
Background
The intelligent electric energy meter is used as one of important terminal equipment for electric power Internet of things construction, and plays a key role in links such as comprehensive sensing of power grid construction states, efficient information processing and the like. With the high-speed development of China for years, a large number of intelligent electric energy meters in operation at present reach the period of the rotation period. The DLT448-2016 technical management regulations of electric energy metering devices clearly and mainly maintains the operation reliability of the electric energy metering devices in the modes of field inspection, periodic rotation sampling inspection and the like, and the JJG596-2012 electronic alternating current electric energy meter prescribes that the verification period of electric energy meters of 1 level, 2 level and the like is generally 8 years. Due to technical progress and improvement of manufacturing technology, the actual service life of the intelligent electric energy meter can reach 10-15 years generally, a large number of intelligent electric energy meters with good performance are forcedly replaced based on JJG596-2012 verification rules, resources are wasted, power-off treatment is needed, and a lot of inconvenience is brought to customers, so that the development of reliability research on the intelligent electric energy meter is very necessary.
Reliability evaluation of intelligent electric energy meters is one of important methods for realizing the conversion from periodic rotation to state replacement of intelligent electric energy meters, and in recent years, many experts develop and research on reliability engineering of intelligent electric energy meters. At present, part of scholars research the reliability of the intelligent electric energy meter based on a component stress method, but the method depends on a reliability prediction manual, and the updating speed of the manual is generally delayed from that of the intelligent electric energy meter, so that the accuracy of a final result can be influenced; part of students analyze the service life of the intelligent electric energy meter by a method of accelerating the life test, but the test process is complex, time-consuming and labor-consuming, and the test result has errors due to the deviation of the test conditions and the real running environment;
therefore, it is necessary to design an intelligent ammeter reliability evaluation method based on survival analysis.
Disclosure of Invention
The invention aims to solve the technical problem of providing an intelligent ammeter reliability evaluation method based on survival analysis, which can realize reliability analysis of the intelligent ammeter and provide theoretical support for the conversion of the intelligent ammeter from periodic rotation to state replacement.
The technical proposal of the invention is as follows:
a reliability evaluation method of an intelligent electric energy meter based on survival analysis comprises the following steps:
step S101, acquiring the original data of the electric energy meter, carrying out category characteristic numerical processing on the original data of the intelligent electric energy meter by using a target code and a box diagram, and eliminating abnormal values, wherein the basic information, the production information and the operation information of the intelligent electric energy meter are acquired through category characteristic numerical processing;
basic information comprises voltage, current, accuracy, communication protocol and the like, production information comprises manufacturers, batches and the like, and operation information comprises operation areas, states, fault phenomena, table ages and the like;
step S102, performing dimension reduction processing on the category characteristics;
performing dimension reduction on the processed category characteristics by adopting Elastic Net-Cox, screening characteristic indexes related to the performance of the intelligent electric energy meter, and reducing the influence of redundant characteristics on the calculation efficiency and accuracy of the reliability evaluation model of the intelligent electric energy meter;
step S103, evaluating reliability of the batch intelligent ammeter
Adopting a three-parameter Weibull model to evaluate the reliability of the batch intelligent electric energy meter;
step S104, evaluating the reliability of the intelligent ammeter individual
On the basis of obtaining the reliability of the batch intelligent electric energy meters, in order to further realize accurate replacement of the intelligent electric energy meters, individual survival rate calculation is carried out on certain batches of intelligent electric energy meters with low overall reliability, a random survival forest model is taken as a basic framework, super parameters in the random survival forest are optimized by taking Bayes as an optimization algorithm, a method for optimizing the random survival forest based on the Bayes is provided for evaluating the survival rate of each intelligent electric energy meter, a survival rate curve of the intelligent electric energy meter individuals is obtained, and the reliability condition of the intelligent electric energy meter individuals is reflected accordingly.
2. The intelligent ammeter reliability evaluation method based on survival analysis according to claim 1, wherein the target coding is a coding method of class variables based on variable values of class features and corresponding dependent variables, the class features are replaced by a combination of a dependent variable target expected value given a specific class value and target expected values of dependent variables on all training data, and the target coding does not cause change of feature quantity;
X k ' Primary_prob× (1-smoove) +smoove×condition_prob (formula 1)
Wherein:
X k ' represents the code value of k in the category of the intelligent ammeter category characteristic X;
the prior probability of the target variable is represented by the prior_prob, and in python, the value is taken through the prior_prob=train_y, wherein train_y represents a tag of a training set in intelligent electric energy meter data and consists of the running time and the running state of the intelligent electric energy meter;
condition_prob represents the conditional probability of the target variable;
smoove is a reliability coefficient expressed as
Wherein n represents the number of samples with k categories in the intelligent ammeter category characteristic X, n + The sample number which indicates the class k and has positive results in the class characteristic X of the intelligent electric energy meter, wherein min_sample_leaf and smoothing are self-defined parameters, the value of min_sample_leaf is the minimum sample number when the class average value of the intelligent electric energy meter is calculated, and the default is 1, smoothsAnd (3) long is a smoothing coefficient for balancing the classified average value and the prior average value, and the larger the value is, the stronger the regularization is, and the default value is 1.
3. The method for evaluating reliability of intelligent ammeter based on survival analysis according to claim 1, wherein a box diagram is used as a statistical diagram showing data dispersion condition, and a maximum value Q of data is utilized 4 Upper quartile Q 3 Median Q 2 Lower quartile Q 1 And minimum value Q 0 These 5 statistics describe the data distribution, the principle of which is shown in FIG. 2. And detecting and eliminating abnormal values in batches by using the box line diagram by taking the running time t of the intelligent electric energy meter in the same batch as a detection object.
The upper threshold UL and the lower threshold LL are expressed as
UL＝Q 3 +1.5XIQR (equation 4)
LL＝Q 1 -1.5 XIQR (equation 5)
In the formula, IQR represents the upper quartile Q 3 And lower quartile Q 1 The difference between the two, i.e. the quartile range, is calculated as: iqr=q 3 －Q 1 And eliminating abnormal data which are larger than a threshold upper limit UL or smaller than a threshold lower limit LL from the data.
4. The method for evaluating reliability of intelligent ammeter based on survival analysis according to claim 1, wherein in step 102, the feature after the digitizing is subjected to dimension reduction processing by using Elastic Net-Cox, and variable screening of fault factors of the intelligent ammeter is completed, and the Cox model is as follows:
wherein: h represents a risk function, and t represents the running time of the intelligent electric energy meter;
X＝(X 1 ,X 2 ,…,X p ) T the intelligent electric energy meter is a covariate, and is characterized by comprising fault phenomena, constants, communication protocols, currents, sorters and electric energy meter gauges, wherein the category characteristics are obtained through data processing and feature screening in the step 1 and the step 2At least 2 of grid, wiring mode, manufacturer code, accuracy, type, metering direction, rate, manufacturing standard, model, tear-back unit, hardware version and user category;
X i ＝(X i1 ,X i2 ,…,X ip ) T p covariates for the ith individual;
h 0 (t) is a reference risk function, which takes the risk function value when all covariate values at time t are 0;
ω＝(ω 1 ,ω 2 ,…,ω p ) T is a regression coefficient; the regression coefficient is the coefficient required, and is solved through a formula 6;
parameters of the Cox model are generally calculated by using partial likelihood functions, and the partial likelihood functions of the Cox model are as follows:
wherein delta i Delta when event is deleted as an indicative function i =0, delta at event occurrence i ＝1，H i At t i Risk sets of individuals at the moment; defining an Elastic Net estimate of the Cox model by minimizing the opposite numbers of the partial log likelihood function and adding an appropriate penalty term;
in the formula, v 1 And v 2 Is a non-negative adjustment parameter, and determines upsilon by a 5-fold cross validation method 1 And v 2 The values were 0.81 and 0.05 respectively, I omega I 2 The L2 norm is represented by the number, I omega I 1 Representing the L1 norm.
Dimension reduction principle: the Elastic Net-Cox method achieves the purpose of eliminating redundant features and performing feature screening by compressing the regression coefficient omega of the redundant features to 0, thereby realizing dimension reduction.
In step S103, a three-parameter Weibull model is adopted to evaluate the reliability of the batch intelligent electric energy meter; and calculating to obtain a time node of the intelligent electric energy meter entering the loss failure period, introducing an average rank and median rank method to optimize the calculation of reliability, and solving scale parameters, shape parameters and position parameters in a three-parameter Weibull model by combining a gray estimation method, thereby obtaining quantitative results such as a reliability function, a reliability curve and the like of the batch intelligent electric energy meters to reflect the reliability of the batch intelligent electric energy meters.
The important evaluation index of the reliability of the intelligent electric energy meter is reliability, the reliability is expressed as probability of completing a specified task under specified time and conditions, and the reliability function of the three-parameter Weibull model is as follows:
wherein R is reliability; t represents the running time of the intelligent electric energy meter, beta is a scale parameter, the larger the scale parameter is, the more dispersed the scale parameter is, alpha is a shape parameter, and gamma is a position parameter;
the three parameters of the three-parameter Weibull distribution model are 3 parameters
The calculation process comprises the following steps:
assuming that T represents the life of the intelligent electric energy meter, let l=t- τ represent the remaining life of the intelligent electric energy meter after the time τ, the distribution function and probability density function of the remaining life can be expressed as:
the expectations of the remaining life of the intelligent ammeter after the moment tau are:
and modeling and solving the scale parameter beta, the shape parameter alpha and the position parameter gamma in the Weibull model by adopting a gray estimation method.
Opposite typeConversion is carried out to obtain
The reliability function is transformed to obtain:
order the
x i ＝ln(-ln(R(t i ) ) and (formula 9)
Assuming β=c, 1/α= -a, γ=b, then equation (10) is transformed into
ln is the natural logarithm underlying e.
Suppose (T) ti ,x i ) And solving the scale parameter beta, the shape parameter alpha and the position parameter gamma according to the gray model theorem in order to accord with the time sequence of the gray model. Let the running time of m intelligent electric energy meters be t 1 ,t 2 ,…,t m Its original data is T (0) ＝(t (0) (t 1 ),t (0) (t 2 ),…,t (0) (t m ) Calculated from the properties of the gray model).
From this, y and z are calculated, b=z/y, a=y, and the parameters a, b are substituted into equation (11) to calculate c, the 3 parameters of the three-parameter weibull distribution model are
The formula (15) is substituted into the formula (7) to obtain the reliability function of the intelligent electric energy meter, so that the reliability of the batch intelligent electric energy meter at each time can be known.
In step 101, in order to avoid the situation that the numerical value of a feature differs greatly or the variance of a certain feature is several orders of magnitude larger than that of other features, so that some algorithms cannot learn other features, a min-max standardization method is adopted to normalize data, the original data is transformed to map the data between [0,1], and a min-max standardization formula is as follows:
wherein max is the maximum value of a row, min is the minimum value of a row, and the new sequence y 1 ,y 2 ,…,y n Is [0,1]And a value in between.
The random living forest is composed of binary living trees, and the basic ideas are consistent with decision trees. When data passes through the splitting nodes of the tree, the splitting standard is the maximum survival difference, and the specific process of the random survival forest algorithm is as follows:
(1) Raw data was randomly extracted as sub-sample data by boottrap, and 63% of the data in the samples were set as experimental data, and 37% of the data were set as out-of-bag data (OOB).
(2) And each sub-sample grows into a binary survival tree model, at each split node, the split standard is that the survival difference is maximized, the log-rank score is adopted as the split standard, the tree model is grown when the terminal node is not smaller than the threshold set by the algorithm, and otherwise, the tree model is stopped to grow.
(3) A cumulative risk function (CHF) for each tree is calculated using the Nelson-Aalen estimate, and then an average cumulative risk function for the model may be calculated using the cumulative risk function for each tree.
(4) And (3) measuring the prediction effect of the model by adopting a consistency index (C-index) through the prediction accuracy of the out-of-bag data (OOB) test model by using the random survival forest.
Meanwhile, in order to improve the accuracy of the model, a Bayesian method is adopted to carry out super-parameter adjustment on the forest which randomly survives. The Bayesian parameter adjustment adopts a Gaussian process, the prior parameter information is considered, the prior is continuously updated, the iteration times are small, and the speed is high.
The beneficial effects are that:
the invention discloses an intelligent ammeter reliability assessment method based on survival analysis, which relates to the field of intelligent ammeter reliability engineering and comprises the following steps: step S101, detaching the intelligent meter quality analysis system from MDS of a power grid company to acquire basic information such as voltage, current, accuracy, communication protocol and the like of the intelligent electric energy meter, production information such as factories, batches and the like, and operation information such as operation areas, states, fault phenomena, meter ages and the like, and performing category characteristic numerical processing on the original data of the intelligent electric energy meter by utilizing a target code and a box diagram and eliminating abnormal values; step S102, performing dimension reduction on the processed characteristics by adopting an Elastic Net-cox, screening characteristic indexes related to the performance of the intelligent electric energy meter, and reducing the influence of redundant characteristics on the calculation efficiency and accuracy of a reliability evaluation model of the intelligent electric energy meter; step S103, evaluating the reliability of the batch intelligent electric energy meters by adopting a three-parameter Weibull model, calculating to obtain a time node of the entry loss failure period of the intelligent electric energy meters, introducing a method of average rank and median rank to optimize the calculation of the reliability, and solving scale parameters, shape parameters and position parameters in the three-parameter Weibull model by combining a gray estimation method, thereby obtaining quantitative results such as a reliability function and a reliability curve of the batch intelligent electric energy meters to reflect the reliability of the batch intelligent electric energy meters; step S104, on the basis of obtaining the reliability of the batch intelligent electric energy meters, in order to further realize accurate replacement of the intelligent electric energy meters, individual survival rate calculation is performed on the intelligent electric energy meters in certain batches, which are low in overall reliability, a random survival forest model is taken as a basic framework, super parameters in the random survival forest are optimized by taking Bayes as an optimization algorithm, a Bayes-based random survival forest optimizing method is provided, survival rate of each intelligent electric energy meter is evaluated, survival rate curves of the intelligent electric energy meters are obtained, and accordingly reliability conditions of the intelligent electric energy meters are reflected. The method can realize reliability analysis of the batch intelligent electric energy meters, evaluate the survival rate of each intelligent electric energy meter, help the electric power department to find abnormal conditions of the intelligent electric energy meters in time, provide theoretical support for the intelligent electric energy meters to realize transition from 'due rotation' to 'state replacement', and improve the lean operation and maintenance management level of the intelligent electric energy meters.
According to the intelligent electric energy meter reliability evaluation method, a reliability analysis model is built according to the running state, fault data, attribute information and other data of the intelligent electric energy meter, and the reliability of the intelligent electric energy meter is evaluated based on the data driving method, so that accurate replacement and lean management of the intelligent electric energy meter are facilitated.
According to the method, class characteristic numerical processing is carried out through a target coding method and a box diagram according to data such as running states, fault data and attribute information of the intelligent electric energy meter, abnormal values are removed, elastic Net-cox is adopted for carrying out dimension reduction processing on the processed characteristics, characteristic indexes relevant to the performance of the intelligent electric energy meter are screened, reliability of the intelligent electric energy meter is estimated through a three-parameter Weibull model, time nodes of the intelligent electric energy meter entering a loss failure period are obtained through calculation, calculation of the reliability is optimized through a method of introducing average rank and median rank, scale parameters, shape parameters and position parameters in the three-parameter Weibull model are solved through a gray estimation method, a Bayesian optimization random forest survival algorithm is adopted, accuracy of survival rate assessment of each intelligent electric energy meter is improved, an intelligent electric energy meter reliability assessment method based on survival analysis is finally provided, reliability analysis of the intelligent electric energy meter is achieved, and theoretical support is provided for conversion of the intelligent electric energy meter from periodic rotation to state replacement.
According to the invention, class characteristic numerical processing is carried out based on a target coding method and a box diagram, abnormal values are removed, the processed characteristics are subjected to dimension reduction processing by adopting an Elastic Net-cox, characteristic indexes related to the performance of the intelligent electric energy meter are screened, the reliability of the intelligent electric energy meter is estimated by utilizing a three-parameter Weibull model, a time node of the intelligent electric energy meter entering a loss failure period is obtained by calculation, the calculation of the reliability is optimized by introducing a method of average rank and median rank, the scale parameters, shape parameters and position parameters in the three-parameter Weibull model are solved by combining a gray estimation method, a Bayesian optimization random survival forest algorithm is adopted, the accuracy of survival rate estimation of each intelligent electric energy meter is improved, finally, an intelligent electric energy meter reliability estimation method based on survival analysis is provided, the reliability analysis of the intelligent electric energy meter is realized, and theoretical support is provided for the conversion of the intelligent electric energy meter from periodic rotation to state replacement.
Drawings
FIG. 1 is a flow chart of an embodiment of the present invention;
FIG. 2 is a schematic diagram of a box plot with outlier rejection
FIG. 3 is a feature after variable screening;
FIG. 4 is a reliability evaluation model of the intelligent ammeter;
FIG. 5 is a reliability evaluation curve of a batch intelligent ammeter;
FIG. 6 random survival forest network construction;
FIG. 7 is a survival curve of an individual smart power meter.
Detailed Description
The invention will be described in further detail with reference to the accompanying drawings and specific examples:
example 1:
the technical scheme of the invention will be described in further detail with reference to fig. 1-5.
The invention provides a method for accurately evaluating the reliability of an intelligent electric energy meter, and provides a method for evaluating the reliability of the intelligent electric energy meter based on survival analysis, wherein theoretical analysis and actual measurement results show that: according to the running state, fault data, attribute information and other data of the intelligent electric energy meter, category characteristic numerical processing is carried out through a target coding method and a box diagram, abnormal values are removed, elastic Net-cox dimension reduction processing is carried out on the processed characteristics, characteristic indexes relevant to the performance of the intelligent electric energy meter are screened, reliability of the intelligent electric energy meter is evaluated through a three-parameter Weibull model, a time node of the intelligent electric energy meter in a loss failure period is obtained through calculation, calculation of the reliability is optimized through introducing an average rank and a median rank method, scale parameters, shape parameters and position parameters in the three-parameter Weibull model are solved through a gray estimation method, a Bayesian optimization random survival forest algorithm is adopted, accuracy of survival rate assessment of each intelligent electric energy meter is improved, an intelligent electric energy meter reliability theory assessment method based on survival analysis is finally provided, reliability analysis of the intelligent electric energy meter is achieved, and support is provided for the intelligent electric energy meter to be converted from periodic rotation to state replacement.
Step S101, the intelligent meter quality analysis system is disassembled from the MDS of the power grid company to acquire basic information such as voltage, current, accuracy, communication protocol and the like of the intelligent electric energy meter, production information such as factories, batches and the like, and operation information such as operation areas, states, fault phenomena, meter ages and the like, and category characteristic numerical processing is carried out on the original data of the intelligent electric energy meter by utilizing a target code and a box diagram, and abnormal values are removed.
The method comprises the steps that original data and information of an intelligent electric energy meter are required to be removed, missing and messy data are required to be removed, low variance features and co-linearity features existing in the data are removed, preliminary data dimension reduction is carried out on the data, the existing intelligent electric energy meter data set comprises a plurality of types of feature data such as wiring modes, communication protocols, manufacturer codes, disassembly units, fault phenomena, manufacturing standards and intelligent electric energy meter states, the intelligent electric energy meter type features are quantized by adopting a target coding method, the target coding method is a coding method based on variable values of the type features and type variables of corresponding dependent variables, the type features are replaced by combination of dependent variable target expected values given a certain specific type value and target expected values of dependent variables on all training data, and feature quantity change cannot be caused after target coding.
X k ' Primary_prob× (1-smoove) +smoove×condition_prob (formula 1)
Wherein X is k ' represent the coding value of k in the intelligent electric energy meter class characteristic X, the priority_prob represents the prior probability of the target variable, in python, the value is taken through the priority_prob=train_y.mean (), wherein train_y represents the label of the training set in the intelligent electric energy meter data, which is composed of the running time and the state of the intelligent electric energy meter, the condition_prob represents the conditional probability of the target variable, the smoove is the reliability coefficient, and the expression is
Wherein n represents the number of samples with k categories in the intelligent ammeter category characteristic X, n + The sample number of the class k and positive result in the class characteristic X of the intelligent electric energy meter is represented, the min_sample_leaf and the smoothing are self-defined parameters, the value of the min_sample_leaf is the minimum sample number when the class average value of the intelligent electric energy meter is calculated, the default is 1, the smoothing is a smoothing coefficient used for balancing the class average value and the prior average value, the larger the value is, the stronger the regularization is, and the default value is 1.
In order to avoid the situation that the numerical value of a feature is larger in difference or the variance of a feature is larger by several orders of magnitude than other features, so that some algorithms cannot learn other features, a min-max standardization method is adopted to normalize data, the original data is transformed to map the data between [0,1], and a min-max standardization formula is as follows:
wherein max is the maximum value of a row, min is the minimum value of a row, and the new sequence y 1 ,y 2 ,…,y n Is [0,1]And a value in between.
The box diagram is used as a statistical diagram for displaying the data dispersion condition, and the maximum value Q of the data is utilized 4 Upper quartile Q 3 Median Q 2 Lower quartile Q 1 And minimum value Q 0 These 5 statistics describe the data distribution, considering outliers as extreme values in the data sequence, the principle of which is shown in FIG. 2.
The upper threshold UL and the lower threshold LL are expressed as
UL＝Q 3 +1.5XIQR (equation 5)
LL＝Q 1 -1.5 XIQR (equation 6)
Abnormal data which is larger than a threshold upper limit UL or smaller than a threshold lower limit LL in the data are rejected.
Step S102, performing dimension reduction on the processed characteristics by adopting Elastic Net-cox, screening characteristic indexes related to the performance of the intelligent electric energy meter, and reducing the influence of redundant characteristics on the calculation efficiency and accuracy of the reliability evaluation model of the intelligent electric energy meter.
And performing dimension reduction treatment on the treated characteristics by adopting Elastic Net-Cox to finish variable screening of fault factors of the intelligent electric energy meter, wherein the Cox model is as follows:
where h represents a risk function, x= (X) 1 ,X 2 ,…,X p ) T X is covariate, X i ＝(X i1 ,X i2 ,…,X ip ) T P covariates for the ith individual, h 0 (t) is a reference risk function, which takes the risk function value, ω= (ω), when all the covariates take the value 0 at time t 1 ,ω 2 ,…,ω p ) T Is regression toCoefficients.
Parameters of the Cox model are generally calculated by using partial likelihood functions, and the partial likelihood functions of the Cox model are as follows:
wherein delta i Delta when event is deleted as an indicative function i =0, delta at event occurrence i ＝1，H i At t i Risk set of individuals at the moment.
Minimizing the opposite numbers of the partial log likelihood function and adding the appropriate penalty term defines the Elastic Net estimate of the Cox model.
In the formula, v 1 And v 2 Is a non-negative adjustment parameter, and determines upsilon by a 5-fold cross validation method 1 And v 2 The values were 0.81 and 0.05, respectively.
After the dimension reduction treatment is carried out by adopting the Elastic Net-cox, 17 characteristics are obtained by screening from 38 characteristics, and the characteristic coefficient value after variable screening is shown in figure 3.
Step S103, the reliability of the batch intelligent electric energy meters is evaluated by adopting a three-parameter Weibull model, a time node of the entry loss failure period of the intelligent electric energy meters is obtained by calculation, the calculation of the reliability is optimized by introducing a method of average rank and median rank, and the scale parameter, the shape parameter and the position parameter in the three-parameter Weibull model are solved by combining a gray estimation method, so that the reliability of the batch intelligent electric energy meters is reflected by quantitative results such as reliability functions, reliability curves and the like of the batch intelligent electric energy meters.
The important evaluation index of the reliability of the intelligent electric energy meter is reliability, the reliability represents the probability of completing a specified task under specified time and conditions, and the reliability function of the three-parameter Weibull model is as follows:
wherein t represents the running time of the intelligent electric energy meter, beta is a scale parameter, the larger the scale parameter is, the more dispersed the scale parameter is, alpha is a shape parameter, and gamma is a position parameter.
The reliability function is transformed to obtain:
order the
x i ＝In(-In(R(t i ) ) and (equation 13)
Assuming β=c, 1/α= -a, γ=b, then equation (14) is transformed into
Suppose (T) ti ,x i ) And solving the scale parameter beta, the shape parameter alpha and the position parameter gamma according to the gray model theorem in order to accord with the time sequence of the gray model. Let the running time of m intelligent electric energy meters be t 1 ,t 2 ,…,t m Its original data is T (0) ＝(t (0) (t 1 ),t (0) (t 2 ),…,t (0) (t m ) Calculated from the properties of the gray model).
From this, y and z are calculated, b=z/y, a=y, and the parameters a, b are substituted into equation (15) to calculate c, the 3 parameters of the three-parameter weibull distribution model are
The formula (19) is substituted into the formula (11) to obtain the reliability function of the intelligent electric energy meter, so that the reliability of the batch intelligent electric energy meter at each time can be known.
According to the above analysis, the proposed weibull model based on gray-three parameters is shown in fig. 4. According to the method, the cleaned intelligent ammeter age data are analyzed, and the scale parameter beta, the shape parameter alpha and the position parameter gamma are calculated by using a gray estimation method and are shown in the table 1.
Table 1 parameter values calculated by each method
And obtaining the reliability curve of the batch intelligent electric energy meter according to the result of the scale parameter beta, the shape parameter alpha and the position parameter gamma obtained in the table 1 and the established Weibull model for reliability evaluation, wherein the reliability curve is shown in figure 5.
Step S104, on the basis of obtaining the reliability of the batch intelligent electric energy meters, in order to further realize accurate replacement of the intelligent electric energy meters, individual survival rate calculation is performed on the intelligent electric energy meters in certain batches, which are low in overall reliability, a random survival forest model is taken as a basic framework, super parameters in the random survival forest are optimized by taking Bayes as an optimization algorithm, a Bayes-based random survival forest optimizing method is provided, survival rate of each intelligent electric energy meter is evaluated, survival rate curves of the intelligent electric energy meters are obtained, and accordingly reliability conditions of the intelligent electric energy meters are reflected.
The random living forest is composed of binary living trees, and the basic ideas are consistent with decision trees. When data passes through the splitting nodes of the tree, the splitting standard is the maximum survival difference, and the specific process of the random survival forest algorithm is as follows:
(1) Raw data was randomly extracted as sub-sample data by boottrap, and 63% of the data in the samples were set as experimental data, and 37% of the data were set as out-of-bag data (OOB).
(2) And each sub-sample grows into a binary survival tree model, at each split node, the split standard is that the survival difference is maximized, the log-rank score is adopted as the split standard, the tree model is grown when the terminal node is not smaller than the threshold set by the algorithm, and otherwise, the tree model is stopped to grow.
(3) A cumulative risk function (CHF) for each tree is calculated using the Nelson-Aalen estimate, and then an average cumulative risk function for the model may be calculated using the cumulative risk function for each tree.
(4) And (3) measuring the prediction effect of the model by adopting a consistency index (C-index) through the prediction accuracy of the out-of-bag data (OOB) test model by using the random survival forest.
According to the above analysis, the proposed random living forest network construction framework is shown in fig. 6.
Meanwhile, in order to improve the accuracy of the model, a Bayesian method is adopted to carry out super-parameter adjustment on the forest which randomly survives. The Bayesian parameter adjustment adopts a Gaussian process, the prior parameter information is considered, the prior is continuously updated, the iteration times are small, and the speed is high. The search range and the optimal value of each parameter are shown in the following table.
TABLE 2 optimization and selection of structural parameters
After determining parameters of a random survival forest model, randomly selecting 2 intelligent electric energy meter input models, obtaining survival curves of the 2 intelligent electric energy meters as shown in fig. 7, wherein the running performance and reliability of different intelligent electric energy meters at the early stage of installation are not greatly different, but the survival rate curves of different intelligent electric energy meters are different along with the increase of running time, when the number 0 intelligent electric energy meter is operated for 1800 days along with the increase of running time, the accumulated risk starts to rise gradually, and the survival rate is reduced to be lower than 0.1 in 2000 days of operation, so that the number 0 intelligent electric energy meter is indicated to fail at the moment of high probability during the period, and the intelligent electric energy meter is replaced timely. In the actual running condition of the intelligent electric energy meter, the manual inspection data shows that the No. 0 intelligent electric energy meter fails in 1930 days, and the three meters, namely No. 1, no. 2 and No. 3 intelligent electric energy meters, still do not fail in 1939 days, 1898 days and 2020 days, and are right deleted data.
While the invention has been disclosed in terms of preferred embodiments, the embodiments are not limiting of the invention. Any equivalent changes or modifications can be made without departing from the spirit and scope of the present invention, and are intended to be within the scope of the present invention. The scope of the invention should therefore be determined by the following claims.
Claims (5)
1. The intelligent ammeter reliability evaluation method based on survival analysis is characterized by comprising the following steps of:
step S101, collecting original data of electric energy meter and performing category characteristic numerical processing
Basic information, production information and operation information of the intelligent electric energy meter are collected, category characteristic numerical processing is carried out on the original data of the intelligent electric energy meter by utilizing a target code and a box diagram, and abnormal values are removed;
step S102, performing dimension reduction processing on the category characteristics;
performing dimension reduction on the processed category characteristics by adopting Elastic Net-Cox, screening characteristic indexes related to the performance of the intelligent electric energy meter, and reducing the influence of redundant characteristics on the calculation efficiency and accuracy of the reliability evaluation model of the intelligent electric energy meter;
step S103, evaluating reliability of the batch intelligent ammeter
Adopting a three-parameter Weibull model to evaluate the reliability of the batch intelligent electric energy meter;
step S104, evaluating the reliability of the intelligent ammeter individual
On the basis of obtaining the reliability of the batch intelligent electric energy meters, in order to further realize accurate replacement of the intelligent electric energy meters, individual survival rate calculation is carried out on certain batches of intelligent electric energy meters with low overall reliability, a random survival forest model is taken as a basic framework, super parameters in the random survival forest are optimized by taking Bayes as an optimization algorithm, a method for optimizing the random survival forest based on the Bayes is provided for evaluating the survival rate of each intelligent electric energy meter, a survival rate curve of the intelligent electric energy meter individuals is obtained, and the reliability condition of the intelligent electric energy meter individuals is reflected accordingly.
2. The intelligent ammeter reliability evaluation method based on survival analysis according to claim 1, wherein the target coding is a coding method of class variables based on variable values of class features and corresponding dependent variables, the class features are replaced by a combination of a dependent variable target expected value given a specific class value and target expected values of dependent variables on all training data, and the target coding does not cause change of feature quantity;
X k ' Primary_prob× (1-smoove) +smoove×condition_prob (formula 1)
Wherein:
X k ' represents the code value of k in the category of the intelligent ammeter category characteristic X;
the prior probability of the target variable is represented by the prior_prob, and in python, the value is taken through the prior_prob=train_y, wherein train_y represents a tag of a training set in intelligent electric energy meter data and consists of the running time and the running state of the intelligent electric energy meter;
condition_prob represents the conditional probability of the target variable;
smoove is a reliability coefficient expressed as
Wherein n represents the number of samples with k categories in the intelligent ammeter category characteristic X, n + The sample number of the class k and positive result in the class characteristic X of the intelligent electric energy meter is represented, the min_sample_leaf and the smoothing are self-defined parameters, the value of the min_sample_leaf is the minimum sample number when the class average value of the intelligent electric energy meter is calculated, the default is 1, the smoothing is a smoothing coefficient used for balancing the class average value and the prior average value, the larger the value is, the stronger the regularization is, and the default value is 1.
3. The method for evaluating reliability of intelligent ammeter based on survival analysis according to claim 1, wherein a box diagram is used as a statistical diagram showing data dispersion condition, and a maximum value Q of data is utilized 4 Upper quartile Q 3 Median Q 2 Lower quartile Q 1 And minimum value Q 0 The 5 statistics describe data distribution, the running time t of the intelligent electric energy meter in the same batch is taken as a detection object, and abnormal values are detected and removed in batches by utilizing the box line graph.
The upper threshold UL and the lower threshold LL are expressed as
UL＝Q 3 +1.5XIQR (equation 4)
LL＝Q 1 -1.5 XIQR (equation 5)
In the formula, IQR represents the upper quartile Q 3 And lower quartile Q 1 The difference between the two, i.e. the quartile range, is calculated as: iqr=q 3 －Q 1 And eliminating abnormal data which are larger than a threshold upper limit UL or smaller than a threshold lower limit LL from the data.
4. The method for evaluating reliability of intelligent ammeter based on survival analysis according to claim 1, wherein in step 102, the feature after the digitizing is subjected to dimension reduction processing by using Elastic Net-Cox, and variable screening of fault factors of the intelligent ammeter is completed, and the Cox model is as follows:
wherein: h represents a risk function, and t represents the running time of the intelligent electric energy meter;
X＝(X 1 ,X 2 ,…,X p ) T the category characteristics of the intelligent electric energy meter are obtained through data processing and feature screening in the step 1 and the step 2, wherein the category characteristics comprise at least 2 of fault phenomena, constants, communication protocols, currents, sorters, electric energy meter specifications, wiring modes, manufacturer codes, accuracy, types, metering directions, rates, manufacturing standards, models, dismantling units, hardware versions and user categories;
X i ＝(X i1 ,X i2 ,…,X ip ) T p covariates for the ith individual;
h 0 (t) is a reference risk function, which takes the risk function value when all covariate values at time t are 0;
ω＝(ω 1 ,ω 2 ,…,ω p ) T is a regression coefficient;
parameters of the Cox model are generally calculated by using partial likelihood functions, and the partial likelihood functions of the Cox model are as follows:
wherein delta i Delta when event is deleted as an indicative function i =0, delta at event occurrence i ＝1，H i At t i Risk sets of individuals at the moment; defining an Elastic Net estimate of the Cox model by minimizing the opposite numbers of the partial log likelihood function and adding an appropriate penalty term;
in the formula, v 1 And v 2 Is a non-negative adjustment parameter, and determines upsilon by a 5-fold cross validation method 1 And v 2 The values were 0.81 and 0.05 respectively, I omega I 2 The L2 norm is represented by the number, I omega I 1 Representing the L1 norm.
5. The method for evaluating the reliability of the intelligent ammeter based on survival analysis according to claim 1, wherein in step S103, the reliability of the batch intelligent ammeter is evaluated by adopting a three-parameter weibull model;
the important evaluation index of the reliability of the intelligent electric energy meter is reliability, the reliability is expressed as probability of completing a specified task under specified time and conditions, and the reliability function of the three-parameter Weibull model is as follows:
wherein R is reliability; t represents the running time of the intelligent electric energy meter, beta is a scale parameter, the larger the scale parameter is, the more dispersed the scale parameter is, alpha is a shape parameter, and gamma is a position parameter;
the three parameters of the three-parameter Weibull distribution model are 3 parameters
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310485924.4A CN116522241A (en) | 2023-04-28 | 2023-04-28 | Intelligent ammeter reliability assessment method based on survival analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310485924.4A CN116522241A (en) | 2023-04-28 | 2023-04-28 | Intelligent ammeter reliability assessment method based on survival analysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116522241A true CN116522241A (en) | 2023-08-01 |
Family
ID=87395507
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310485924.4A Pending CN116522241A (en) | 2023-04-28 | 2023-04-28 | Intelligent ammeter reliability assessment method based on survival analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116522241A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118209905A (en) * | 2024-05-20 | 2024-06-18 | 国网山西省电力公司晋城供电公司 | Online fault prediction method for distribution transformer based on Internet of things perception |
-
2023
- 2023-04-28 CN CN202310485924.4A patent/CN116522241A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118209905A (en) * | 2024-05-20 | 2024-06-18 | 国网山西省电力公司晋城供电公司 | Online fault prediction method for distribution transformer based on Internet of things perception |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111709490B (en) | Fan health state assessment method based on GRU neural network | |
| CN116757534B (en) | Intelligent refrigerator reliability analysis method based on neural training network | |
| CN109389145B (en) | Electric energy meter manufacturer evaluation method based on metering big data clustering model | |
| CN110232203B (en) | Knowledge distillation optimization RNN short-term power failure prediction method, storage medium and equipment | |
| CN111105332A (en) | Highway intelligent pre-maintenance method and system based on artificial neural network | |
| CN111813084A (en) | Mechanical equipment fault diagnosis method based on deep learning | |
| CN113126019B (en) | Remote estimation method, system, terminal and storage medium for error of intelligent ammeter | |
| CN111401599A (en) | Water level prediction method based on similarity search and L STM neural network | |
| CN113792939B (en) | Electric energy meter reliability prediction method and device based on mixed Weibull distribution | |
| CN116150897A (en) | Machine tool spindle performance evaluation method and system based on digital twin | |
| CN111178585A (en) | Fault reporting amount prediction method based on multi-algorithm model fusion | |
| CN110826237B (en) | Wind power equipment reliability analysis method and device based on Bayesian belief network | |
| CN112329990A (en) | User power load prediction method based on LSTM-BP neural network | |
| CN115438726A (en) | Device life and fault type prediction method and system based on digital twin technology | |
| CN113408659A (en) | Building energy consumption integrated analysis method based on data mining | |
| CN116522241A (en) | Intelligent ammeter reliability assessment method based on survival analysis | |
| CN114065605A (en) | Intelligent electric energy meter running state detection and evaluation system and method | |
| CN110705859A (en) | PCA-self-organizing neural network-based method for evaluating running state of medium and low voltage distribution network | |
| CN115034485A (en) | Wind power interval prediction method and device based on data space | |
| CN114912716A (en) | Short-term power load prediction method and system based on bidirectional long-term and short-term memory network | |
| CN114154716B (en) | Enterprise energy consumption prediction method and device based on graph neural network | |
| CN116467658A (en) | Equipment fault tracing method based on Markov chain | |
| CN110781206A (en) | Method for predicting whether electric energy meter in operation fails or not by learning meter-dismantling and returning failure characteristic rule | |
| CN115096357A (en) | Indoor environment quality prediction method based on CEEMDAN-PCA-LSTM | |
| CN113762591A (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM counterstudy |
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 |