CN112380759A - Smart meter service life prediction method based on deep learning and CoxPH model - Google Patents
Smart meter service life prediction method based on deep learning and CoxPH model Download PDFInfo
- Publication number
- CN112380759A CN112380759A CN201911185780.0A CN201911185780A CN112380759A CN 112380759 A CN112380759 A CN 112380759A CN 201911185780 A CN201911185780 A CN 201911185780A CN 112380759 A CN112380759 A CN 112380759A
- Authority
- CN
- China
- Prior art keywords
- meter
- ammeter
- abnormal
- electric meter
- life
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 70
- 238000013135 deep learning Methods 0.000 title claims abstract description 29
- 230000002159 abnormal effect Effects 0.000 claims abstract description 117
- 230000004083 survival effect Effects 0.000 claims abstract description 107
- 230000006870 function Effects 0.000 claims abstract description 82
- 238000012549 training Methods 0.000 claims abstract description 51
- 238000012217 deletion Methods 0.000 claims abstract description 45
- 230000037430 deletion Effects 0.000 claims abstract description 45
- 230000005856 abnormality Effects 0.000 claims description 37
- 230000005611 electricity Effects 0.000 claims description 28
- 238000012423 maintenance Methods 0.000 claims description 22
- 238000013528 artificial neural network Methods 0.000 claims description 14
- 238000004590 computer program Methods 0.000 claims description 12
- 238000003860 storage Methods 0.000 claims description 8
- 150000001875 compounds Chemical class 0.000 claims 1
- 230000003068 static effect Effects 0.000 abstract description 7
- 238000009826 distribution Methods 0.000 description 13
- 238000004458 analytical method Methods 0.000 description 10
- 238000004891 communication Methods 0.000 description 10
- 230000008569 process Effects 0.000 description 9
- 238000009434 installation Methods 0.000 description 8
- 230000006399 behavior Effects 0.000 description 5
- 238000010801 machine learning Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000012795 verification Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 230000002301 combined effect Effects 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000006386 memory function Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- 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
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Software Systems (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Health & Medical Sciences (AREA)
- Educational Administration (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The embodiment of the invention provides a smart meter service life prediction method based on deep learning and a CoxPH model, which comprises the following steps: inputting abnormal data of an ammeter to be predicted into an ammeter life prediction model, and outputting an ammeter survival curve corresponding to the abnormal data of the ammeter to be predicted, wherein the ammeter survival curve is a curve of the relationship between the survival probability of the ammeter to be predicted and time; the electric meter life prediction model is obtained by training based on electric meter abnormal sample data and a predetermined electric meter life label and a deletion label, and a loss function in the electric meter life prediction model training is formed by participation of a logarithmic part risk function in a CoxPH model; and predicting the service life of the electric meter to be predicted based on the electric meter survival curve and a preset survival probability threshold value. The method provided by the embodiment of the invention avoids the problem that the service life prediction model of the intelligent ammeter in the prior art is too static, and improves the service life prediction reliability of the intelligent ammeter.
Description
Technical Field
The invention relates to the technical field of machine learning, in particular to a smart meter service life prediction method based on deep learning and a CoxPH model.
Background
The smart electric meter is used as a key metering device in a modern electric power system, and the high reliability of the smart electric meter is an important guarantee for normal maintenance and operation of a power grid system. With the access of massive intelligent electric meters in the power distribution network, multi-source intelligent meter big data is formed, and the multi-source intelligent meter big data contains abundant user energy consumption and equipment operation and maintenance information. And the operation and maintenance of the intelligent electric meters in the power distribution network always have challenges. Since there is currently no reasonably proactive approach to assessing meter reliability, the grid operator can only take the strategy of expiring total replacement by making a uniform restriction on the meter's run time. Such maintenance is undoubtedly wasteful, as a large amount of potentially meter uptime is a margin to ensure stable operation of the system. In parallel, an operation and maintenance mode is that a field operation and maintenance technician finds that a meter has a fault during regular maintenance, and reports the fault to be replaced or maintained. Such a passive operation and maintenance strategy faces practical challenges of huge installation quantity of electric meters, occasional faults, reduction of service quality of the power consumption end before the faults are discovered, and the like.
Researchers have done a series of studies on the analysis of electric meter big data to date. For example, a method for establishing a mathematical model of a functional module of the electric energy meter is proposed, stress levels are reasonably estimated by inquiring a component failure manual, and failure rates of the functional modules are calculated, so that the service life and reliability analysis of the intelligent electric meter is completed. A method for pre-assuming that the survival function of the electric meter conforms to Weibull distribution is proposed, so that parameters of the Weibull distribution are calculated by utilizing maximum likelihood estimation after actual production data and operation data of the electric meter are divided into batches; and obtaining expected life nodes according to the calculated parameters and a preset survival rate threshold value so as to establish an early warning mechanism, or adopting a Cox linear model in a survival analysis theory to carry out fitting on a survival function.
However, the method for estimating the reliability of the failure rate of the components is lack of combination with actual metering statistical data, and cannot reflect the complex physical characteristics and the operating environment of the electric meter; the formed model is static and cannot be adjusted and adapted according to the change of the operating environment and the physical information; the proposed assumptions are not reasonable. Some researches are established on the basis of stronger hypothesis, such as a highly simplified model, only partial functional modules are considered, and physical reality is separated; or the survival function is assumed to conform to a certain specific function distribution form; alternatively, it is assumed that the combined effect of covariates can be expressed as a linear combination of values of the covariates. These assumptions are so strong that it is difficult to characterize the physical characteristics and operating environment of a real electricity meter, and finally, research is in a form; the method is based on a basic Cox linear model method, so that covariate types which are possibly influenced need to be manually screened in advance when fitting calculation is carried out, the process depends on manpower analysis and experience, and the obtained result is seriously influenced by the screening working quality.
Therefore, how to avoid the formed prediction model of the service life of the smart meter from being too static and improve the reliability of the service life prediction of the smart meter is still a problem to be solved by those skilled in the art.
Disclosure of Invention
The embodiment of the invention provides a smart meter service life prediction method based on deep learning and a CoxPH model, which is used for solving the problem that the service life prediction reliability of the smart meter is low in the conventional smart meter service life prediction model.
In a first aspect, an embodiment of the present invention provides a smart meter life prediction method based on deep learning and a cox ph model, including:
inputting abnormal data of an ammeter to be predicted into an ammeter life prediction model, and outputting an ammeter survival curve corresponding to the abnormal data of the ammeter to be predicted, wherein the ammeter survival curve is a curve of the relationship between the survival probability of the ammeter to be predicted and time;
the electric meter life prediction model is obtained by training based on electric meter abnormal sample data and a predetermined electric meter life label and a deletion label, and a loss function in the electric meter life prediction model training is formed by participation of a logarithmic part risk function in a CoxPH model;
and predicting the service life of the electric meter to be predicted based on the electric meter survival curve and a preset survival probability threshold value.
Preferably, in the method, the ammeter abnormal sample data is obtained by sorting the collected overhaul data of the ammeter and the abnormal alarm historical record of the ammeter.
Preferably, in the method, the collected maintenance data of the electric meter and the collected abnormal alarm history of the electric meter are sorted to obtain the abnormal sample data of the electric meter, and the method specifically includes:
firstly, the overhaul data and the abnormal alarm historical record of the electric meter are described in the form of the following formula:
In the above-mentioned formula,is the abnormal type of the t-th meter abnormality occurred in the meter of the meter i,is the existing time when the electricity meter of the electricity meter i has the abnormality for the t-th time, d(i)Is the life of the meter, i(i)Deletion tag, Meta, of a meter that is meter i(i)Is the physical information of the meter, T, for meter i(i)The number of abnormal records in an observation period is shown, and omega represents a set of the electric meters of the collected maintenance data and the abnormal alarm historical records;
forming a list of data described by the formula, wherein a main key is the number of the ammeter, an ith row corresponds to abnormal data of the ammeter i, and the list comprises an abnormal column, a deletion column and a life column, wherein the abnormal value of the mth column is a normalized value of the number of times of the mth type of abnormality of the ammeter i observed in an observation period, the deletion column takes a value of 0 or 1, the value of 0 represents that the ammeter has a fault, the value of 1 represents that the ammeter is still alive after the observation time, and the value of the life column is the life of the ammeter i.
Preferably, in the method, determining a normalized numerical value of the number of m-th type of abnormality occurring to the electric meter of the electric meter i observed in the observation period specifically includes:
the actual number of m-th type of abnormality of the electric meter i isThe normalized number of the m-th type abnormity of the ammeter i ismin (m) and max (m) are the minimum and maximum values, respectively, for the m-th column anomaly,
and determining the normalized numerical value of the number of m-th type abnormality of the electric meter i observed in the observation period based on the formula.
Preferably, in the method, the formula of the loss function l (θ) is as follows:
wherein h isθ(x) Representing the logarithmic partial risk function, x representing a physical information parameter vector of the electricity meter, xiPhysical information parameter vector, x, of electric meter representing electric meter ijRepresenting a physical information parameter vector of meter j, theta represents a parameter of a neural network training the logarithmic partial risk function, EiIs a deletion, EiThe meter i, denoted 1, survived until the observation time, TiRefers to the life of the meter, R (T), for meter ii) Is TiSet of meters that remain alive over time, j being R (T)i) The meters j in the set.
In a second aspect, an embodiment of the present invention provides an apparatus for predicting a lifetime of a smart meter based on deep learning and a cox ph model, including:
the device comprises a survival curve determining unit, a life prediction unit and a life prediction unit, wherein the survival curve determining unit is used for inputting abnormal data of the ammeter to be predicted into an ammeter life prediction model and outputting an ammeter survival curve corresponding to the abnormal data of the ammeter to be predicted, and the ammeter survival curve is a curve of the relationship between the survival probability of the ammeter to be predicted and time; the electric meter life prediction model is obtained by training based on electric meter abnormal sample data and a predetermined electric meter life label and a deletion label, and a loss function in the electric meter life prediction model training is formed by participation of a logarithmic part risk function in a CoxPH model;
and the service life prediction unit is used for predicting the service life of the electric meter to be predicted based on the electric meter survival curve and a preset survival probability threshold value.
Preferably, in the device, the ammeter abnormal sample data is obtained by sorting the collected ammeter overhaul data and ammeter abnormal alarm history records.
Preferably, in the device, the sorting of the collected maintenance data of the electric meter and the abnormal alarm history of the electric meter to obtain the abnormal sample data of the electric meter specifically includes:
firstly, the overhaul data and the abnormal alarm historical record of the electric meter are described in the form of the following formula:
In the above-mentioned formula,is the abnormal type of the t-th meter abnormality occurred in the meter of the meter i,is the existing time when the electricity meter of the electricity meter i has the abnormality for the t-th time, d(i)Is the life of the meter, i(i)Deletion tag, Meta, of a meter that is meter i(i)Is the physical information of the meter, T, for meter i(i)The number of abnormal records in an observation period is shown, and omega represents a set of the electric meters of the collected maintenance data and the abnormal alarm historical records;
forming a list of data described by the formula, wherein a main key is the number of the ammeter, an ith row corresponds to abnormal data of the ammeter i, and the list comprises an abnormal column, a deletion column and a life column, wherein the abnormal value of the mth column is a normalized value of the number of times of the mth type of abnormality of the ammeter i observed in an observation period, the deletion column takes a value of 0 or 1, the value of 0 represents that the ammeter has a fault, the value of 1 represents that the ammeter is still alive after the observation time, and the value of the life column is the life of the ammeter i.
In a third aspect, an embodiment of the present invention provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor executes the computer program to implement the steps of the method for predicting the lifetime of a smart meter based on deep learning and a cox ph model according to the first aspect.
In a fourth aspect, an embodiment of the present invention provides a non-transitory computer readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the steps of the smart meter life prediction method based on deep learning and a cox ph model as provided in the first aspect.
The intelligent ammeter service life prediction method based on deep learning and the CoxPH model, provided by the embodiment of the invention, inputs abnormal data of an ammeter to be predicted into an ammeter service life prediction model, and outputs an ammeter survival curve corresponding to the abnormal data of the ammeter to be predicted, wherein the ammeter survival curve is a curve of a relationship between survival probability and time of the ammeter to be predicted; the electric meter life prediction model is obtained after training is carried out on the basis of electric meter abnormal sample data and a predetermined electric meter life label and a deletion label, a loss function in the electric meter life prediction model training process is formed by participation of a logarithm part risk function in a CoxPH model, the electric meter life prediction model for intelligent electric meter life prediction is dynamically generated through deep learning, and the electric meter life prediction model can be retrained through updating the training sample and the label to obtain a more accurate prediction model. Therefore, the method provided by the embodiment of the invention avoids the situation that the service life prediction model of the intelligent ammeter in the prior art is too static, and improves the service life prediction reliability of the intelligent ammeter.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, a brief description will be given below of the drawings required for the embodiments or the technical solutions in the prior art, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
Fig. 1 is a schematic flowchart of a method for predicting the service life of an intelligent electric meter based on deep learning and a cox ph model according to an embodiment of the present invention;
FIG. 2 is a schematic flowchart of a method for training an electricity meter life prediction model based on machine learning and a CoxPH model according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of an intelligent electric meter life prediction device based on deep learning and a cox ph model according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without inventive effort based on the embodiments of the present invention, are within the scope of the present invention.
The CoxPH model-based intelligent electric meter life prediction method is based on a survival analysis theory, the survival analysis is a statistical analysis method for researching a distribution rule of survival time and a relation between the survival time and relevant factors, and the method is widely applied to the fields of patient life analysis, fault-time analysis of machine equipment and the like. First, some basic terms are introduced in combination with a smart meter application scenario:
event (Event): in the service life analysis of the electric meter, the electric meter is replaced after the electric meter is failed/reaches the specified maximum service time.
Survival Time (survivval Time): the time duration from the time of installation to the time of dismounting the electric meter after the electric meter is determined to have a fault/cause to reach the maximum specified use time.
Deletion (Censoring) refers to the situation where the time of the occurrence of the electricity lasts until the last observation time node, and the event has not yet occurred. In a broad sense, three conditions of left deletion, right deletion and interval deletion exist, but for the intelligent electric meter, only the right deletion is considered, namely the actual survival time of the electric meter is longer than the observed time.
Covariate (Covariate) a variable factor that affects the time of occurrence of an event, such as for electricity meters, the manufacturer or the number of abnormal alarms.
Survival Function (survivval Function) s (t): the Probability that the survival time of an individual exceeds T is defined as s (T) Probability (T > T).
Survival Curve (Survival Curve): the survival rates at each time point were connected to form a curve. Generally, the x-axis represents time-to-live and the y-axis represents probability-to-live.
Risk Function (Hazard Function) λ (t): the instantaneous probability of death is characterized and defined as follows:
Classical methods of survival analysis can be divided into three categories: parametric, nonparametric and semiparametric methods. The parameter method needs to assume or determine a distribution model of the survival time in advance, then estimates model parameters according to data, and finally calculates the survival rate by using the distribution model, wherein the commonly used assumed distribution model comprises Poisson distribution and Weibull distribution. The nonparametric model does not need to assume or calculate a survival time distribution model, the survival rate is directly estimated according to the sample statistics, a Kaplan-Meier estimator is a common method, but the obtained survival function estimation cannot correct the influence caused by the change of covariates. The semi-parametric method also does not need to know the distribution of the survival time, but finally needs to evaluate the factors influencing the survival rate through a model, most commonly a Cox regression model, which can consider the influence of a plurality of covariates.
The Cox regression model, also known as the Cox Proportional Hazard (Cox Proportional Hazard) model, defines a risk function having the form:
wherein λ is0(t) is a reference risk function, X ═ X1,...,xm) Vector of covariates, h (X) is the log partial risk function. The CoxPH model assumes that the common influence of a plurality of covariates can be characterized by a logarithm part risk function, so the effect influence of the covariates is comprehensively considered, and the survival function of the product can be effectively described under the condition of assumed conformity. When h (X) is θ · X, θ is (θ)1,...,θm) When the coefficients of the linear model, i.e. the logarithmic risk function, can be expressed as a linear combination of covariate values, the model at this time is called a linear cox ph model.
The relationship between the known risk function and the survival function is:
From the above relationship, it can be known that:
in the above-mentioned formula, the first and second groups,referred to as cumulative baseline risk; s0(t)=exp(-H0(t)) is referred to as the baseline survival function. For H0(t) the most widely used method is to use a Breslow estimator, defined by the following equation:
in the above formula, if tiIs the time of occurrence of the event (i.e., the occurrence of the meter failure), thenWherein R (t)i) Is tiSet of individuals still under observation study within the time, otherwise, λ0(ti)=0。
Therefore, as long as the logarithmic partial risk function h (X) is determined, the survival function S (t) can be derived, and the service life of the electric meter can be predicted.
The conventional intelligent ammeter service life prediction method based on the CoxPH model generally has the problems that an ammeter service life prediction model is too static and prediction reliability is low. Therefore, the embodiment of the invention provides a service life prediction method of an intelligent electric meter based on a CoxPH model. Fig. 1 is a schematic flow chart of a method for predicting the service life of an intelligent electric meter based on a cox ph model according to an embodiment of the present invention, as shown in fig. 1, the method includes:
Specifically, abnormal data of the ammeter to be predicted are input into an ammeter life prediction model, and an ammeter survival curve corresponding to the abnormal data of the ammeter to be predicted is output. The abnormal data of the electric meter to be predicted comprises the type of the abnormal occurrence of the electric meter, such as: specific abnormal types such as short circuit of an ammeter mainboard, abnormal display screen or abnormal battery power supply and the like; also included is the installed time from the installation time when various types of anomalies occur, such as: the distance from the ammeter installation time when the main board of the ammeter is short-circuited has already been 567 days, the distance from the ammeter installation time when the display screen is abnormal has already been 678 days, the distance from the ammeter installation time when the battery power supply is abnormal has already been 345 days, and the like. The survival curve is a curve of the survival Probability of the electric meter in relation to time, and the survival curve s (T) is expressed by a formula, namely s (T) ═ Proavailability (T > T).
The electric meter life prediction model is obtained after training based on electric meter abnormal sample data and a predetermined electric meter life label and a deletion label, and a loss function in the electric meter life prediction model training is formed by participation of a logarithmic part risk function in a CoxPH model. From the foregoing, the survival function s (t) is derived by determining the logarithmic partial risk function h (x) in the cox ph model-based electric meter life prediction model, where the training of the electric meter life prediction model is to fit the logarithmic partial risk function h (x) to the nonlinear function h (x) using the training of the neural network. Therefore, the loss function adopted in the process of training the life prediction model of the electric meter is constructed on the basis of the logarithmic partial risk function h (X), so that the h (X) value output in the training process is more and more accurate, and the accuracy of the h (X) value is represented by the accuracy of S (t) derived from the h (X) value.
And step 120, predicting the service life of the electric meter to be predicted based on the electric meter survival curve and a preset survival probability threshold value.
Specifically, the service life of the electric meter to be predicted is predicted based on an electric meter survival curve output by the electric meter life prediction model and a preset survival probability threshold. For example, the preset threshold value of the survival probability is 80%, that is, the maximum survival time of the electric meter is determined according to the survival curve of the electric meter under the condition that the survival rate of the electric meter reaches 80%, for example, the survival probability of the electric meter is just 80% when the survival time is 1000 days according to the survival curve of the electric meter output by the life prediction model of the electric meter, and then 1000 days is the life of the electric meter under the condition that the preset threshold value of the survival probability is 80%.
The method provided by the embodiment of the invention comprises the steps of inputting abnormal data of an ammeter to be predicted into an ammeter life prediction model, and outputting an ammeter survival curve corresponding to the abnormal data of the ammeter to be predicted, wherein the ammeter survival curve is a curve of the relationship between the survival probability of the ammeter to be predicted and time; the electric meter life prediction model is obtained after training is carried out on the basis of electric meter abnormal sample data and a predetermined electric meter life label and a deletion label, a loss function in the electric meter life prediction model training process is formed by participation of a logarithm part risk function in a CoxPH model, the electric meter life prediction model for intelligent electric meter life prediction is dynamically generated through deep learning, and the electric meter life prediction model can be retrained through updating the training sample and the label to obtain a more accurate prediction model. Therefore, the method provided by the embodiment of the invention avoids the situation that the service life prediction model of the intelligent ammeter in the prior art is too static, and improves the service life prediction reliability of the intelligent ammeter.
Based on the embodiment, in the method, the ammeter abnormal sample data is obtained by sorting the collected ammeter overhaul data and the ammeter abnormal alarm historical record.
Specifically, the ammeter abnormity sample data is obtained from overhaul data of the ammeter and an abnormity alarm historical record of the ammeter, and the data is required to be classified and sorted.
Based on the above embodiment, in the method, the collected maintenance data of the electric meter and the abnormal alarm history of the electric meter are sorted to obtain the abnormal sample data of the electric meter, and the method specifically includes:
firstly, the overhaul data and the abnormal alarm historical record of the electric meter are described in the form of the following formula:
In the above-mentioned formula,is the abnormal type of the t-th meter abnormality occurred in the meter of the meter i,is the existing time when the electricity meter of the electricity meter i has the abnormality for the t-th time, d(i)Is the life of the meter, i(i)Deletion tag, Meta, of a meter that is meter i(i)Is the physical information of the meter, T, for meter i(i)The number of abnormal records in an observation period is shown, and omega represents the collected maintenance data and abnormal alarmA set of historically recorded electricity meters;
forming a list of data described by the formula, wherein a main key is the number of the ammeter, an ith row corresponds to abnormal data of the ammeter i, and the list comprises an abnormal column, a deletion column and a life column, wherein the abnormal value of the mth column is a normalized value of the number of times of the mth type of abnormality of the ammeter i observed in an observation period, the deletion column takes a value of 0 or 1, the value of 0 represents that the ammeter has a fault, the value of 1 represents that the ammeter is still alive after the observation time, and the value of the life column is the life of the ammeter i.
Specifically, firstly, the maintenance data and the abnormal alarm history of the electric meter are uniformly described in a format, and the following formula is specifically adopted:
In the above-mentioned formula,is the abnormal type of the t-th meter abnormality occurred in the meter of the meter i,is the existing time when the electricity meter of the electricity meter i has the abnormality for the t-th time, d(i)Is the life of the meter, i(i)Deletion tag, Meta, of a meter that is meter i(i)Is the physical information of the meter, T, for meter i(i)The number of abnormal records in an observation period is shown, and omega represents the collection of the electric meters of the collected overhaul data and the abnormal alarm historical records. The physical information of the electricity meter generally includes various physical parameters and specifications of the electricity meter.
Then, the data are sorted in a table form, table 1 is a preliminary sorting mode, as shown in table 1, the primary key of the table is { table number × abnormal time }, and each row represents an abnormal record of the electric meter corresponding to one table number.
TABLE 1 preliminary finishing mode
In Table 1, each row corresponds to one abnormal condition of the electric meter corresponding to one table number, namely Meta of the electric meter(i)The information includes the attribute, manufacturer, specification, version, and installation time of the electric meter, different exception types are represented by different exception codes, the exception time is the elapsed time from the installation time to the occurrence of the exception, table 1 further includes a deletion column and a life column, wherein the deleted value is 0 or 1, the deleted value is 0, the electric meter has a fault, when the deleted value of the electric meter is 0, the life value represents the life of the electric meter, when the deleted value is 1, the electric meter is still in operation, and when the deleted value of the electric meter is 1, the life value represents the duration of the continuous operation of the electric meter until the observation time.
The data in table 1 were further processed so that the data could be fitted to a CoxPH model. Table 2 shows the final arrangement, and the primary key in table 2 is { table number }. As shown in table 2, one electric meter corresponds to one table number, one table number corresponds to one row in table 2, and the value of each column of the anomaly represents the result obtained by normalizing the number of times that the anomaly is observed in the observation period. For example, the value of abnormality 3 of the electricity meter with the table number "00002" indicates the normalized value of the number of times of occurrence of abnormality 3 of the electricity meter with the table number "00002".
TABLE 2 Final finishing modes
Based on any one of the above embodiments, in the method, determining a normalized numerical value of the number of times that the m-th abnormality occurs to the electric meter of the electric meter i observed in the observation period specifically includes:
the actual number of m-th type of abnormality of the electric meter i isMeter generation for Meter iThe normalized number of m-th class of anomalies ismin (m) and max (m) are the minimum and maximum values, respectively, for the m-th column anomaly,
and determining the normalized numerical value of the number of m-th type abnormality of the electric meter i observed in the observation period based on the formula.
Specifically, the times of the different types of anomalies occurring in each electric meter are normalized, for example, for the mth type of anomaly, the times of the mth type of anomaly occurring in all the electric meters are traversed, and the maximum value max (m) and the minimum value min (m) are found, where the actual times of the mth type of anomaly occurring in the electric meter of the electric meter i isThe normalized number of the m-th type abnormity of the ammeter i isUsing the following formula:
In any of the above embodiments, the formula of the loss function l (θ) in the method is as follows:
wherein h isθ(x) Representing the logarithmic partial risk function, x representing a physical information parameter vector of the electricity meter, xiPhysical information parameter vector, x, of electric meter representing electric meter ijRepresenting a physical information parameter vector of meter j, theta represents a parameter of a neural network training the logarithmic partial risk function, EiIs a deletion, EiThe meter i, denoted 1, survived until the observation time, TiRefers to the life of the meter, R (T), for meter ii) Is TiSet of meters that remain alive over time, j being R (T)i) The meters j in the set.
Specifically, the formula of the loss function l (θ) is further specified here as follows:
wherein h isθ(x) Representing the logarithmic partial risk function, x representing a physical information parameter vector of the electricity meter, xiPhysical information parameter vector, x, of electric meter representing electric meter ijRepresenting a physical information parameter vector of meter j, theta represents a parameter of a neural network training the logarithmic partial risk function, EiIs a deletion, EiThe meter i, denoted 1, survived until the observation time, TiRefers to the life of the meter, R (T), for meter ii) Is TiSet of meters that remain alive over time, j being R (T)i) The meters j in the set. The training of the ammeter life prediction model in the embodiment of the invention is essentially to perform neural network training on a logarithmic part risk function in the ammeter life prediction model, hθ(x) Representing the logarithm part risk function, theta represents the parameter of the neural network for training the logarithm part risk function, and when the parameter theta of the neural network for training the logarithm part risk function is obtained, the logarithm part risk function h can be determinedθ(x) Then, by the above-mentioned conversion relationship from the logarithmic partial risk function to the survival function, a survival curve can be derived.
Based on any one of the above embodiments, an embodiment of the present invention provides a method for training an electric meter life prediction model based on machine learning and a CoxPH model, and fig. 2 is a schematic flow chart of the method for training an electric meter life prediction model based on machine learning and a CoxPH model according to the embodiment of the present invention. Before the training of the ammeter life prediction model shown in fig. 2 is carried out, cleaning and statistical analysis are carried out on ammeter abnormal data and fault data to form a data table containing deletion labels, and the data table is used as a data sample set of subsequent machine learning training. Then, the process shown in fig. 2 is entered, a deep neural network is defined, a parameter structure of the deep neural network is initialized, data of a data sample set which is sorted before is used as input, specifically, an input covariate vector X is an N × 1 vector formed by normalized values corresponding to N anomaly types in table 2, and a target output is an input tag, that is, two columns of data of life and deletion in table 2. The neural network training process is to train network parameters to fit a nonlinear logarithm part risk function h (X), the output of each unit of a hidden layer and an output layer of the neural network in autumn in the training process is obtained to obtain an expression of h (X), a reference risk function is obtained through calculation according to the relation between the risk function and the logarithm part risk function, so that a complete survival function is obtained, then a consistency index C-index is calculated, if the numerical value of the C-index can be accepted, the training is completed, the ammeter life prediction is carried out by using an ammeter life prediction model obtained through current training, and if the numerical value of the C-index can not be accepted, the ammeter life prediction is carried out through the following formula
A loss function is calculated and then the neural network parameters are adjusted using a gradient descent method.
Based on any one of the above embodiments, an embodiment of the present invention provides a smart meter life prediction apparatus based on deep learning and a cox ph model, and fig. 3 is a schematic structural diagram of the smart meter life prediction apparatus based on deep learning and the cox ph model according to the embodiment of the present invention. As shown in fig. 3, the apparatus includes a survival curve determining unit 310 and a life predicting unit 320, wherein,
the survival curve determining unit 310 is configured to input the abnormal data of the to-be-predicted electric meter into an electric meter life prediction model, and output an electric meter survival curve corresponding to the abnormal data of the to-be-predicted electric meter, where the electric meter survival curve is a curve of a relationship between a survival probability of the to-be-predicted electric meter and time; the electric meter life prediction model is obtained by training based on electric meter abnormal sample data and a predetermined electric meter life label and a deletion label, and a loss function in the electric meter life prediction model training is formed by participation of a logarithmic part risk function in a CoxPH model;
the life prediction unit 320 is configured to predict the life of the electric meter to be predicted based on the electric meter survival curve and a preset survival probability threshold.
According to the device provided by the embodiment of the invention, abnormal data of the ammeter to be predicted are input into an ammeter life prediction model, and an ammeter survival curve corresponding to the abnormal data of the ammeter to be predicted is output, wherein the ammeter survival curve is a curve of the relationship between the survival probability of the ammeter to be predicted and time; the electric meter life prediction model is obtained after training is carried out on the basis of electric meter abnormal sample data and a predetermined electric meter life label and a deletion label, a loss function in the electric meter life prediction model training process is formed by participation of a logarithm part risk function in a CoxPH model, the electric meter life prediction model for intelligent electric meter life prediction is dynamically generated through deep learning, and the electric meter life prediction model can be retrained through updating the training sample and the label to obtain a more accurate prediction model. Therefore, the device provided by the embodiment of the invention avoids the phenomenon that the service life prediction model of the intelligent ammeter in the prior art is too static, and improves the service life prediction reliability of the intelligent ammeter.
Based on any one of the embodiments, in the device, the ammeter abnormal sample data is obtained by sorting the collected ammeter overhaul data and the ammeter abnormal alarm historical record.
Based on any one of the above embodiments, in the device, the collected maintenance data of the electric meter and the collected abnormal alarm history of the electric meter are sorted to obtain the abnormal sample data of the electric meter, which specifically includes:
firstly, the overhaul data and the abnormal alarm historical record of the electric meter are described in the form of the following formula:
In the above-mentioned formula,is the abnormal type of the t-th meter abnormality occurred in the meter of the meter i,is the existing time when the electricity meter of the electricity meter i has the abnormality for the t-th time, d(i)Is the life of the meter, i(i)Deletion tag, Meta, of a meter that is meter i(i)Is the physical information of the meter, T, for meter i(i)The number of abnormal records in an observation period is shown, and omega represents a set of the electric meters of the collected maintenance data and the abnormal alarm historical records;
forming a list of data described by the formula, wherein a main key is the number of the ammeter, an ith row corresponds to abnormal data of the ammeter i, and the list comprises an abnormal column, a deletion column and a life column, wherein the abnormal value of the mth column is a normalized value of the number of times of the mth type of abnormality of the ammeter i observed in an observation period, the deletion column takes a value of 0 or 1, the value of 0 represents that the ammeter has a fault, the value of 1 represents that the ammeter is still alive after the observation time, and the value of the life column is the life of the ammeter i.
Based on any one of the above embodiments, in the apparatus, determining a normalized numerical value of the number of times that the m-th type abnormality occurs to the electric meter of the electric meter i observed in the observation period specifically includes:
the actual number of m-th type of abnormality of the electric meter i isThe normalized number of the m-th type abnormity of the ammeter i ismin (m) and max (m) are the minimum and maximum values, respectively, for the m-th column anomaly,
and determining the normalized numerical value of the number of m-th type abnormality of the electric meter i observed in the observation period based on the formula.
In the apparatus according to any of the above embodiments, the formula of the loss function l (θ) is as follows:
wherein h isθ(x) Representing the logarithmic partial risk function, x representing a physical information parameter vector of the electricity meter, xiPhysical information parameter vector, x, of electric meter representing electric meter ijRepresenting a physical information parameter vector of meter j, theta represents a parameter of a neural network training the logarithmic partial risk function, EiIs a deletion, EiThe meter i, denoted 1, survived until the observation time, TiRefers to the life of the meter, R (T), for meter ii) Is TiSet of meters that remain alive over time, j being R (T)i) The meters j in the set.
Fig. 4 is a schematic entity structure diagram of an electronic device according to an embodiment of the present invention, and as shown in fig. 4, the electronic device may include: a processor (processor)401, a communication Interface (communication Interface)402, a memory (memory)403 and a communication bus 404, wherein the processor 401, the communication Interface 402 and the memory 403 complete communication with each other through the communication bus 404. The processor 401 may call a computer program stored in the memory 403 and executable on the processor 401 to execute the method for predicting the service life of the smart meter based on deep learning and the CoxPH model provided in the foregoing embodiments, for example, the method includes: inputting abnormal data of an ammeter to be predicted into an ammeter life prediction model, and outputting an ammeter survival curve corresponding to the abnormal data of the ammeter to be predicted, wherein the ammeter survival curve is a curve of the relationship between the survival probability of the ammeter to be predicted and time; the electric meter life prediction model is obtained by training based on electric meter abnormal sample data and a predetermined electric meter life label and a deletion label, and a loss function in the electric meter life prediction model training is formed by participation of a logarithmic part risk function in a CoxPH model; and predicting the service life of the electric meter to be predicted based on the electric meter survival curve and a preset survival probability threshold value.
In addition, the logic instructions in the memory 403 may be implemented in the form of software functional units and stored in a computer readable storage medium when the software functional units are sold or used as independent products. Based on such understanding, the technical solutions of the embodiments of the present invention may be essentially implemented or make a contribution to the prior art, or may be implemented in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the methods described in the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The embodiment of the present invention further provides a non-transitory computer-readable storage medium, on which a computer program is stored, where the computer program is implemented to, when executed by a processor, perform the method for predicting the lifetime of a smart meter based on deep learning and a cox ph model provided in the foregoing embodiments, for example, the method includes: inputting abnormal data of an ammeter to be predicted into an ammeter life prediction model, and outputting an ammeter survival curve corresponding to the abnormal data of the ammeter to be predicted, wherein the ammeter survival curve is a curve of the relationship between the survival probability of the ammeter to be predicted and time; the electric meter life prediction model is obtained by training based on electric meter abnormal sample data and a predetermined electric meter life label and a deletion label, and a loss function in the electric meter life prediction model training is formed by participation of a logarithmic part risk function in a CoxPH model; and predicting the service life of the electric meter to be predicted based on the electric meter survival curve and a preset survival probability threshold value.
Embodiments of the present invention further provide a non-transitory computer-readable storage medium, on which a computer program is stored, where the computer program is implemented to perform the identity verification method provided in the foregoing embodiments when executed by a processor, for example, the method includes: sending a verification request to a server to request a maze image to be verified from the server; the maze image to be verified comprises a maze path in a disconnected state; receiving the maze image to be verified, and acquiring maze path communication behavior information corresponding to the maze image to be verified; the maze path communication behavior information represents the rotation behavior of a maze subimage in the maze image to be verified; and returning the maze path communication behavior information to the server to request the server to determine an identity verification result based on the maze path communication behavior information.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A smart meter service life prediction method based on deep learning and a CoxPH model is characterized by comprising the following steps:
inputting abnormal data of an ammeter to be predicted into an ammeter life prediction model, and outputting an ammeter survival curve corresponding to the abnormal data of the ammeter to be predicted, wherein the ammeter survival curve is a curve of the relationship between the survival probability of the ammeter to be predicted and time;
the electric meter life prediction model is obtained by training based on electric meter abnormal sample data and a predetermined electric meter life label and a deletion label, and a loss function in the electric meter life prediction model training is formed by participation of a logarithmic part risk function in a CoxPH model;
and predicting the service life of the electric meter to be predicted based on the electric meter survival curve and a preset survival probability threshold value.
2. The smart meter life prediction method based on deep learning and the CoxPH model according to claim 1, wherein the meter abnormality sample data is obtained by arranging collected maintenance data of the meter and abnormality alarm history records of the meter.
3. The smart meter life prediction method based on deep learning and a CoxPH model according to claim 2, wherein the method for collecting the maintenance data of the smart meter and the abnormal alarm history of the smart meter to obtain the abnormal sample data of the smart meter comprises the following steps:
firstly, the overhaul data and the abnormal alarm historical record of the electric meter are described in the form of the following formula:
In the above-mentioned formula,is the abnormal type of the t-th meter abnormality occurred in the meter of the meter i,is the existing time when the electricity meter of the electricity meter i has the abnormality for the t-th time, d(i)Is the life of the meter, i(i)Deletion tag, Meta, of a meter that is meter i(i)Is the physical information of the meter, T, for meter i(i)The number of abnormal records in an observation period is shown, and omega represents a set of the electric meters of the collected maintenance data and the abnormal alarm historical records;
forming a list of data described by the formula, wherein a main key is the number of the ammeter, an ith row corresponds to abnormal data of the ammeter i, and the list comprises an abnormal column, a deletion column and a life column, wherein the abnormal value of the mth column is a normalized value of the number of times of the mth type of abnormality of the ammeter i observed in an observation period, the deletion column takes a value of 0 or 1, the value of 0 represents that the ammeter has a fault, the value of 1 represents that the ammeter is still alive after the observation time, and the value of the life column is the life of the ammeter i.
4. The smart meter life prediction method based on deep learning and the CoxPH model according to claim 3, wherein the determining of the normalized number of m-th type anomalies of the meter i observed in the observation period specifically comprises:
the actual number of m-th type of abnormality of the electric meter i isThe normalized number of the m-th type abnormity of the ammeter i ismin (m) and max (m) are the minimum and maximum values, respectively, for the m-th column anomaly,
and determining the normalized numerical value of the number of m-th type abnormality of the electric meter i observed in the observation period based on the formula.
5. The method for predicting the life span of a smart meter based on deep learning and a CoxPH model according to any one of claims 1-4, wherein the formula of the loss function l (θ) is as follows:
wherein h isθ(x) Representing the logarithmic partial risk function, x representing a physical information parameter vector of the electricity meter, xiPhysical information parameter vector, x, of electric meter representing electric meter ijRepresenting a physical information parameter vector of meter j, theta represents a parameter of a neural network training the logarithmic partial risk function, EiIs a deletion, EiThe meter i, denoted 1, survived until the observation time, TiRefers to the life of the meter, R (T), for meter ii) Is TiSet of meters that remain alive over time, j being R (T)i) The meters j in the set.
6. The utility model provides a smart electric meter life-span prediction device based on deep learning and CoxPH model which characterized in that includes:
the device comprises a survival curve determining unit, a life prediction unit and a life prediction unit, wherein the survival curve determining unit is used for inputting abnormal data of the ammeter to be predicted into an ammeter life prediction model and outputting an ammeter survival curve corresponding to the abnormal data of the ammeter to be predicted, and the ammeter survival curve is a curve of the relationship between the survival probability of the ammeter to be predicted and time; the electric meter life prediction model is obtained by training based on electric meter abnormal sample data and a predetermined electric meter life label and a deletion label, and a loss function in the electric meter life prediction model training is formed by participation of a logarithmic part risk function in a CoxPH model;
and the service life prediction unit is used for predicting the service life of the electric meter to be predicted based on the electric meter survival curve and a preset survival probability threshold value.
7. The intelligent ammeter life prediction device based on deep learning and CoxPH model as claimed in claim 6, wherein said ammeter abnormality sample data is obtained by collecting overhaul data of the ammeter and abnormality alarm history of the ammeter.
8. The smart meter life prediction device based on deep learning and a CoxPH model according to claim 7, wherein the device for collecting maintenance data of the smart meter and abnormal alarm history of the smart meter to obtain the abnormal sample data of the smart meter comprises:
firstly, the overhaul data and the abnormal alarm historical record of the electric meter are described in the form of the following formula:
The above mentionedIn the formula (I), the compound is shown in the specification,is the abnormal type of the t-th meter abnormality occurred in the meter of the meter i,is the existing time when the electricity meter of the electricity meter i has the abnormality for the t-th time, d(i)Is the life of the meter, i(i)Deletion tag, Meta, of a meter that is meter i(i)Is the physical information of the meter, T, for meter i(i)The number of abnormal records in an observation period is shown, and omega represents a set of the electric meters of the collected maintenance data and the abnormal alarm historical records;
forming a list of data described by the formula, wherein a main key is the number of the ammeter, an ith row corresponds to abnormal data of the ammeter i, and the list comprises an abnormal column, a deletion column and a life column, wherein the abnormal value of the mth column is a normalized value of the number of times of the mth type of abnormality of the ammeter i observed in an observation period, the deletion column takes a value of 0 or 1, the value of 0 represents that the ammeter has a fault, the value of 1 represents that the ammeter is still alive after the observation time, and the value of the life column is the life of the ammeter i.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program implements the steps of the deep learning and cox ph model based smart meter life prediction method according to any of claims 1-5.
10. A non-transitory computer readable storage medium, having stored thereon a computer program, wherein the computer program, when executed by a processor, implements the steps of the deep learning and cox ph model based smart meter lifetime prediction method of any of claims 1-5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911185780.0A CN112380759B (en) | 2019-11-27 | 2019-11-27 | Smart electric meter service life prediction method based on deep learning and CoxPH model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911185780.0A CN112380759B (en) | 2019-11-27 | 2019-11-27 | Smart electric meter service life prediction method based on deep learning and CoxPH model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112380759A true CN112380759A (en) | 2021-02-19 |
CN112380759B CN112380759B (en) | 2024-01-09 |
Family
ID=74586030
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911185780.0A Active CN112380759B (en) | 2019-11-27 | 2019-11-27 | Smart electric meter service life prediction method based on deep learning and CoxPH model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112380759B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112906994A (en) * | 2021-04-19 | 2021-06-04 | 拉扎斯网络科技(上海)有限公司 | Order meal delivery time prediction method and device, electronic equipment and storage medium |
CN116302661A (en) * | 2023-05-15 | 2023-06-23 | 合肥联宝信息技术有限公司 | Abnormality prediction method and device, electronic equipment and storage medium |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CA2602867A1 (en) * | 2007-09-18 | 2009-03-18 | Blue Line Innovations Inc. | System and method for reading power meters |
CN110261811A (en) * | 2019-07-05 | 2019-09-20 | 北京志翔科技股份有限公司 | Intelligent electric meter batch method for early warning and system |
-
2019
- 2019-11-27 CN CN201911185780.0A patent/CN112380759B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CA2602867A1 (en) * | 2007-09-18 | 2009-03-18 | Blue Line Innovations Inc. | System and method for reading power meters |
CN110261811A (en) * | 2019-07-05 | 2019-09-20 | 北京志翔科技股份有限公司 | Intelligent electric meter batch method for early warning and system |
Non-Patent Citations (3)
Title |
---|
严晶晶;王海巍;张卫欣;陈红芳;胡长安;: "智能电能表软件可靠性测试研究", 电测与仪表, no. 07, pages 97 - 102 * |
田正其;祝宇楠;刘建;徐晴;: "基于贝叶斯的智能电能表可靠性评估方法", 计算机测量与控制, no. 10, pages 274 - 277 * |
贺宁;: "智能电表故障大数据分析探究", 中小企业管理与科技(上旬刊), no. 07, pages 142 - 145 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112906994A (en) * | 2021-04-19 | 2021-06-04 | 拉扎斯网络科技(上海)有限公司 | Order meal delivery time prediction method and device, electronic equipment and storage medium |
CN116302661A (en) * | 2023-05-15 | 2023-06-23 | 合肥联宝信息技术有限公司 | Abnormality prediction method and device, electronic equipment and storage medium |
CN116302661B (en) * | 2023-05-15 | 2023-10-13 | 合肥联宝信息技术有限公司 | Abnormality prediction method and device, electronic equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN112380759B (en) | 2024-01-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7107187B1 (en) | Method for modeling system performance | |
US7082381B1 (en) | Method for performance monitoring and modeling | |
US8918345B2 (en) | Network analysis system | |
CN111314173B (en) | Monitoring information abnormity positioning method and device, computer equipment and storage medium | |
CN107992410B (en) | Software quality monitoring method and device, computer equipment and storage medium | |
US7369967B1 (en) | System and method for monitoring and modeling system performance | |
CN106940678B (en) | System real-time health degree evaluation and analysis method and device | |
CN111262750B (en) | Method and system for evaluating baseline model | |
CN111796233B (en) | Method for evaluating secondary errors of multiple voltage transformers in double-bus connection mode | |
CN112381258B (en) | Meter operation life cycle prediction method and device | |
CN111539493A (en) | Alarm prediction method and device, electronic equipment and storage medium | |
CN112380759B (en) | Smart electric meter service life prediction method based on deep learning and CoxPH model | |
CN109460567A (en) | A kind of maintaining method and system of multi-part equipment | |
CN116049146A (en) | Database fault processing method, device, equipment and storage medium | |
CN111985816A (en) | Power equipment maintenance mode risk cost assessment method, device, terminal and medium | |
CN111767193A (en) | Server data anomaly detection method and device, storage medium and equipment | |
CN107480703B (en) | Transaction fault detection method and device | |
JP2022127958A (en) | Business improvement support device, program, and storage medium storing program | |
CN115114124A (en) | Host risk assessment method and device | |
CN110413482B (en) | Detection method and device | |
CN111931323B (en) | Memory, hydrocracking equipment fault prediction method, device and equipment | |
CN117061335A (en) | Cloud platform equipment health management and control method and device, storage medium and electronic equipment | |
JP7062505B2 (en) | Equipment management support system | |
CN115794479A (en) | Log data processing method and device, electronic equipment and storage medium | |
US20230244754A1 (en) | Automatic anomaly thresholding for machine learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |