[go: up one dir, main page]

CN107817404B - Portable metering automation terminal fault diagnosis device and diagnosis method thereof - Google Patents

Portable metering automation terminal fault diagnosis device and diagnosis method thereof Download PDF

Info

Publication number
CN107817404B
CN107817404B CN201711149412.1A CN201711149412A CN107817404B CN 107817404 B CN107817404 B CN 107817404B CN 201711149412 A CN201711149412 A CN 201711149412A CN 107817404 B CN107817404 B CN 107817404B
Authority
CN
China
Prior art keywords
data
module
fault
training
automation terminal
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201711149412.1A
Other languages
Chinese (zh)
Other versions
CN107817404A (en
Inventor
陈俊
龙东
韦杏秋
李捷
潘俊涛
唐志涛
何涌
郭小璇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Electric Power Research Institute of Guangxi Power Grid Co Ltd
Original Assignee
Electric Power Research Institute of Guangxi Power Grid Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Electric Power Research Institute of Guangxi Power Grid Co Ltd filed Critical Electric Power Research Institute of Guangxi Power Grid Co Ltd
Priority to CN201711149412.1A priority Critical patent/CN107817404B/en
Publication of CN107817404A publication Critical patent/CN107817404A/en
Application granted granted Critical
Publication of CN107817404B publication Critical patent/CN107817404B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • 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
    • 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
    • Y04S10/52—Outage or fault management, e.g. fault detection or location

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Test And Diagnosis Of Digital Computers (AREA)
  • Testing Or Calibration Of Command Recording Devices (AREA)

Abstract

本发明涉及计量自动化终端异常诊断技术，具体涉及一种便携式计量自动化终端故障诊断装置及其诊断方法，具体包括检测子模块、电源管理模块和人机交互模块；本发明通过对神经网络和支持向量机技术相结合，首先利用训练好的Ababoost分类器和神经网络对计量自动化终端的各类数据进行分类并提取特征，再利用支持向量机为所提取的数据特征找到最优的分类面，对终端的故障原因和故障所在点进行精确定位，提高了计量自动化终端设备的诊断效率。

Figure 201711149412

The present invention relates to a measurement automation terminal abnormality diagnosis technology, in particular to a portable measurement automation terminal fault diagnosis device and a diagnosis method thereof, specifically including a detection sub-module, a power management module and a human-computer interaction module; the invention uses a neural network and a support vector Combining computer technology, first use the trained Ababoost classifier and neural network to classify and extract the features of various types of data from the metering automation terminal, and then use the support vector machine to find the optimal classification surface for the extracted data features. The cause of the fault and the location of the fault can be accurately located, which improves the diagnostic efficiency of the metering automation terminal equipment.

Figure 201711149412

Description

Portable metering automation terminal fault diagnosis device and diagnosis method thereof
Technical Field
The invention relates to a metering automation terminal abnormality diagnosis technology, in particular to a portable metering automation terminal fault diagnosis device and a diagnosis method thereof.
Background
At present, a metering automation system in a power grid company in China can be built on a large scale, and a large number of matched metering automation terminal devices are installed at metering nodes. Along with the continuous expansion of the scale of the metering automation system, the metering automation terminals in the field operation are continuously increased, and the fault processing work of the field terminals is also continuously increased, so that a power grid company needs to face larger maintenance cost pressure, and how to rapidly and accurately realize the analysis and diagnosis of the terminal faults is an important problem to be solved by the current power enterprises. The portable mobile detection terminal is applied to a metering automation system, intelligent automatic diagnosis of common faults such as a screen blacking state, off-line state, incapability of collecting data and the like of the terminal is realized, rapid processing of the faults of the terminal is realized, the on-line rate and the collection integrity rate of the terminal are improved, the operation and maintenance efficiency is improved, and guarantee is provided for on-site operation maintenance and fault processing of the metering automation terminal.
The metering automation terminal abnormality diagnosis technology comprises the implementation of metering automation terminal field wiring and interface fault abnormality diagnosis, terminal communication channel abnormality, communication protocol compliance field diagnosis and the like. In the existing metering automation terminal diagnosis technology, in the face of complex terminal systems and various fault phenomena, it is often difficult to accurately and rapidly judge and locate the root cause and fault point of the fault.
Disclosure of Invention
In order to solve the problems, the invention provides a portable metering automation terminal fault diagnosis device and a diagnosis method thereof, wherein the portable metering automation terminal fault diagnosis device is firstly based on data information such as real-time voltage, current, power, electric energy, calendar day, calendar month, load curve and the like which are preprocessed by a neural network classifier, and then a support vector machine is used for learning a new classification sample set as a final classifier to carry out more accurate judgment, so that the diagnosis efficiency of the metering automation terminal equipment fault is improved. The specific technical scheme is as follows:
the portable metering automation terminal fault diagnosis device comprises a detection sub-module, a power management module and a man-machine interaction module; the detection sub-module is connected with the man-machine interaction module; the power management module is respectively connected with the detection sub-module and the man-machine interaction module; the man-machine interaction module is used for inputting related instructions to carry out configuration information, and the detection sub-module is used for detecting terminal faults; the power management module provides power for the detection sub-module and the man-machine interaction module; the detection submodule comprises a main control module, a local communication module, a remote communication module, an on-off module, an alternating current sampling module and a meter reading module; the main control module is respectively connected with the local communication module, the remote communication module, the on-off module, the alternating current sampling module, the meter reading module and the man-machine interaction module;
the main control module is used for controlling the fault detection, diagnosis and communication of the terminal;
the local communication module comprises a carrier communication module and a micropower wireless communication module;
the remote communication module is used for uploading collected electric energy data;
the on-off module is provided with a plurality of paths of switch inputs and a plurality of paths of switch outputs and is used for testing whether the switching value input and output of the metering automation terminal are normal or not;
the alternating current sampling module is used for collecting three-phase voltage data and three-phase current data;
the meter reading module is used for simulating an ammeter and providing data of the ammeter read by the metering automation terminal.
A diagnosis method of a portable metering automation terminal fault diagnosis device comprises the following steps:
(1) Collecting data: the detection sub-module collects various data of the metering automation terminal equipment, and specifically comprises control data, remote signaling data, freezing data, time setting data and voltage qualification rate data;
(2) And (3) information normalization processing: the detection sub-module performs standard normalization processing on various data of the acquired metering automation terminal equipment, and converts a data sample of the various data into a numerical value in the range of [0,1] to serve as an input variable of various Adaboost weak classifiers;
(3) Training weights of various Adaboost classifiers: initializing training data weights for N data samples of various types of data, wherein each training sample is given the same weight: 1/N, through inputting fuzzy collected abnormal data and different fault types of output, through repeated iterative computation, a data sample set with updated weight is used for training a next data sample set, each Adaboost weak classifier obtained through training is combined into an Adaboost strong classifier, the weight of the Adaboost weak classifier with small classification error is increased, and the weight of the Adaboost weak classifier with large classification error rate is reduced;
(4) Training by adopting a neural network on the basis of the Adaboost strong classifier: taking various fault reasons and fault possibilities of the metering automation terminal as input quantities, and taking various collected data obtained by a weighted Adaboost strong classifier as output; during training, various fault reasons are taken as input variables, various fault history data are taken as output variables, and a neural network is utilized for feature extraction and training;
(5) And classifying various data acquired by the metering automation terminal by using a trained Ababoost strong classifier and a neural network, extracting characteristics of the received signals, and finding an optimal classification surface for the extracted characteristics by using a support vector machine, namely, classifying the most accurate fault types.
Further, in the step (2), standard normalization processing is performed on various types of data collected by the metering automation terminal device, and data samples of various types of data are converted into numerical values in the range of [0,1], and the method specifically comprises the following steps:
is provided with K evaluation data indexes and M times of acquisition data, x ij The value r of the j-th data acquisition of the normalized ith data index is obtained by carrying out standard normalization processing on the value of the j-th data acquisition of the ith data index ij ：
Figure BDA0001473149910000021
Wherein->
Figure BDA0001473149910000022
The maximum value of the jth data acquisition in the kth data indexValues and minimums.
Further, the abnormal data in the step (3) comprise voltage data, current data, read-write data flow of a local communication module, data flow of a remote communication module and switching value input/output state data.
Further, the fault types in the step (3) include voltage-current wiring errors, abnormality of a local communication module, abnormality of a remote communication module, abnormality of a switching value input/output module, and the like.
Further, the training of the weights of the various Adaboost classifiers in the step (3) specifically includes the following steps:
1) Initializing training data weights for N data samples of various types of data, wherein each training sample is given the same weight: 1/N; d1 is a weight matrix of training samples:
Figure BDA0001473149910000031
2) Performing multiple iterations, wherein m represents the iteration times, and in the whole iteration process, learning a data sample set with weight distribution aiming at different fault types caused by different collected abnormal data to obtain a basic classifier Gm (x):
Figure BDA0001473149910000032
3) Calculating a classification error rate e of Gm (x) on the learning data sample set m I.e. by G m (x) The sum of the weights of the misclassified samples:
Figure BDA0001473149910000033
wherein w is mi The weight of each training sample is iterated for the mth time.
4) Calculating coefficients of Gm (x), wherein α m Represents the importance of Gm (x) in analyzing the Adaboost classifier for the final failure cause:
Figure BDA0001473149910000034
5) Updating the distribution of training value weights for the next round of weight updating, wherein Zm is a weight normalization factor:
D m+1 ＝(w m+1,1 ,w m+1,2 ...w m+1,i ...,w m+1,N )；
Figure BDA0001473149910000035
6) After the iteration is completed, each weak classifier is combined to obtain a final strong classifier:
Figure BDA0001473149910000036
further, the specific steps of feature extraction in the step (4) by using the neural network are as follows:
1) Given a label-free metering automation terminal fault sample set x= { X i |1≤i≤L}，x i Representing an ith sample in the fault sample set, wherein the sample length is m; mapping the input fault vector by an automatic encoder, wherein the output vector set is Y= { Y i |1≤i≤L}，h i Representing the feature vector corresponding to the ith failure sample, h=f (W,b) (X)＝s f (WX+b), W is the weight matrix of the input layer and the hidden layer of the neural network, b is the bias matrix between the input layer and the hidden layer, s f An activation function for the encoder portion;
2) Reconstructing the hidden layer output variable obtained by the encoder into an original input variable: the output vector set is
Figure BDA0001473149910000041
The length of the output vector is the same as the length of the fault vector before decoding, and the mathematical analysis formula of the decoder is as follows:
Figure BDA0001473149910000042
s g activating a function for neurons of the decoder section;
3) By continuously minimizing reconstruction errors between the output vector and the input vector
Figure BDA0001473149910000043
The purpose of extracting features is achieved, and the reconstruction error is +.>
Figure BDA0001473149910000044
And continuously adjusting the weight matrix and the bias matrix of the input layer and the hidden layer of the neural network by using a gradient descent method to minimize the reconstruction error, wherein the specific implementation formula is as follows:
Figure BDA0001473149910000045
Figure BDA0001473149910000046
wherein o is the learning rate of the neural network;
4) For the reconstruction error between the output vector and the input vector calculated by the formula, the weight and the bias { W ] are continuously adjusted by an error back propagation algorithm 1 ,b 1 ,W 1 ',b 1 ' minimize the construction error, complete the training of the first stage of the nerve; then, reserving the encoder part of the current stage, wherein the characteristic layer output vector is used as the input vector of the input layer of the next stage neural network;
5) Training the second-stage neural network according to the same steps 1) to 4), repeating the training process of the steps 1) to 4), and finishing the training of the last-stage neural network, wherein when all the previous training is finished, the hidden layer output of the last layer is the final feature vector.
The beneficial effects of the invention are as follows: according to the invention, by combining the neural network and the support vector machine technology, the trained Ababoost classifier and the neural network are utilized to classify and extract the characteristics of various data of the metering automation terminal, and the support vector machine is utilized to find the optimal classification surface for the extracted data characteristics, so that the fault reason and the fault point of the terminal are accurately positioned, and the diagnosis efficiency of the metering automation terminal equipment is improved.
Drawings
FIG. 1 is a schematic diagram of a portable metering automation terminal fault diagnosis device in the invention;
FIG. 2 is a schematic diagram of the training steps of the Adaboost classifier;
fig. 3 is a block diagram of a neural network including an automatic encoder according to the present invention.
Detailed Description
For a better understanding of the present invention, reference is made to the following description of the invention, taken in conjunction with the accompanying drawings and specific examples:
as shown in fig. 1, a portable metering automation terminal fault diagnosis device comprises a detection sub-module, a power management module and a man-machine interaction module; the detection sub-module is connected with the man-machine interaction module; the power management module is respectively connected with the detection sub-module and the man-machine interaction module; the man-machine interaction module is used for inputting related instructions to carry out configuration information, and the detection sub-module is used for detecting terminal faults; the power management module provides power for the detection sub-module and the man-machine interaction module; the detection submodule comprises a main control module, a local communication module, a remote communication module, an on-off module, an alternating current sampling module and a meter reading module; the main control module is respectively connected with the local communication module, the remote communication module, the on-off module, the alternating current sampling module, the meter reading module and the man-machine interaction module.
The main control module is used for controlling the fault detection and diagnosis of the terminal and communication, and specifically, three-phase voltage and three-phase current data acquisition and opening and closing amount management: e.g. 4-way switching value input/output management; communication management: such as electric energy meters, concentrator communications (RS 232, RS485 and infrared); USB management, wireless communication management, etc., and local/remote communication module testing, etc.
The local communication module comprises a carrier communication module and a micropower wireless communication module.
The remote communication module is used for uploading collected electric energy data and comprises a remote GPRS/4G communication module.
The on-off module is provided with 4-way switch input and 4-way switch output and is used for testing whether the switching value input and output of the metering automation terminal are normal.
The alternating current sampling module is used for collecting three-phase voltage data and three-phase current data, wherein the three-phase current sampling can be compatible with a three-phase three-wire system and a three-phase four-wire system.
The meter reading module is used for simulating the ammeter and providing data of the ammeter read by the metering automation terminal so as to judge whether meter reading of the metering automation terminal is normal or not.
A diagnosis method of a portable metering automation terminal fault diagnosis device comprises the following steps:
1. collecting data: the detection sub-module collects various data of the metering automation terminal equipment, and specifically comprises control data, remote signaling data, freezing data, time setting data and voltage qualification rate data.
2. And (3) information normalization processing: the detection sub-module performs standard normalization processing on various data of the acquired metering automation terminal equipment, and converts a data sample of the various data into a numerical value in the range of [0,1] to serve as an input variable of various Adaboost weak classifiers; the method comprises the following steps of carrying out standard normalization processing on various data of the acquired metering automation terminal equipment, converting a data sample of the various data into a numerical value in a range of [0,1], and specifically comprising the following steps:
is provided with K evaluation data indexes and M times of acquisition data, x ij The value r of the j-th data acquisition of the normalized ith data index is obtained by carrying out standard normalization processing on the value of the j-th data acquisition of the ith data index ij ：
Figure BDA0001473149910000061
Wherein the method comprises the steps of
Figure BDA0001473149910000062
The maximum value and the minimum value of the jth data acquisition values in the kth data index are respectively.
3. Training weights of various Adaboost classifiers: initializing training data weights for N data samples of various types of data, wherein each training sample is given the same weight: 1/N, through inputting fuzzy collected abnormal data and different fault types of output, through repeated iterative computation, a data sample set with updated weight is used for training a next data sample set, each Adaboost weak classifier obtained through training is combined into an Adaboost strong classifier, the weight of the Adaboost weak classifier with small classification error is increased, and the weight of the Adaboost weak classifier with large classification error rate is reduced.
The abnormal data comprise voltage data, current data, read-write data flow of a local communication module, data flow of a remote communication module and switching value input and output state data. The fault types comprise voltage and current wiring errors, abnormal local communication modules, abnormal remote communication modules and abnormal switching value output and output modules.
As shown in fig. 2, the training of the weights of the Adaboost classifiers specifically includes the following steps:
1) Initializing training data weights for N data samples of various types of data, wherein each training sample is given the same weight: 1/N; d1 is a weight matrix of training samples:
Figure BDA0001473149910000063
2) Performing multiple iterations, wherein m represents the iteration times, and in the whole iteration process, learning a data sample set with weight distribution aiming at different fault types caused by different collected abnormal data to obtain a basic classifier Gm (x):
Figure BDA0001473149910000064
3) Calculating a classification error rate e of Gm (x) on the learning data sample set m I.e. by G m (x) The sum of the weights of the misclassified samples:
Figure BDA0001473149910000065
wherein w is mi The weight of each training sample is iterated for the mth time.
4) Calculating coefficients of Gm (x), wherein α m Represents the importance of Gm (x) in analyzing the Adaboost classifier for the final failure cause:
Figure BDA0001473149910000066
5) Updating the distribution of training value weights for the next round of weight updating, wherein Zm is a weight normalization factor:
D m+1 ＝(w m+1,1 ,w m+1,2 ...w m+1,i ...,w m+1,N )
Figure BDA0001473149910000071
6) After the iteration is completed, each weak classifier is combined to obtain a final strong classifier:
Figure BDA0001473149910000072
4. training the neural network on the basis of the Adaboost strong classifier: taking various fault reasons and fault possibilities of the metering automation terminal as input quantities, and taking various collected data obtained by a weighted Adaboost strong classifier as output; during training, various fault reasons are taken as input variables, various fault history data are taken as output variables, and the neural network is utilized for feature extraction and training.
FIG. 3 is a schematic diagram of a neural network structure including an automatic encoder, wherein the specific steps of feature extraction using the neural network are as follows:
1) Given a label-free metering automation terminal fault sample set x= { X i |1≤i≤L}，x i Representing an ith sample in the fault sample set, wherein the sample length is m; mapping the input fault vector by an automatic encoder, wherein the output vector set is Y= { Y i |1≤i≤L}，h i Representing the feature vector corresponding to the ith failure sample, h=f (W,b) (X)＝s f (WX+b), W is the weight matrix of the input layer and the hidden layer of the neural network, b is the bias matrix between the input layer and the hidden layer, s f Is an activation function of the encoder portion.
2) Reconstructing the hidden layer output variable obtained by the encoder into an original input variable: the output vector set is
Figure BDA0001473149910000073
The length of the output vector is the same as the length of the fault vector before decoding, and the mathematical analysis formula of the decoder is as follows:
Figure BDA0001473149910000074
s g the function is activated for neurons of the decoder part.
3) By continuously minimizing reconstruction errors between the output vector and the input vector
Figure BDA0001473149910000075
The purpose of extracting features is achieved, and the reconstruction error is +.>
Figure BDA0001473149910000076
And continuously adjusting the weight matrix and the bias matrix of the input layer and the hidden layer of the neural network by using a gradient descent method to minimize the reconstruction error, wherein the specific implementation formula is as follows:
Figure BDA0001473149910000077
Figure BDA0001473149910000078
where o is the learning rate of the neural network.
4) For the reconstruction error between the output vector and the input vector calculated by the above formula, the weight and bias { W ] are continuously adjusted by using an error back propagation algorithm 1 ,b 1 ,W 1 ',b 1 ' minimize the construction error, complete the training of the first stage of the nerve; the encoder portion of the present stage is then preserved, with the feature layer output vector being the input vector of the next stage neural network input layer.
5) Training the second-stage neural network according to the same steps 1) to 4), repeating the training process of the steps 1) to 4), and finishing the training of the last-stage neural network, wherein when all the previous training is finished, the hidden layer output of the last layer is the final feature vector.
5. The method comprises the steps of classifying various data acquired by a metering automation terminal by using a trained Ababoost classifier and a neural network, extracting characteristics of received signals, and finding an optimal classification surface for the extracted characteristics by using a support vector machine, namely, classifying the most accurate fault types. Among the N types of faults for metering automation terminal faults, for example the common types: for training samples (x) after feature extraction by using a neural network, a remote communication module failure, a meter reading module failure, an alternating current sampling module failure, an input/output module failure, and the like i ,y i ),i＝1,2,…,N，y i Is a category label, y i ∈(1,-1)。
6. By using a support vector machine classifier with two classes, an N-class classifier can be constructed by the following steps:
1) Constructing N support vector machine SVM classifier rules with two classifications: constructing a classification function f of the training sample j (x) J=1, 2..n, the j-th class of samples is separated from the training samples of the other classes (if training sample x i Belonging to the j-th class, sgn [ f ] j (x i )]=1, otherwise sgn [ f ] j (x i )]＝-1)。
2) By selecting a function f j (x) J=1, 2. In N in N categories class F (x) corresponding to maximum value i )＝argmax{f 1 (x i ).,..f N (x i ) And constructing an N-type classifier which can separate each type from the rest N-1 type fault samples, thereby realizing the purpose of diagnosing and classifying the faults of the metering automatic terminal.
The present invention is not limited to the specific embodiments described above, but is to be construed as being limited to the preferred embodiments of the present invention, and any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention are intended to be included in the scope of the present invention.

Claims (5)

1.一种便携式计量自动化终端故障诊断方法，其特征在于：包括以下步骤：1. A method for fault diagnosis of a portable automated metering terminal, characterized by comprising the following steps: (1)采集数据：检测子模块采集计量自动化终端设备的各类数据，具体包括控制数据、遥信数据、冻结数据、对时数据、电压合格率数据；(1) Data collection: The detection submodule collects various types of data from the metering automation terminal equipment, including control data, remote signaling data, freeze data, time synchronization data, and voltage qualification rate data. (2)信息归一化处理：检测子模块对采集到的计量自动化终端设备的各类数据进行标准归一化处理，将各类数据的数据样本转化为[0,1]范围内的数值，以作为各类Adaboost弱分类器的输入变量；(2) Information normalization processing: The detection submodule performs standard normalization processing on the various types of data collected from the metering automation terminal equipment, and transforms the data samples of various types of data into values in the range of [0,1], which are used as input variables for various Adaboost weak classifiers; (3)各类Adaboost分类器权重的训练：对各类数据的N个数据样本初始化训练数据权重，每一个训练样本开始被赋予相同的权值：1/N，通过输入模糊化的采集到的异常数据和输出的不同故障类型，通过多次迭代计算，权重更新过的数据样本集用于训练下一个数据样本集，并将训练得到的各个Adaboost弱分类器组合成Adaboost强分类器，并加大分类误差小的Adaboost弱分类器的权重，降低分类误差率大的Adaboost弱分类器的权重；各类Adaboost分类器权重的训练具体包括以下步骤：(3) Training of Adaboost classifier weights: Initialize training data weights for N data samples of each data type. Each training sample is initially assigned the same weight: 1/N. By inputting fuzzy collected abnormal data and outputting different fault types, through multiple iterations, the weight-updated data sample set is used to train the next data sample set. The trained Adaboost weak classifiers are combined into an Adaboost strong classifier, and the weights of Adaboost weak classifiers with small classification errors are increased, while the weights of Adaboost weak classifiers with large classification error rates are decreased. The specific training of Adaboost classifier weights includes the following steps: 1)对各类数据的N个数据样本初始化训练数据权重，每一个训练样本开始被赋予相同的权值：1/N；D1为训练样本的权值矩阵：1) Initialize the training data weights for N data samples of each data type. Each training sample is initially assigned the same weight: 1/N; D1 is the weight matrix of the training samples: D1＝(w11,w12...w1i...,w1N),
Figure FDA0004119503630000011
D1＝(w 11 ,w 12 ...w 1i ...,w 1N ),
Figure FDA0004119503630000011
 2)进行多次迭代，用m代表迭代次数，在整个迭代过程中，针对采集到的不同异常数据导致的不同故障类型，采用具有权值分布的数据样本集学习，得到基本分类器Gm(x)：2) Perform multiple iterations, with m representing the number of iterations. Throughout the iteration process, for different fault types caused by different collected abnormal data, learn using a data sample set with weighted distribution to obtain the basic classifier Gm(x):
Figure FDA0004119503630000012
Figure FDA0004119503630000012
 3)计算Gm(x)在学习数据样本集上的分类误差率em，即被Gm(x)误分类样本的权重之和：3) Calculate the classification error rate e <sub>m </sub> of G<sub>m</sub>(x) on the training data sample set, which is the sum of the weights of the samples misclassified by G <sub>m </sub>(x):
Figure FDA0004119503630000021
Figure FDA0004119503630000021
 其中wmi为第m次迭代各个训练样本的权重；Where w_mi is the weight of each training sample in the m-th iteration; 4)计算Gm(x)的系数，其中αm表示Gm(x)在分析最终故障原因的Adaboost分类器的重要程度：4) Calculate the coefficients of Gm(x), where αm represents the importance of Gm(x) in the Adaboost classifier for analyzing the final cause of failure:
Figure FDA0004119503630000022
Figure FDA0004119503630000022
 5)更新训练值权重的分布，用于下一轮的权重更新，其中Zm是权重规范化因子：5) Update the distribution of training weights for the next round of weight updates, where Zm is the weight normalization factor: Dm+1＝(wm+1,1,wm+1,2...wm+1,i...,wm+1,N)；D m+1 =(w m+1,1 ,w m+1,2 ...w m+1,i ...,w m+1,N );
Figure FDA0004119503630000023
Figure FDA0004119503630000023
 6)迭代完成后，组合各个弱分类器从而得到最终的强分类器：6) After iteration, combine the weak classifiers to obtain the final strong classifier:
Figure FDA0004119503630000024
Figure FDA0004119503630000024
 (4)在所述的Adaboost强分类器基础上对神经网络进行训练：将计量自动化终端的各种故障原因和故障可能性作为输入量，通过加权Adaboost强分类器得到的采集的各类数据作为输出；训练时，将各类故障原因作为输入变量，各类故障的历史数据作为输出变量，利用神经网络进行特征提取和训练；(4) Train the neural network based on the Adaboost strong classifier: take the various fault causes and fault probabilities of the metering automation terminal as input, and take the various types of data collected by the weighted Adaboost strong classifier as output; during training, take the various fault causes as input variables, take the historical data of various faults as output variables, and use the neural network for feature extraction and training. (5)利用训练好的Ababoost强分类器和神经网络对计量自动化终端采集的各类数据进行分类，并对于接收的信号进行提取特征，利用支持向量机为所提取的特征找到最优的分类面，即最准确的故障类型分类。(5) Use the trained Ababoost strong classifier and neural network to classify various types of data collected by the metering automation terminal, extract features from the received signals, and use support vector machine to find the optimal classification surface for the extracted features, that is, the most accurate fault type classification. 2.根据权利要求1所述的一种便携式计量自动化终端故障诊断方法，其特征在于：所述步骤(2)中对采集到的计量自动化终端设备的各类数据进行标准归一化处理，将各类数据的数据样本转化为[0,1]范围内的数值，具体包括以下步骤：设有K个评价数据指标和M次的采集数据，xij为第i项数据指标的第j次数据采集的值，对其进行标准归一化处理得归一化后的第i项数据指标的第j次数据采集的值rij：
Figure FDA0004119503630000031
2. A method for fault diagnosis of a portable automated metering terminal according to claim 1, characterized in that: in step (2), the various types of data collected from the automated metering terminal are subjected to standard normalization processing, and the data samples of various types of data are converted into values in the range of [0,1]. Specifically, this includes the following steps: Given K evaluation data indicators and M collection data, x <sub>ij</sub> is the value of the j-th data collection of the i-th data indicator. Standard normalization processing is performed on x<sub>ij</sub> to obtain the normalized value r <sub>ij </sub> of the j-th data collection of the i-th data indicator.
Figure FDA0004119503630000031
 其中
Figure FDA0004119503630000032
分别为第K项数据指标中第j次数据采集的值的最大值与最小值。in
Figure FDA0004119503630000032
These are the maximum and minimum values of the j-th data collection in the K-th data indicator. 3.根据权利要求1所述的一种便携式计量自动化终端故障诊断方法，其特征在于：所述步骤(3)中的异常数据包括电压数据、电流数据、本地通信模块读写数据流、远程通信模块数据流、开关量输入输出状态数据。3. The portable metering automation terminal fault diagnosis method according to claim 1, characterized in that: the abnormal data in step (3) includes voltage data, current data, local communication module read/write data stream, remote communication module data stream, and switch input/output status data. 4.根据权利要求1所述的一种便携式计量自动化终端故障诊断方法，其特征在于：所述步骤(3)中的故障类型包括电压电流接线错误、本地通信模块异常、远程通信模块异常、开关量输入输出模块异常。4. The method for fault diagnosis of a portable metering automation terminal according to claim 1, characterized in that: the fault types in step (3) include voltage and current wiring errors, local communication module malfunctions, remote communication module malfunctions, and switch input/output module malfunctions. 5.用于实施权利要求1-4任一所述的一种便携式计量自动化终端故障诊断方法的装置，其特征在于：包括检测子模块、电源管理模块和人机交互模块；检测子模块与人机交互模块连接；电源管理模块分别与检测子模块和人机交互模块连接；所述人机交互模块用于输入相关指令进行配置信息，所述检测子模块用于检测终端故障；所述电源管理模块为检测子模块和人机交互模块提供电源；所述检测子模块包括主控模块、本地通信模块、远程通信模块、开入开出模块、交流采样模块和抄表模块；所述主控模块分别与本地通信模块、远程通信模块、开入开出模块、交流采样模块、抄表模块和人机交互模块连接；5. An apparatus for implementing the portable metering automation terminal fault diagnosis method according to any one of claims 1-4, characterized in that: it comprises a detection submodule, a power management module, and a human-machine interaction module; the detection submodule is connected to the human-machine interaction module; the power management module is connected to both the detection submodule and the human-machine interaction module; the human-machine interaction module is used to input relevant instructions for configuration information; the detection submodule is used to detect terminal faults; the power management module provides power to the detection submodule and the human-machine interaction module; the detection submodule comprises a main control module, a local communication module, a remote communication module, an input/output module, an AC sampling module, and a meter reading module; the main control module is connected to the local communication module, the remote communication module, the input/output module, the AC sampling module, the meter reading module, and the human-machine interaction module. 所述主控模块用于控制终端故障检测、诊断以及通信；The main control module is used to control terminal fault detection, diagnosis, and communication. 所述本地通信模块包括载波通信模块、微功率无线通信模块；The local communication module includes a carrier communication module and a low-power wireless communication module; 所述远程通信模块用于上传采集的电能量数据；The remote communication module is used to upload the collected electrical energy data; 所述开入开出模块具有若干路开关输入、若干路开关输出，用于测试计量自动化终端开关量输入输出是否正常；The input/output module has several switch inputs and several switch outputs, which are used to test whether the switch input and output of the metering automation terminal are normal. 所述交流采样模块用于采集三相电压数据、三相电流数据；The AC sampling module is used to collect three-phase voltage data and three-phase current data; 所述抄表模块用于模拟电表，用于提供计量自动化终端抄读电表的数据。The meter reading module is used to simulate an electricity meter and to provide data for the meter reading automation terminal.
CN201711149412.1A 2017-11-18 2017-11-18 Portable metering automation terminal fault diagnosis device and diagnosis method thereof Active CN107817404B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711149412.1A CN107817404B (en) 2017-11-18 2017-11-18 Portable metering automation terminal fault diagnosis device and diagnosis method thereof

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711149412.1A CN107817404B (en) 2017-11-18 2017-11-18 Portable metering automation terminal fault diagnosis device and diagnosis method thereof

Publications (2)

Publication Number Publication Date
CN107817404A CN107817404A (en) 2018-03-20
CN107817404B true CN107817404B (en) 2023-06-20

Family

ID=61609342

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711149412.1A Active CN107817404B (en) 2017-11-18 2017-11-18 Portable metering automation terminal fault diagnosis device and diagnosis method thereof

Country Status (1)

Country Link
CN (1) CN107817404B (en)

Families Citing this family (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108535572B (en) * 2018-05-10 2020-10-23 长沙理工大学 Metering system secondary circuit monitoring method and device based on fundamental wave zero sequence characteristics
CN109086889B (en) * 2018-09-30 2021-05-11 广东电网有限责任公司 Terminal fault diagnosis method, device and system based on neural network
CN109490816B (en) * 2018-10-16 2021-04-20 广西电网有限责任公司电力科学研究院 A kind of split metering automatic terminal fault detection equipment and detection method
JP7044175B2 (en) * 2019-03-26 2022-03-30 東芝三菱電機産業システム株式会社 Abnormality judgment support device
CN110728836A (en) * 2019-10-22 2020-01-24 深圳供电局有限公司 Concentrator meter reading abnormity analysis method and system
CN111131237B (en) * 2019-12-23 2020-12-29 深圳供电局有限公司 Microgrid attack identification method and grid-connected interface device based on BP neural network
CN111157899B (en) * 2020-01-20 2022-09-13 南京邮电大学 Method for estimating SOC of battery based on model fusion idea
CN111447106B (en) * 2020-03-18 2022-05-03 深圳市普威技术有限公司 Fault detection method, device, storage medium and communication equipment
CN113515684B (en) * 2020-04-09 2025-10-14 阿里巴巴集团控股有限公司 Abnormal data detection method and device
CN111551888A (en) * 2020-04-23 2020-08-18 宁夏隆基宁光仪表股份有限公司 Electric energy meter metering data fault analysis method based on improved AdaBoost algorithm
CN112269907B (en) * 2020-11-02 2021-09-03 山东万里红信息技术有限公司 Processing method of health big data of Internet of things
CN113110044A (en) * 2021-03-29 2021-07-13 华北电力大学 Intelligent BIT design method for heavy-duty gas turbine control system controller module based on Elman neural network and SVM
CN115115245A (en) * 2022-07-07 2022-09-27 广西电网有限责任公司 Intelligent electric meter evaluation analysis method based on Gauss-Adaboost

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103941207A (en) * 2014-03-03 2014-07-23 广州供电局有限公司 Electric power measurement automation terminal detection method and system
CN104569576A (en) * 2014-12-19 2015-04-29 上海交通大学 Distributed detection system for illegal power utilization behavior of smart power grid
CN105184084A (en) * 2015-09-14 2015-12-23 深圳供电局有限公司 Method and system for predicting fault type of electric power metering automation terminal
WO2016033883A1 (en) * 2014-09-04 2016-03-10 国家电网公司 Power transmission line gallop risk early-warning method based on adaboost
CN105511452A (en) * 2016-01-11 2016-04-20 深圳供电局有限公司 Automatic testing arrangement of measurement
CN105610132A (en) * 2014-11-13 2016-05-25 郭振威 Line and bus protection method based on initial fault angle, transition resistance and machine learning

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA2805747A1 (en) * 2010-07-16 2012-01-19 The Trustees Of Columiba University In The City Of New York Machine learning for a power grid
CN207516464U (en) * 2017-11-18 2018-06-19 广西电网有限责任公司电力科学研究院 A kind of Portable metering automatization terminal trouble-shooter

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103941207A (en) * 2014-03-03 2014-07-23 广州供电局有限公司 Electric power measurement automation terminal detection method and system
WO2016033883A1 (en) * 2014-09-04 2016-03-10 国家电网公司 Power transmission line gallop risk early-warning method based on adaboost
CN105610132A (en) * 2014-11-13 2016-05-25 郭振威 Line and bus protection method based on initial fault angle, transition resistance and machine learning
CN104569576A (en) * 2014-12-19 2015-04-29 上海交通大学 Distributed detection system for illegal power utilization behavior of smart power grid
CN105184084A (en) * 2015-09-14 2015-12-23 深圳供电局有限公司 Method and system for predicting fault type of electric power metering automation terminal
CN105511452A (en) * 2016-01-11 2016-04-20 深圳供电局有限公司 Automatic testing arrangement of measurement

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
胡金海 等.基于支持向量机的组合分类方法及应用.《推进技术》.2007,第28卷(第6期),669-673页. *
赖敏.基于AdaBoost迭代学习的支持向量机分类算法.《中国优秀硕士学位论文全文数据库信息科技辑》.2012,(第3期),第1页. *

Also Published As

Publication number Publication date
CN107817404A (en) 2018-03-20

Similar Documents

Publication Publication Date Title
CN107817404B (en) Portable metering automation terminal fault diagnosis device and diagnosis method thereof
CN109635928B (en) Identification method of voltage sag causes based on deep learning model fusion
CN111060838B (en) A fault diagnosis method for switching power supply of medical electronic equipment based on multi-dimensional feature fusion
CN110458230A (en) A Method for Discriminating Abnormality of Distribution Transformer Application Data Based on Multi-criteria Fusion
CN110598854A (en) GRU model-based transformer area line loss rate prediction method
CN108761377A (en) A kind of electric energy metering device method for detecting abnormality based on long memory models in short-term
CN113988210B (en) Method, device and storage medium for repairing distorted data of structural monitoring sensor network
CN110533007B (en) Intelligent identification and extraction method for bridge vehicle-mounted strain influence line features
CN111981635B (en) Central air conditioner fault prediction and diagnosis method adopting double intelligent algorithms
CN112507479A (en) Oil drilling machine health state assessment method based on manifold learning and softmax
CN115358347B (en) A Remaining Lifetime Prediction Method for Smart Meters under Different Subsystems
CN113850320A (en) Transformer fault detection method based on improved support vector machine regression algorithm
CN116317937A (en) A Distributed Photovoltaic Power Plant Operation Fault Diagnosis Method
CN113139601A (en) Turnout fault detection method based on LSTM and attention mechanism
CN118052326A (en) A method for predicting water consumption of campus water supply network based on neural network
CN111551888A (en) Electric energy meter metering data fault analysis method based on improved AdaBoost algorithm
CN115184734B (en) A method and system for detecting power grid line faults
CN119179919A (en) On-line monitoring and diagnosing system for hydroelectric equipment
CN119644234A (en) A cloud-edge collaborative fault diagnosis and positioning method and system for smart electric meters in substations
CN120372525A (en) Substation evaluation method and system integrating attention mechanism neural network model
CN115564182A (en) Power grid analysis method based on power grid resource service middling platform
CN118036819A (en) Impact of environmental factors on distribution network operation and equipment maintenance system and method
CN116660678A (en) A Fault Location and Fault Evaluation System for Transmission Network
CN115713162A (en) Distributed low-carbon energy station health state early warning prediction method
CN207516464U (en) A kind of Portable metering automatization terminal trouble-shooter

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