CN117368598B - Electric energy quality monitoring method and device - Google Patents
Electric energy quality monitoring method and device Download PDFInfo
- Publication number
- CN117368598B CN117368598B CN202311298050.8A CN202311298050A CN117368598B CN 117368598 B CN117368598 B CN 117368598B CN 202311298050 A CN202311298050 A CN 202311298050A CN 117368598 B CN117368598 B CN 117368598B
- Authority
- CN
- China
- Prior art keywords
- monitoring
- power quality
- preset
- information
- layer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000012544 monitoring process Methods 0.000 title claims abstract description 218
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000004364 calculation method Methods 0.000 claims abstract description 31
- 238000007781 pre-processing Methods 0.000 claims abstract description 26
- 230000009466 transformation Effects 0.000 claims abstract description 26
- 230000006870 function Effects 0.000 claims description 53
- 238000012545 processing Methods 0.000 claims description 43
- 238000012549 training Methods 0.000 claims description 41
- 239000011159 matrix material Substances 0.000 claims description 39
- 238000011176 pooling Methods 0.000 claims description 37
- 239000013598 vector Substances 0.000 claims description 17
- 238000013519 translation Methods 0.000 claims description 14
- 238000012937 correction Methods 0.000 claims description 13
- 238000012806 monitoring device Methods 0.000 claims description 13
- 238000013528 artificial neural network Methods 0.000 claims description 11
- 238000004458 analytical method Methods 0.000 claims description 8
- 230000001052 transient effect Effects 0.000 claims description 7
- 238000013507 mapping Methods 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 abstract 1
- 238000003860 storage Methods 0.000 description 5
- 230000002159 abnormal effect Effects 0.000 description 4
- 230000008901 benefit Effects 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 239000003990 capacitor Substances 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 230000007257 malfunction Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000005236 sound signal Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
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
- G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
- G01D21/00—Measuring or testing not otherwise provided for
- G01D21/02—Measuring two or more variables by means not covered by a single other subclass
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2131—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on a transform domain processing, e.g. wavelet transform
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2123/00—Data types
- G06F2123/02—Data types in the time domain, e.g. time-series data
-
- 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
Abstract
The invention provides a method and a device for monitoring electric energy quality, which relate to the technical field of electric energy quality monitoring and comprise the following steps: acquiring an original monitoring signal set and an original monitoring time set contained in different power grid sections; preprocessing an original monitoring signal set to obtain a preprocessed monitoring signal set; preprocessing the original monitoring time set to obtain a preprocessed monitoring time set; performing time difference calculation on the preprocessed monitoring signal set to obtain a monitoring signal difference data set; performing wavelet transformation on the monitoring signal differential data set and the preprocessing monitoring time set to obtain a monitoring signal transformation data set; and inputting the monitoring signal conversion data set into a preset power quality monitoring model to obtain a power quality monitoring label. The method starts from the relevance of various electric power events in the electric energy quality, and improves the accuracy and the monitoring classification capability of the monitoring classification.
Description
Technical Field
The invention relates to the technical field of power quality monitoring, in particular to a power quality monitoring method and device.
Background
Along with the continuous upgrading of consumption and production technology, the requirements on the quality of electric energy are continuously improved, and poor electric energy can not only lead to the reduction of the quality of people's production life, but also influence industrial production, for example, when the electric energy is unstable, a large number of unqualified products can be generated, and when serious, electric or electronic equipment can also be caused to malfunction. In a modern power system, almost every event can leave own labels on the power quality of a power grid, and the labels of every event can be found through detection and classification of the power quality of the power grid, so that the running condition of the power grid is mastered. However, the existing power quality monitoring method is single, and monitoring signal analysis is only carried out from the single classification angle, so that the monitoring classification accuracy is low, and the relevance of various power events in the power quality is ignored.
Disclosure of Invention
The invention aims to provide a power quality monitoring method and device, which solve the problem of singleness of the existing power quality monitoring method, realize that the relevance of various power events in the power quality is started, and improve the accuracy and the monitoring classification capability of monitoring classification.
In order to achieve the above purpose, the technical scheme adopted by the invention is as follows:
in a first aspect, the present application provides a method for monitoring power quality, the method comprising:
acquiring an original monitoring signal set and an original monitoring time set contained in different power grid sections;
preprocessing the original monitoring signal set to obtain a preprocessed monitoring signal set;
preprocessing the original monitoring time set to obtain a preprocessed monitoring time set;
performing time difference calculation on the preprocessed monitoring signal set to obtain a monitoring signal difference data set;
performing wavelet transformation on the monitoring signal differential data set and the preprocessing monitoring time set to obtain a monitoring signal transformation data set;
and inputting the monitoring signal transformation data set into a preset power quality monitoring model to obtain a power quality monitoring tag, wherein the power quality monitoring tag is used for responding to power quality monitoring scheduling, and the power quality monitoring model is a convolution model established based on a multi-target deep convolution neural network.
In a second aspect, the present application also provides a power quality monitoring device, the device comprising:
the acquisition module is used for acquiring an original monitoring signal set and an original monitoring time set which are contained in different power grid sections;
the first processing module is used for preprocessing the original monitoring signal set to obtain a preprocessed monitoring signal set;
the second processing module is used for preprocessing the original monitoring time set to obtain a preprocessed monitoring time set;
the third processing module is used for carrying out time difference calculation on the preprocessed monitoring signal set to obtain a monitoring signal difference data set;
the fourth processing module is used for carrying out wavelet transformation on the monitoring signal differential data set and the preprocessing monitoring time set to obtain a monitoring signal transformation data set;
and the fifth processing module is used for inputting the monitoring signal transformation data set into a preset power quality monitoring model to obtain a power quality monitoring label, wherein the power quality monitoring label is used for responding to power quality monitoring scheduling, and the power quality monitoring model is a convolution model established based on a multi-target deep convolution neural network.
The beneficial effects of the invention are as follows:
the abnormal signals caused by the power event are often formed by superposition of various factors, but each abnormal signal in the multiple power quality anomalies is not mutually exclusive and independent, and has certain relevance, for example, voltage dip is often accompanied by harmonic waves, voltage flicker is also often accompanied by transient pulses and the like. The method comprises the steps of firstly carrying out time differential calculation on monitoring signals to establish time relation of various power events from the angle of signal period, then carrying out label classification on a convolution model established by a multi-target deep convolution neural network, and carrying out deep mining on the relation of various power events through a weight relation function. The method greatly improves the accuracy and the monitoring classification capacity of monitoring classification, and provides guidance for later-stage power quality monitoring and scheduling.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of the embodiments of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims thereof as well as the appended drawings.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic flow chart of a power quality monitoring method based on a deep convolutional neural network according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a power quality monitoring device based on a deep convolutional neural network according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a fifth process module according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a power quality monitoring device based on a deep convolutional neural network according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be noted that: like reference numerals and letters denote like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures. Meanwhile, in the description of the present invention, the terms "first", "second", and the like are used only to distinguish the description, and are not to be construed as indicating or implying relative importance.
Example 1:
the embodiment provides a power quality monitoring method, referring to fig. 1, which shows that the method includes steps S1 to S6, specifically includes:
s1, acquiring an original monitoring signal set and an original monitoring time set which are contained in different power grid sections;
in step S1, the original monitoring signal set includes: the original monitoring time set comprises one or more of indexes such as steady-state voltage, steady-state current, harmonic voltage, harmonic current, frequency and voltage three-phase unbalance, and the original monitoring time set comprises one or more of indexes such as power grid power supply time, power grid peak-to-valley time distribution index and the like. The original monitoring signal set and the original monitoring time set are obtained by monitoring systems such as a DC4KV sub-power grid, an AC390V sub-power grid and the like. In addition, when the original monitoring signal sets and the original monitoring time contained in different power grid sections are obtained, steady-state electric quantity parameters can be displayed by the upper computer, so that power grid monitoring staff can master the real-time state of the power grid, and the power grid monitoring staff is assisted in post-processing.
S2, preprocessing the original monitoring signal set to obtain a preprocessed monitoring signal set;
the preprocessing of the original monitoring signal sets contained in different power grid sections comprises amplitude standardization and clipping processing, so that the numerical range of the original monitoring signal sets falls into a preset interval after being scaled, wherein the amplitude standardization and clipping processing are the prior art means.
S3, preprocessing the original monitoring time set to obtain a preprocessed monitoring time set;
the preprocessing of the original monitoring time sets contained in different power grid sections comprises filling of data missing sections according to time sequences, specifically, the original monitoring time sets are arranged according to the time sequences, when the data missing sections occur, weighted average and interpolation is carried out according to adjacent data before and after the missing sections, and therefore integrity of the original monitoring time sets on the time sequences is guaranteed.
S4, performing time difference calculation on the preprocessed monitoring signal set to obtain a monitoring signal difference data set;
since the abnormal signal length varies greatly in the power quality signal, such as: the transient signal may last only 1 period, while the steady state signal may last more than 20 periods. But the power quality signal will show periodic characteristics if it is distorted. Therefore, the method provides that the time difference calculation is carried out on the preprocessing monitoring signal set so as to determine the period information of the faults in different occurrence stages.
Further, for the specific process of explicit time difference calculation, the step S4 includes the following steps:
s41, carrying out one-dimensional signal analysis on the pretreatment monitoring signal set to obtain one-dimensional power quality information corresponding to the pretreatment monitoring signal set;
s42, arranging the one-dimensional power quality information according to a preset period to obtain two-dimensional matrix information corresponding to the one-dimensional power quality information;
in step S42, the preset period is determined according to the sample length in the monitoring signal set, and the two-dimensional matrix information is X (N×L) Where N represents the number of preset periods and L represents the number of sampling points for one period.
S43, translating the two-dimensional matrix information to obtain translated two-dimensional matrix information;
in step S43, the specific calculation formula is:
X j ＝V j X； (1)
in the above formula (1), X j Representing translated two-dimensional matrix information, X represents two-dimensional matrix information corresponding to one-dimensional power quality information, V j Representing a preset translation matrix, which specifically is:
in the above formula (2), V j The method comprises the steps of representing a preset translation matrix, wherein I represents an identity matrix, j represents the number of units of translation, and N represents the number of preset periods.
And S44, performing matrix projection on the translated two-dimensional matrix information to obtain a monitoring signal differential data set.
In step S44, the translated two-dimensional matrix information X j The method comprises the steps of carrying out conventional matrix projection, wherein the method introduces translation units, when different translation units are set, the difference information of preprocessed monitoring signal sets among j periods can be obtained, so that the monitoring signal difference information corresponding to the j periods is obtained, and after complete matrix projection is carried out, a data set containing all translation units, namely a monitoring signal difference data set is obtained.
Therefore, by introducing time difference calculation in the step S4, the monitoring signal characteristics of the long sample can be ensured to be more concentrated; in addition, the characteristic information of the time difference signal can obviously improve the size change range and the classification accuracy of the network input signal.
S5, carrying out wavelet transformation on the monitoring signal differential data set and the preprocessing monitoring time set to obtain a monitoring signal transformation data set;
because the monitoring signal differential data set and the preprocessing monitoring time set are high-dimension feature vectors, dimension degradation is needed to realize feature extraction of the power quality monitoring signal. The method introduces wavelet transformation to decompose signals under different resolutions and different scales, analyzes localization of time and space frequency, decomposes the signals layer by layer through wavelet functions, further obtains all detail components of the original signals under different frequency bands, has good time-frequency analysis capability, and is also suitable for non-stationary signal analysis of abrupt change characteristics. The wavelet transformation in this step is well known in the art.
And S6, inputting the monitoring signal transformation data set into a preset power quality monitoring model to obtain a power quality monitoring label, wherein the power quality monitoring label is used for responding to power quality monitoring scheduling, and the power quality monitoring model is a convolution model established based on a multi-target deep convolution neural network.
Since the wavelet transformation in step S5 is only a preliminary processing of the monitoring signal, when the monitoring condition of the grid signal at this time needs to be further understood, the monitoring signal transformation data set needs to be identified and classified, and here, the method introduces step S6.
Specifically, the power quality monitoring model in the step S6 is a convolution model established based on the multi-target deep convolution neural network. The neural network has the greatest advantage that the characteristic semantic information of the data can be analyzed through self-learning and self-correction without data relation. In case the sample is sufficient, its classification ability can certainly be ensured. The convolution neural network is determined to be used, and is a feedforward neural network taking convolution as a core, wherein the convolution is a basic operation in analysis mathematics, a function for operating on input data is called as a convolution kernel, and the convolution operation refers to a process that the convolution kernel slides on each position of the input data according to the number of steps, and data are weighted and summed every time when the convolution kernel reaches one position.
According to the method, an electric energy quality monitoring training label is obtained through training according to the electric energy quality monitoring model, and after the monitoring signal transformation data set is input in the later period, the electric energy quality monitoring label corresponding to the electric energy quality monitoring training label is obtained, wherein the calculation of the electric energy quality monitoring training label comprises the following steps:
s61, acquiring a multi-target training signal set;
s62, carrying out preset shared layer training on the multi-target training signal set to obtain shared characteristic information of the multi-target training signal set, wherein the shared layer comprises four convolution layers and two pooling layers;
in step S62, the sharing layer is a network layer shared by all targets to capture a large amount of common feature information from all multi-target training signal sets. Here, the method determines that the shared layer includes four convolutional layers and two pooling layers, specifically includes: convolution layer 1-pooling layer 1-convolution layer 2-convolution layer 3-convolution layer 4-pooling layer 2, wherein the convolution kernels in the convolution layers 1 to 4 are sequentially reduced, and the pooling kernels in the pooling layer 1 and the pooling layer 2 are equal. When the pooling kernels in the pooling layer 1 and the pooling layer 2 are equal, overfitting can be avoided so as to prevent feature information from being lost.
S63, training the shared characteristic information in a preset specific layer to obtain characteristic classification information of a multi-target training signal set, wherein the specific layer comprises five convolution layers and three pooling layers;
in step S63, the specific layer passes through different convolution layers for each learning object to extract characteristic information specific to the object. The method determines that the specific layer comprises five layers of convolution layers and three layers of pooling layers, and specifically comprises the following steps: the method comprises the steps of constructing a convolution layer 1-pooling layer 2-convolution layer 3-convolution layer 4-pooling layer 3, wherein the convolution layer 1 and the convolution layer 3 are set for steady-state signals, the convolution layer 2 is set for transient signals, and the convolution layers 3 to 5 are set for voltage distortions in the transient signals; furthermore, the convolution kernels in the convolution layers 1 to 5 decrease in sequence, and the pooling kernels in the pooling layers 1 and 2 decrease in sequence. And when the pooling cores in the pooling layer 1 and the pooling layer 2 are sequentially reduced, the dimension of the data is effectively reduced.
And S64, performing label mapping on the characteristic classification information through a preset Dense layer to obtain an electric energy quality monitoring training label.
In step S64, the Dense layer maps the feature classification information of the multi-target training signal set into the tag space to complete the final classification of the power quality monitoring training tag. The method determines that the preset Dense layer is five connecting layers so as to sequentially realize the classification of unbalanced heavy load, switching of a capacitor bank, harmonic wave, transient signals and steady-state signals.
In step S64, to enhance the association between feature classifications of different individual targets, a weight association function is included, where the weight association function is used to associate feature classifications between individual targets, and the calculation of the weight association function includes S641 to 644, specifically includes:
s641, acquiring learning prediction probability of a single target and entropy loss function difference information of the single target;
s642, calculating the learning prediction probability of the single target and the entropy loss function difference information of the single target through a preset cross model to obtain a loss function of the single target;
wherein, the preset cross model is:
L i (y i ,p i )＝-y i log(p i )-(1-y i )log(1-p i )； (3)
in the above formula (3), L i (y i ,p i ) Representing the loss function of a single target, y i Entropy loss function difference information, p, representing individual targets i Representing the learned prediction probability of a single target.
S643, calculating according to the loss function of the single target and a preset weight factor to obtain loss function correction information of the single target;
the calculation of the loss function correction information of the single target is as follows:
L i ′＝L i (y i ,p i )×α i ； (4)
in the above formula (4), L i ' loss function correction information representing a single target, L i (y i ,p i ) Representing the loss function, alpha, of a single target i Representing the preset weight factor of the ith task.
S644, solving the loss function correction information of the single target through a preset weight relation model to obtain a weight relation function.
The preset weight relation model is as follows:
in the above formula (5), L represents a weight linking function, m represents the total classification target number, L i ' loss function repair representing a single targetPositive information, alpha i Representing the preset weight factor of the ith task.
Therefore, by introducing the weight contact function, on one hand, the learning of each target is balanced, the association information among the labels is acquired in the training process, and on the other hand, the classification accuracy of the network is ensured.
After step S6, to analyze the fault interval of the power grid section for the power quality monitoring tag, steps S7 to S9 are included, specifically:
s7, positioning the section positions according to the power quality monitoring tag to obtain a plurality of power quality fault sections;
and indexing the original monitoring signal sets and the original monitoring time sets which are initially acquired and contained in different power grid sections corresponding to the power quality monitoring labels so as to obtain a plurality of power quality fault sections, wherein the power quality fault sections comprise fault position section information and fault time information.
S8, comparing each power quality fault section with a preset reference power grid section to obtain adjacent sections corresponding to each power quality fault section;
therefore, the reference power grid sections in the normal state are compared, each power quality fault section is amplified, and the adjacent sections of each power quality fault section are conveniently studied in a later period.
And S9, performing fault judgment on adjacent sections corresponding to each power quality fault section to obtain a fault section of each adjacent section.
Specifically, to clarify the specific content of the fault judgment, the step S9 includes the following steps:
s91, vectorizing calculation is carried out on adjacent sections corresponding to each power quality fault section, and association vectors corresponding to each adjacent section are obtained;
s92, carrying out feature calculation on the associated vector corresponding to each adjacent section to obtain a feature value of each associated vector;
in this step, the existing similarity calculation principle may be adopted to calculate the cosine value between the vector of each power quality fault section and the associated vector corresponding to the adjacent section, so as to correspondingly obtain the feature value of each associated vector.
And S93, comparing the characteristic value of each association vector with a preset characteristic interval to obtain a fault interval of each adjacent section.
The preset characteristic interval is calculated according to an empirical formula corresponding to the fault signal, and when the characteristic value of each correlation vector is matched with the preset characteristic interval, the fault interval of the adjacent section is an abnormal monitoring signal containing the fault section with the corresponding power quality, and the adjacent section is an expansion area of the fault interval.
Therefore, the embodiment can locate the power grid section corresponding to the power quality monitoring tag and amplify the section on the basis of the power grid section to realize the amplification and location of the event source position corresponding to the power quality monitoring abnormality in the power grid, so that the power grid section is more beneficial to comprehensively diagnosing the power event.
Example 2:
as shown in fig. 2, the present embodiment provides a power quality monitoring device, which includes:
an acquisition module 1, configured to acquire an original monitoring signal set and an original monitoring time set included in different power grid sections;
the first processing module 2 is used for preprocessing the original monitoring signal set to obtain a preprocessed monitoring signal set;
the second processing module 3 is used for preprocessing the original monitoring time set to obtain a preprocessed monitoring time set;
the third processing module 4 is used for performing time difference calculation on the preprocessed monitoring signal set to obtain a monitoring signal difference data set;
the fourth processing module 5 is configured to perform wavelet transformation on the monitoring signal differential data set and the preprocessed monitoring time set to obtain a monitoring signal transformation data set;
and the fifth processing module 6 is configured to input the monitoring signal transformation data set into a preset power quality monitoring model, so as to obtain a power quality monitoring tag, where the power quality monitoring tag is used for responding to power quality monitoring scheduling, and the power quality monitoring model is a convolution model built based on a multi-target deep convolution neural network.
In one disclosed embodiment of the present invention, the third processing module 4 includes:
a first processing unit 41, configured to perform one-dimensional signal analysis on the pre-processing monitoring signal set, so as to obtain one-dimensional power quality information corresponding to the pre-processing monitoring signal set;
the second processing unit 42 is configured to arrange the one-dimensional power quality information according to a preset period, so as to obtain two-dimensional matrix information corresponding to the one-dimensional power quality information;
a third processing unit 43, configured to translate the two-dimensional matrix information to obtain translated two-dimensional matrix information;
the fourth processing unit 44 is configured to perform matrix projection on the translated two-dimensional matrix information, so as to obtain a differential data set of the monitoring signal.
As shown in fig. 3, in one embodiment of the disclosed method, the fifth processing module 6 includes:
a first obtaining unit 61, configured to obtain a multi-target training signal set;
a fifth processing unit 62, configured to perform preset shared layer training on the multi-target training signal set to obtain shared feature information of the multi-target training signal set, where the shared layer includes four convolution layers and two pooling layers;
a sixth processing unit 63, configured to perform preset specific layer training on the shared feature information to obtain feature classification information of a multi-target training signal set, where the specific layer includes five convolution layers and three pooling layers;
and a seventh processing unit 64, configured to perform label mapping on the feature classification information through a preset Dense layer, so as to obtain an electric energy quality monitoring training label.
In one embodiment of the present disclosure, the seventh processing unit 64 includes a weight contact function, where the weight contact function is used to contact feature classifications between individual targets, and the calculating of the weight contact function includes:
a second acquisition unit 641 for acquiring a learning prediction probability of the single target and entropy loss function difference information of the single target;
a first calculating unit 642, configured to calculate, through a preset cross model, a learning prediction probability of the single target and entropy loss function difference information of the single target, so as to obtain a loss function of the single target;
a second calculating unit 643, configured to calculate according to the loss function of the single target and a preset weight factor, to obtain loss function correction information of the single target;
and a third calculation unit 644, configured to solve the loss function correction information of the single target through a preset weight association model, so as to obtain a weight association function.
In one embodiment of the present disclosure, after the fifth processing module 6, the method further includes:
the sixth processing module 7 is configured to perform zone location positioning according to the power quality monitoring tag, so as to obtain a plurality of power quality fault zones;
the seventh processing module 8 is configured to compare each power quality fault section with a preset reference power grid section to obtain an adjacent section corresponding to each power quality fault section;
and the eighth processing module 9 is configured to perform fault judgment on the adjacent sections corresponding to each power quality fault section, so as to obtain a fault section of each adjacent section.
In one embodiment of the disclosed method, the eighth processing module 9 comprises:
a fourth calculation unit 91, configured to perform vectorization calculation on adjacent sections corresponding to each power quality fault section, so as to obtain an association vector corresponding to each adjacent section;
a fifth calculating unit 92, configured to perform feature calculation on the associated vector corresponding to each adjacent segment, so as to obtain a feature value of each associated vector;
and a sixth calculating unit 93, configured to compare the feature value of each association vector with a preset feature interval to obtain a fault interval of each adjacent section.
It should be noted that, regarding the apparatus in the above embodiments, the specific manner in which the respective modules perform the operations has been described in detail in the embodiments regarding the method, and will not be described in detail herein.
Example 3:
corresponding to the above method embodiment, there is also provided in this embodiment a power quality monitoring device, which can be referred to in correspondence with the power quality monitoring method described in embodiment 1.
Specifically, fig. 4 shows a block diagram of a power quality monitoring device 800. The power quality monitoring device 800 may include: a processor 801, a memory 802. The deep convolutional neural network-based power quality monitoring device 800 may also include one or more of a multimedia component 803, an i/O interface 804, and a communication component 805.
Wherein the processor 801 is configured to control the overall operation of the power quality monitoring apparatus 800 to perform all or part of the steps of the power quality monitoring method described above.
Memory 802 is used to store various types of data to support operation at the power quality monitoring device 800, which may include instructions for any application or method operating on the power quality monitoring device 800, as well as application-related data, such as contact data, messages, pictures, audio, video, and the like. The multimedia component 803 may include a screen and an audio component. Wherein the screen may be, for example, a touch screen, the audio component being for outputting and/or inputting audio signals. The I/O interface 804 provides an interface between the processor 801 and other interface modules, which may be a keyboard, mouse, buttons, etc. The communication component 805 is configured to perform wired or wireless communication between the power quality monitoring device 800 and other devices.
In another exemplary embodiment, a computer readable storage medium is also provided comprising program instructions which, when executed by a processor, implement the steps of the above-described power quality monitoring method. For example, the computer readable storage medium may be the memory 802 including program instructions described above, which are executable by the processor 801 of the power quality monitoring device 800 to perform the power quality monitoring method described above.
Example 4:
corresponding to the above method embodiments, there is also provided in this embodiment a readable storage medium which can be referred to in correspondence with the above described power quality monitoring method.
Specifically, the readable storage medium stores a computer program, and the computer program when executed by a processor realizes the steps of the power quality monitoring method.
The readable storage medium may be a usb disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, and the like.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
The foregoing is merely illustrative of the present invention, and the present invention is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present invention. Therefore, the protection scope of the invention is subject to the protection scope of the claims.
Claims (6)
1. A method of power quality monitoring, comprising:
acquiring an original monitoring signal set and an original monitoring time set contained in different power grid sections;
preprocessing the original monitoring signal set to obtain a preprocessed monitoring signal set;
preprocessing the original monitoring time set to obtain a preprocessed monitoring time set;
performing time difference calculation on the preprocessed monitoring signal set to obtain a monitoring signal difference data set, wherein the time difference calculation comprises the following steps:
carrying out one-dimensional signal analysis on the pretreatment monitoring signal set to obtain one-dimensional power quality information corresponding to the pretreatment monitoring signal set;
arranging the one-dimensional power quality information according to a preset period to obtain two-dimensional matrix information corresponding to the one-dimensional power quality information;
translating the two-dimensional matrix information to obtain translated two-dimensional matrix information; the specific calculation formula is as follows:
；
in the above-mentioned method, the step of,two-dimensional matrix information representing translation, < >>Two-dimensional matrix information representing one-dimensional power quality information,/-for>Representing a preset translation matrix, which specifically is:
；
in the above-mentioned method, the step of,representing a preset translation matrix,/->Representing an identity matrix>Representing the number of units of translation, +.>Representing the number of preset periods;
performing matrix projection on the translated two-dimensional matrix information to obtain a monitoring signal differential data set;
performing wavelet transformation on the monitoring signal differential data set and the preprocessing monitoring time set to obtain a monitoring signal transformation data set;
inputting the monitoring signal transformation data set into a preset power quality monitoring model to obtain a power quality monitoring tag, wherein the power quality monitoring tag is used for responding to power quality monitoring scheduling, the power quality monitoring model is a convolution model established based on a multi-target deep convolution neural network, the power quality monitoring training tag is obtained according to the power quality monitoring model, and the power quality monitoring method comprises the following steps of:
acquiring a multi-target training signal set;
carrying out preset shared layer training on the multi-target training signal set to obtain shared characteristic information of the multi-target training signal set, wherein the shared layer comprises four convolution layers and two pooling layers;
training the shared characteristic information in a preset specific layer to obtain characteristic classification information of a multi-target training signal set, wherein the specific layer comprises five convolution layers and three pooling layers;
performing label mapping on the feature classification information through a preset Dense layer to obtain an electric energy quality monitoring training label, wherein the electric energy quality monitoring training label comprises a weight contact function, the weight contact function is used for connecting feature classifications among single targets, and the calculation of the weight contact function comprises the following steps:
acquiring learning prediction probability of a single target and entropy loss function difference information of the single target;
calculating the learning prediction probability of the single target and the entropy loss function difference information of the single target through a preset cross model to obtain a loss function of the single target;
calculating according to the loss function of the single target and a preset weight factor to obtain loss function correction information of the single target;
solving the loss function correction information of the single target through a preset weight relation model to obtain a weight relation function, wherein the preset weight relation model is as follows:
；
in the above-mentioned method, the step of,representing a weight linking function->Representing the total number of sorting targets, +.>Loss function correction information representing individual targets, +.>Indicate->Preset weight factors for the individual tasks.
2. The power quality monitoring method according to claim 1, wherein the multi-target training signal set is trained by a preset sharing layer to obtain shared characteristic information of the multi-target training signal set, the sharing layer comprises four convolution layers and two pooling layers, and the arrangement of the sharing layers is as follows: convolution layer 1-pooling layer 1-convolution layer 2-convolution layer 3-convolution layer 4-pooling layer 2, wherein the convolution kernels in convolution layers 1-4 decrease in sequence, and the pooling kernels in pooling layer 1 and pooling layer 2 are equal.
3. The power quality monitoring method according to claim 1, wherein the shared feature information is trained on a preset specific layer to obtain feature classification information of a multi-target training signal set, the specific layer comprises five convolution layers and three pooling layers, and the arrangement of the specific layer is as follows: convolution layer 1-pooling layer 1-convolution layer 2-pooling layer 2-convolution layer 3-convolution layer 4-convolution layer 5-pooling layer 3, wherein convolution layer 1 and convolution layer 3 are set for a steady state signal, convolution layer 2 is set for a transient signal, and convolution layer 3 through convolution layer 5 are set for voltage distortion in the transient signal; the convolution kernels in the convolution layers 1 to 5 are sequentially reduced, and the pooling kernels in the pooling layers 1 and 2 are sequentially reduced.
4. The power quality monitoring method according to claim 1, wherein after inputting the monitoring signal transformation data set into a preset power quality monitoring model to obtain a power quality monitoring label, the method comprises:
performing zone position positioning according to the power quality monitoring tag to obtain a plurality of power quality fault zones;
comparing each power quality fault section with a preset reference power grid section to obtain adjacent sections corresponding to each power quality fault section;
and performing fault judgment on adjacent sections corresponding to each power quality fault section to obtain a fault section of each adjacent section.
5. The power quality monitoring method according to claim 4, wherein performing fault judgment on adjacent sections corresponding to each power quality fault section to obtain a fault section of each adjacent section comprises:
carrying out vectorization calculation on adjacent sections corresponding to each power quality fault section to obtain an association vector corresponding to each adjacent section;
performing feature calculation on the associated vector corresponding to each adjacent section to obtain a feature value of each associated vector;
and comparing the characteristic value of each association vector with a preset characteristic interval to obtain a fault interval of each adjacent section.
6. A power quality monitoring device, comprising:
the acquisition module is used for acquiring an original monitoring signal set and an original monitoring time set which are contained in different power grid sections;
the first processing module is used for preprocessing the original monitoring signal set to obtain a preprocessed monitoring signal set;
the second processing module is used for preprocessing the original monitoring time set to obtain a preprocessed monitoring time set;
the third processing module is used for carrying out time difference calculation on the preprocessed monitoring signal set to obtain a monitoring signal difference data set; wherein the third processing module comprises:
the first processing unit is used for carrying out one-dimensional signal analysis on the pretreatment monitoring signal set to obtain one-dimensional power quality information corresponding to the pretreatment monitoring signal set;
the second processing unit is used for arranging the one-dimensional power quality information according to a preset period to obtain two-dimensional matrix information corresponding to the one-dimensional power quality information;
the third processing unit is used for translating the two-dimensional matrix information to obtain translated two-dimensional matrix information; the specific calculation formula is as follows:
；
in the above-mentioned method, the step of,two-dimensional matrix information representing translation, < >>Two-dimensional matrix information representing one-dimensional power quality information,/-for>Representing a preset translation matrix, which specifically is:
;
in the above-mentioned method, the step of,representing a preset translation matrix,/->Representing an identity matrix>Representing the number of units of translation, +.>Representing the number of preset periods;
the fourth processing unit is used for carrying out matrix projection on the translated two-dimensional matrix information to obtain a monitoring signal differential data set;
the fourth processing module is used for carrying out wavelet transformation on the monitoring signal differential data set and the preprocessing monitoring time set to obtain a monitoring signal transformation data set;
the fifth processing module is configured to input the monitoring signal transformation data set into a preset power quality monitoring model, to obtain a power quality monitoring tag, where the power quality monitoring tag is configured to respond to power quality monitoring scheduling, and the power quality monitoring model is a convolution model built based on a multi-target deep convolution neural network, and the fifth processing module includes:
the first acquisition unit is used for acquiring a multi-target training signal set;
the fifth processing unit is used for carrying out preset shared layer training on the multi-target training signal set to obtain shared characteristic information of the multi-target training signal set, wherein the shared layer comprises four convolution layers and two pooling layers;
the sixth processing unit is used for training the shared characteristic information in a preset specific layer to obtain characteristic classification information of the multi-target training signal set, wherein the specific layer comprises five convolution layers and three pooling layers;
a seventh processing unit, configured to perform tag mapping on the feature classification information through a preset Dense layer to obtain an electric energy quality monitoring training tag, where the seventh processing unit includes a weight contact function, the weight contact function is used to contact feature classifications between single targets, and the calculating of the weight contact function includes:
the second acquisition unit is used for acquiring the learning prediction probability of the single target and the entropy loss function difference information of the single target;
the first calculation unit is used for calculating the learning prediction probability of the single target and the entropy loss function difference information of the single target through a preset cross model to obtain a loss function of the single target;
the second calculation unit is used for calculating according to the loss function of the single target and a preset weight factor to obtain loss function correction information of the single target;
the third calculation unit is configured to solve the loss function correction information of the single target through a preset weight association model, so as to obtain a weight association function, where the preset weight association model is:
；
in the above-mentioned method, the step of,representing a weight linking function->Representing the total number of sorting targets, +.>Loss function correction information representing individual targets, +.>Indicate->Preset weight factors for the individual tasks.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311298050.8A CN117368598B (en) | 2023-10-09 | 2023-10-09 | Electric energy quality monitoring method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311298050.8A CN117368598B (en) | 2023-10-09 | 2023-10-09 | Electric energy quality monitoring method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117368598A CN117368598A (en) | 2024-01-09 |
| CN117368598B true CN117368598B (en) | 2024-04-12 |
Family
ID=89388429
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311298050.8A Active CN117368598B (en) | 2023-10-09 | 2023-10-09 | Electric energy quality monitoring method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117368598B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108154223A (en) * | 2017-12-22 | 2018-06-12 | 北京映翰通网络技术股份有限公司 | Power distribution network operating mode recording sorting technique based on network topology and long timing information |
| CN112041693A (en) * | 2018-05-07 | 2020-12-04 | 美国映翰通网络有限公司 | Power distribution network fault positioning system based on mixed wave recording |
| CN114662251A (en) * | 2020-12-24 | 2022-06-24 | 国网西藏电力有限公司拉萨供电公司 | Power distribution network fault positioning method based on deep neural network |
| CN115508673A (en) * | 2022-08-29 | 2022-12-23 | 沈阳航空航天大学 | Series arc fault detection method based on multi-domain depth feature combination |
| CN115792729A (en) * | 2022-11-30 | 2023-03-14 | 广东粤电科试验检测技术有限公司 | Transformer composite fault diagnosis method, device, equipment and storage medium |
| CN116028893A (en) * | 2023-03-28 | 2023-04-28 | 北京电信易通信息技术股份有限公司 | Intelligent data monitoring method, system and electronic equipment |
-
2023
- 2023-10-09 CN CN202311298050.8A patent/CN117368598B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108154223A (en) * | 2017-12-22 | 2018-06-12 | 北京映翰通网络技术股份有限公司 | Power distribution network operating mode recording sorting technique based on network topology and long timing information |
| CN112041693A (en) * | 2018-05-07 | 2020-12-04 | 美国映翰通网络有限公司 | Power distribution network fault positioning system based on mixed wave recording |
| CN114662251A (en) * | 2020-12-24 | 2022-06-24 | 国网西藏电力有限公司拉萨供电公司 | Power distribution network fault positioning method based on deep neural network |
| CN115508673A (en) * | 2022-08-29 | 2022-12-23 | 沈阳航空航天大学 | Series arc fault detection method based on multi-domain depth feature combination |
| CN115792729A (en) * | 2022-11-30 | 2023-03-14 | 广东粤电科试验检测技术有限公司 | Transformer composite fault diagnosis method, device, equipment and storage medium |
| CN116028893A (en) * | 2023-03-28 | 2023-04-28 | 北京电信易通信息技术股份有限公司 | Intelligent data monitoring method, system and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117368598A (en) | 2024-01-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kaplan et al. | An improved feature extraction method using texture analysis with LBP for bearing fault diagnosis | |
| Jia et al. | A deviation based assessment methodology for multiple machine health patterns classification and fault detection | |
| Zhang et al. | Fault detection and recognition of multivariate process based on feature learning of one-dimensional convolutional neural network and stacked denoised autoencoder | |
| Wang et al. | A combination of residual and long–short-term memory networks for bearing fault diagnosis based on time-series model analysis | |
| CN101614786A (en) | Power electronic circuit on-line intelligence method for diagnosing faults based on FRFT and IFSVC | |
| Zhang et al. | Bearing performance degradation assessment based on time-frequency code features and SOM network | |
| Zhang et al. | A hybrid deep-learning model for fault diagnosis of rolling bearings in strong noise environments | |
| Wang et al. | A hybrid intelligent rolling bearing fault diagnosis method combining WKN-BiLSTM and attention mechanism | |
| CN114548190A (en) | Wind turbine fault diagnosis method based on self-adaptive residual error neural network | |
| Zheng et al. | Recognition method of voltage sag causes based on two‐dimensional transform and deep learning hybrid model | |
| Barreto et al. | Artificial neural network approach for fault detection and identification in power systems with wide area measurement systems | |
| CN117368598B (en) | Electric energy quality monitoring method and device | |
| Lei et al. | Bearings fault diagnosis method based on MAM and deep separable dilated convolutional neural network | |
| CN113530850A (en) | Centrifugal pump fault diagnosis method based on EAS (electronic article surveillance) and stacked capsule self-encoder | |
| Cheng et al. | MMFSL: A novel multi-modal few-shot learning framework for fault diagnosis of industrial bearings | |
| Luo et al. | A fault diagnosis method for rotating machinery with variable speed based on multi-feature fusion and improved ShuffleNet V2 | |
| Liu et al. | An anomaly detection method based on double encoder–decoder generative adversarial networks | |
| Li et al. | A sequential feature extraction method based on discrete wavelet transform, phase space reconstruction, and singular value decomposition and an improved extreme learning machine for rolling bearing fault diagnosis | |
| Yuan et al. | A support vector machine framework for fault detection in molecular pump | |
| Bono et al. | Development of soft sensors based on neural networks for detection of anomaly working condition in automated machinery | |
| Sun et al. | Rolling bearing fault diagnosis method based on multisynchrosqueezing s transform and faster dictionary learning | |
| Li et al. | Research on bearing fault detection based on convolution neural network | |
| Gamero et al. | Qssi: A new similarity index for qualitative time series. Application to classify voltage sags | |
| Li et al. | Spatial-temporal attention and information reinforcement network for machine remaining useful life prediction | |
| Zhang et al. | A TFG-CNN Fault Diagnosis Method for Rolling Bearing |
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 |