CN117330906A - Equipment arc fault detection method, device, equipment and storage medium - Google Patents
Equipment arc fault detection method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN117330906A CN117330906A CN202311178746.7A CN202311178746A CN117330906A CN 117330906 A CN117330906 A CN 117330906A CN 202311178746 A CN202311178746 A CN 202311178746A CN 117330906 A CN117330906 A CN 117330906A
- Authority
- CN
- China
- Prior art keywords
- sequence
- data
- value
- arc fault
- equipment
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 113
- 238000009499 grossing Methods 0.000 claims abstract description 109
- 238000000034 method Methods 0.000 claims abstract description 49
- 238000013136 deep learning model Methods 0.000 claims abstract description 26
- 238000007781 pre-processing Methods 0.000 claims abstract description 16
- 239000007772 electrode material Substances 0.000 claims description 15
- 238000000605 extraction Methods 0.000 claims description 13
- 238000005457 optimization Methods 0.000 claims description 9
- 238000013135 deep learning Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 6
- 230000008859 change Effects 0.000 abstract description 12
- 238000004458 analytical method Methods 0.000 description 14
- 238000010586 diagram Methods 0.000 description 12
- 230000008569 process Effects 0.000 description 8
- 239000002245 particle Substances 0.000 description 7
- 238000012545 processing Methods 0.000 description 7
- 230000006872 improvement Effects 0.000 description 6
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 5
- 229910002804 graphite Inorganic materials 0.000 description 5
- 239000010439 graphite Substances 0.000 description 5
- 238000010606 normalization Methods 0.000 description 5
- 238000013450 outlier detection Methods 0.000 description 5
- 230000001932 seasonal effect Effects 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 238000007418 data mining Methods 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 229910001369 Brass Inorganic materials 0.000 description 2
- 229910001141 Ductile iron Inorganic materials 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 2
- XAGFODPZIPBFFR-UHFFFAOYSA-N aluminium Chemical compound [Al] XAGFODPZIPBFFR-UHFFFAOYSA-N 0.000 description 2
- 229910052782 aluminium Inorganic materials 0.000 description 2
- 239000010951 brass Substances 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 239000007770 graphite material Substances 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000000513 principal component analysis Methods 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 229910001220 stainless steel Inorganic materials 0.000 description 2
- 239000010935 stainless steel Substances 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 239000008186 active pharmaceutical agent Substances 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002085 persistent effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012731 temporal analysis Methods 0.000 description 1
- 238000000700 time series analysis Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- 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
- G01R31/12—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing
- G01R31/1227—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing of components, parts or materials
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Abstract
The invention relates to the technical field of equipment detection, and discloses an equipment arc fault detection method, device, equipment and storage medium based on a time sequence prediction model. The method comprises the following steps: acquiring equipment electrical data in a specific time period, preprocessing to obtain a data point sequence, extracting features of each data point in the data point sequence based on a preset time sequence prediction model to obtain a feature quantity sequence, predicting arc faults of the feature quantity sequence by using a preset deep learning model, calculating the accuracy of a prediction result, and realizing the arc fault detection of the equipment based on the lifting ratio and the accuracy of the feature quantity sequence. According to the scheme, the secondary exponential smoothing algorithm is used for extracting the characteristics, the weight is reduced from near to far according to the exponential rule, and when the data mode changes, the change of the data mode is automatically identified for adjustment, so that the prediction deviation is effectively reduced, and the accuracy of equipment arc fault detection is improved.
Description
Technical Field
The invention relates to the field of circuit protection, in particular to a method, a device, equipment and a storage medium for detecting equipment arc faults based on a time sequence prediction model.
Background
In the current operation process of the power distribution and utilization system, arc faults cannot be identified by common overcurrent protection devices such as fuses and circuit breakers, dangerous accidents occur, the existing arc fault detection method is a wavelet coefficient variance, a differential power processing structure or a series direct current arc fault detection method comprehensively utilizing information such as line voltage, power supply voltage and the like, but most of the existing arc fault detection methods utilize surface features of current or voltage, such as maximum amplitude of current and voltage, frequency of current and voltage, average current and voltage and the like, and experimental samples are continuously increased and feature screening is continuously conducted. The existing method has the defect of correlation analysis among data, so that the accuracy of fault detection is not high. Accordingly, there is a need for an arc fault detection scheme that improves detection accuracy.
Disclosure of Invention
The invention mainly aims to solve the technical problems of large processing data volume and low accuracy in the existing arc fault detection scheme.
The first aspect of the invention provides a device arc fault detection method based on a time sequence prediction model, which comprises the following steps: acquiring equipment electrical data in a specific time period, and preprocessing the equipment electrical data to obtain a data point sequence; based on a preset time sequence prediction model, extracting features of each data point in the data point sequence to obtain a feature quantity sequence of the data point sequence; and detecting the arc faults of the equipment by using a preset deep learning model and the lifting ratio of the characteristic quantity sequence.
Optionally, in a first implementation manner of the first aspect of the present invention, the acquiring device electrical data in a specific time period and preprocessing the device electrical data to obtain a data point sequence includes: based on equipment electrical data in a specific time period, which is acquired from a circuit to be detected by acquisition equipment, arranging the equipment electrical data according to a time mark; and carrying out absolute deviation calculation on the electrical data of each device to obtain outliers, and removing the outliers from each data point to obtain a data point sequence.
Optionally, in a second implementation manner of the first aspect of the present invention, the calculating the absolute deviation of the electrical data of each device to obtain the outlier includes: sorting based on the numerical value of each piece of equipment electrical data to obtain the median of the equipment electrical data, and calculating the difference value between each piece of equipment electrical data and the median to obtain the absolute deviation from each piece of equipment electrical data to the median; comparing each absolute deviation with a preset safety deviation, and judging whether the absolute difference is larger than the preset safety deviation or not; and if the absolute deviation is larger than the preset value, determining the equipment electrical data corresponding to the absolute deviation as an outlier.
Optionally, in a third implementation manner of the first aspect of the present invention, the feature extracting, based on a preset time sequence prediction model, each data point in the data point sequence to obtain a feature quantity sequence of the data point sequence includes: calculating the data point sequence based on an exponential smoothing algorithm to obtain a primary smoothing value of the data point sequence; obtaining a secondary smoothed value of the sequence of data points based on the primary smoothed value and the exponential smoothing algorithm; and inputting the primary smoothing value and the secondary smoothing value into a preset time sequence prediction model to obtain a characteristic quantity sequence of the data point sequence.
Optionally, in a fourth implementation manner of the first aspect of the present invention, the inputting the primary smoothed value and the secondary smoothed value into a preset time sequence prediction model to obtain the feature quantity sequence of the data point sequence includes: and determining a parameter variable value of the current time period based on the primary smoothing value and the secondary smoothing value, taking the parameter variable value as the input of a secondary exponential smoothing model, and outputting a predicted value of the current time period to obtain a characteristic quantity sequence of the data point sequence.
Optionally, in a fifth implementation manner of the first aspect of the present invention, after the inputting the primary smoothed value and the secondary smoothed value into a preset time series prediction model, obtaining a feature quantity sequence of the data point sequence, the method further includes: sequentially acquiring algorithm parameters from corresponding preset numerical value intervals based on preset units, and sequentially combining the algorithm parameters, wherein the algorithm parameters comprise a smoothing coefficient, a smoothing algorithm initial value and a predicted lead period number; based on the combined algorithm parameters, generating a corresponding characteristic quantity waveform curve, and determining an optimal algorithm parameter corresponding to the characteristic quantity waveform curve through an optimization algorithm; and taking the optimal algorithm parameter as a target algorithm parameter of a preset time sequence prediction model.
Optionally, in a sixth implementation manner of the first aspect of the present invention, the detecting an arc fault of the device by using a preset deep learning model and a lifting ratio of the feature sequence includes: acquiring equipment electrical data in a normal state acquired in a current circuit, and calculating an average value in the normal state;
comparing the absolute value of the characteristic quantity data sequence with the absolute value to obtain a lifting ratio of the characteristic quantity sequence, and judging whether the lifting ratio exceeds a preset safety interval or not; if the feature quantity sequence is not exceeded, inputting the feature quantity sequence into a preset deep learning model to obtain a model output result; and if the output result is a fault, confirming the arc fault of the equipment.
Optionally, in a seventh implementation manner of the first aspect of the present invention, before the inputting the feature sequence to a preset deep learning model, before obtaining a model output result, the method further includes: acquiring the lifting ratio of the characteristic quantity sequences in at least two electrode materials, and judging whether the lifting ratio exceeds a preset safety interval or not to obtain a judging result; and comparing the judging result with an actual circuit fault result, and training a deep learning detection model to obtain a preset deep learning model.
The second aspect of the present invention provides an apparatus arc fault detection device based on a time series prediction model, the apparatus arc fault detection device comprising:
the preprocessing module is used for acquiring equipment electrical data in a specific time period, and preprocessing the equipment electrical data to obtain a data point sequence;
the prediction module is used for extracting the characteristics of each data point in the data point sequence based on a preset time sequence prediction model to obtain a characteristic quantity sequence of the data point sequence;
and the determining module is used for detecting the arc faults of the equipment by utilizing a preset deep learning model and the lifting ratio of the characteristic quantity sequence.
Optionally, in a first implementation manner of the second aspect of the present invention, the preprocessing module includes:
the arrangement unit is used for arranging the equipment electrical data according to the time mark based on the equipment electrical data in a specific time period acquired by the acquisition equipment from the circuit to be detected;
and the rejecting unit is used for carrying out absolute deviation calculation on the electrical data of each device to obtain outliers, and rejecting the outliers from each data point to obtain a data point sequence.
Optionally, in a second implementation manner of the second aspect of the present invention, the rejection unit is specifically configured to: sorting based on the numerical value of each piece of equipment electrical data to obtain the median of the equipment electrical data, and calculating the difference value between each piece of equipment electrical data and the median to obtain the absolute deviation from each piece of equipment electrical data to the median; comparing each absolute deviation with a preset safety deviation, and judging whether the absolute difference is larger than the preset safety deviation or not; and if the absolute deviation is larger than the preset value, determining the equipment electrical data corresponding to the absolute deviation as an outlier.
Optionally, in a third implementation manner of the second aspect of the present invention, the prediction module includes:
The first operation unit is used for operating the data point sequence based on an exponential smoothing algorithm to obtain a primary smoothing value of the data point sequence;
a second operation unit, configured to obtain a second smoothed value of the data point sequence based on the first smoothed value and the exponential smoothing algorithm;
and the input unit is used for inputting the primary smooth value and the secondary smooth value into a preset time sequence prediction model to obtain a characteristic quantity sequence of the data point sequence.
Optionally, in a fourth implementation manner of the second aspect of the present invention, the input unit is specifically configured to: and determining a parameter variable value of the current time period based on the primary smoothing value and the secondary smoothing value, taking the parameter variable value as the input of a secondary exponential smoothing model, and outputting a predicted value of the current time period to obtain a characteristic quantity sequence of the data point sequence.
Optionally, in a fifth implementation manner of the second aspect of the present invention, the input unit is further configured to: sequentially acquiring algorithm parameters from corresponding preset numerical value intervals based on preset units, and sequentially combining the algorithm parameters, wherein the algorithm parameters comprise a smoothing coefficient, a smoothing algorithm initial value and a predicted lead period number; based on the combined algorithm parameters, generating a corresponding characteristic quantity waveform curve, and determining an optimal algorithm parameter corresponding to the characteristic quantity waveform curve through an optimization algorithm; and taking the optimal algorithm parameter as a target algorithm parameter of a preset time sequence prediction model.
Optionally, in a seventh implementation manner of the second aspect of the present invention, the determining module includes:
the acquisition unit is used for acquiring the equipment electrical data in the normal state acquired in the current circuit and calculating an average value in the normal state;
the comparison unit is used for comparing the absolute value of the characteristic quantity data sequence with the absolute value to obtain the lifting ratio of the characteristic quantity sequence and judging whether the lifting ratio exceeds a preset safety interval or not;
the result judging unit is used for inputting the characteristic quantity sequence into a preset deep learning model if the characteristic quantity sequence is not exceeded, so as to obtain a model output result; and if the output result is a fault, confirming the arc fault of the equipment.
Optionally, in an eighth implementation manner of the second aspect of the present invention, the result judging unit is specifically configured to: acquiring the lifting ratio of the characteristic quantity sequences in at least two electrode materials, and judging whether the lifting ratio exceeds a preset safety interval or not to obtain a judging result; and comparing the judging result with an actual circuit fault result, and training a deep learning detection model to obtain a preset deep learning model.
A third aspect of the present invention provides an apparatus arc fault detection apparatus based on a time series prediction model, the apparatus arc fault detection apparatus comprising a memory and at least one processor, the memory having instructions stored therein; the at least one processor invokes the instructions in the memory to cause the device arc fault detection device to perform the steps of the device arc fault detection method as described above.
A fourth aspect of the invention provides a computer-readable storage medium having instructions stored thereon which, when executed by a processor, implement the steps of the apparatus arc fault detection method as described above.
According to the technical scheme provided by the invention, the equipment electrical data in a specific time period is obtained, the data point sequence is obtained through preprocessing, the characteristic extraction is carried out on each data point in the data point sequence based on a preset time sequence prediction model to obtain a characteristic quantity sequence, the arc fault prediction is carried out on the characteristic quantity sequence by using a preset deep learning model, the accuracy of a prediction result is calculated, and the arc fault detection of the equipment is realized based on the improvement ratio and the accuracy of the characteristic quantity sequence. According to the scheme, the secondary exponential smoothing algorithm is used for extracting the characteristics, the weight is reduced from near to far according to the exponential law, and when the data mode changes, the change of the data mode is automatically identified to be adjusted, so that the prediction deviation is effectively reduced, the accuracy of equipment arc fault detection is improved, and the equipment arc fault detection cost is reduced.
Drawings
Fig. 1 is a schematic diagram of a first embodiment of an arc fault detection method of an apparatus according to an embodiment of the present invention;
Fig. 2 is a schematic diagram of a second embodiment of an arc fault detection method of a device according to an embodiment of the present invention;
fig. 3 is a schematic flow chart of an arc fault detection method of a device according to an embodiment of the present invention;
FIG. 4 is a diagram showing the comparison of the primary data and the characteristic improvement ratio of the secondary exponential smoothing algorithm under the graphite material according to the embodiment of the present invention;
FIG. 5 is a schematic diagram showing the comparison of the secondary exponential smoothing and the improvement of the accuracy of the wavelet scheme to the DC arc fault detection under different electrode materials according to the embodiment of the present invention;
fig. 6 is a schematic structural diagram of an arc fault detection device of an apparatus according to an embodiment of the present invention;
fig. 7 is a schematic diagram of another structure of an arc fault detection device provided in an embodiment of the present invention;
fig. 8 is a schematic structural diagram of an arc fault detection device provided in an embodiment of the present invention.
Detailed Description
Aiming at the existing equipment arc fault detection mode, the method and the device have the advantages that the equipment electrical data is subjected to feature extraction based on the preset time sequence prediction model, the arc fault prediction is performed by utilizing the deep learning model, the arc fault detection of the equipment is realized based on the lifting ratio and the prediction accuracy, the weight is gradually decreased from near to far according to the exponential law, and when the data mode changes, the change of the data mode is automatically recognized and adjusted, the prediction deviation is effectively reduced, the equipment arc fault detection cost is reduced, and the equipment arc fault detection efficiency and accuracy are improved.
The terms "first," "second," "third," "fourth" and the like in the description and in the claims and in the above drawings, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments described herein may be implemented in other sequences than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, apparatus, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus.
For ease of understanding, the following describes a specific flow of an embodiment of the present invention, and please refer to fig. 1 for a schematic diagram of a first embodiment of an arc fault detection method for an apparatus according to an embodiment of the present invention, where the method specifically includes the following steps:
101. and acquiring the equipment electrical data in a specific time period, and preprocessing the equipment electrical data to obtain a data point sequence.
In this embodiment, the preprocessing may be understood as a process of extracting data points of the electrical data of the device and screening the data points, specifically, firstly dividing the electrical data of the device into multiple segments of data according to a sampling rule, then calculating an average value of each segment of data to obtain corresponding data points, finally identifying abnormality of each data point, namely, removing data points which deviate greatly from other data points in the electrical data of the device, namely removing outliers in the electrical data of the device, wherein the detected outliers in the signal are also called outliers, and the outlier detection also becomes abnormality detection, so that in order to avoid the influence of the outliers on feature extraction significance, a method of detecting and removing the outliers is required to be adopted to preprocess the current signal. Outliers are data that deviate from most of the data in the dataset, while clusters with large outliers are considered outlier clusters, i.e. all objects therein are considered outlier behavior. The outlier is a calculated median absolute deviation value, in which the input original data detects that the median is more than three times different, and the position of the outlier is marked when the outlier is detected. The outlier detection algorithm is convenient to operate and simple. Calculating the absolute deviation of the median of the outliers:
MAD＝median(|X i -median(X)|)
Wherein: MAD is the absolute deviation from data point to median, and is the characteristic data obtained by outlier detection; media (X) is the absolute deviation of the median; x is input current data, and the absolute deviation obtained takes an absolute value. According to the method for calculating the median, the equipment electrical data can be arranged based on time marks, the median value is determined based on the number of the equipment electrical data, the equipment electrical data corresponding to the median value is the median value, further, the equipment electrical data in the current time period can be arranged according to the size, the median value is determined based on the median value, the equipment electrical data can be converted into a coordinate axis, the coordinate axis takes time as an abscissa and takes the size of the equipment electrical data as an ordinate, the position of the equipment electrical data on the coordinate axis is obtained, a normalization straight line is obtained based on the distance between each piece of equipment electrical data and a straight line, the point far from the normalization straight line is determined to be an outlier, a regression curve can be generated based on the equipment electrical data, the distance between each piece of equipment electrical data and the regression curve is calculated, and when the distance is larger than a preset value, the equipment electrical data corresponding to the distance is determined to be the outlier.
102. And carrying out feature extraction on each data point in the data point sequence based on a preset time sequence prediction model to obtain a feature quantity sequence of the data point sequence.
The method aims at extracting features of the data point sequence, extracting features of arc faults from the change of the difference between the data points, and accordingly achieving fault detection. The embodiment uses a secondary exponential smoothing process, where the secondary exponential smoothing is a time series prediction method, and in this embodiment, the function is to obtain a statistical rule from data points in a time series in one end of time through calculation, that is, determine a state or degree of change of current in a circuit along with time, and infer future current data from historical and current data according to a change trend of the data, so as to realize detection of arc faults of the circuit. The exponential smoothing method combines an actual value and a smoothed value by using a weighted average method, and constructs a sequence prediction model through the smoothed value to realize a prediction function.
Based on a preset time sequence prediction model, feature extraction is performed on each data point in the data point sequence, and the feature extraction can be realized by the following steps: the secondary exponential smoothing algorithm is used to predict the future values, in particular, to initialize the smoothing parameters and initial values, and for each time point t, to update the level (Lt), trend (Tt) and seasonal index (St) in the following manner:
Update of level (Lt): lt=α×yt+ (1- α) ×lt-1+tt-1;
updating of trend (Tt): tt=β (Lt-1) + (1- β) Tt-1;
updating seasonal index (St): st=γ (Yt-Lt) + (1- γ) St-m.
Where α is a smoothing parameter, typically valued between 0 and 1, representing the weight on the current observation; beta is a trend smoothing parameter, also typically valued between 0 and 1, representing a weight on the change in level; gamma is a seasonal smoothing parameter, typically ranging from 0 to 1, and m is the period of the season (e.g., m is typically 60 if the data is collected in minutes), indicating a weight on the season. The predictive value of future period number (after m-period) is calculated using a secondary exponential smoothing method, typically using the following formula: yt+m=lt+m+tt+st-m; after obtaining the predicted value yt+m of future period number (after m period), the square thereof can be defined as a feature quantity, namely ft= (yt+m)/(2), which feature quantity Ft can be used in subsequent analysis, modeling or other tasks to help capture certain features or trends in the time series data. It should be noted that squaring may be used to amplify the variation of the predicted values, especially for large prediction errors or outliers, which may make certain modes or trends more pronounced, but may also introduce some noise, thus the advantages and disadvantages of using squaring are weighed against specific problems and data situations when using this approach.
And optimizing a smoothing coefficient in a secondary exponential smoothing model through a particle swarm optimization algorithm for related parameters in the model, calculating a predicted value based on the actual value of each data point and the predicted value of the data point of the last time sequence to obtain a primary smoothing value of the data point of each time sequence, and substituting the smoothing coefficient, the primary smoothing value and the secondary smoothing value of the data point of the last time sequence into the secondary exponential smoothing algorithm to obtain a characteristic quantity data array of each data point.
103. And predicting the arc faults of the characteristic quantity sequence by using a preset deep learning model, calculating the accuracy of the prediction result, and realizing the arc fault detection of the equipment based on the lifting ratio and the accuracy of the characteristic quantity sequence.
For detection analysis of direct current arc fault detection, corresponding direct current fault detection can be performed according to the characteristics that different range noises, different line currents, different power supply voltages and the like are generated when arc faults occur. For example, when an arc fault occurs, high-frequency noise is generated in the frequency domain, a high-frequency sinusoidal signal is input to a dc bus, and the variance of the wavelet coefficient is used to determine the arc fault. Or the differential power processing structure of the photovoltaic system is utilized to detect the direct current arc, the differential power processing structure (Differential Power Processing; DPP) and the maximum power point tracking operation (Maximum Power Point Tracking; MPPT) are combined with an arc fault detection algorithm, the function of the DPP device is improved, and a sensor and additional current are not needed. Or a series direct current arc fault detection method comprehensively utilizing information such as line voltage, power supply voltage and the like, wherein the current in a line and the voltage of power supply are used as characteristic parameters of an arc fault to detect the direct current arc, and the direct current arc fault is detected based on the detection of the average value change rate of the line current, the detection of the line current drop, the detection of the alternating current component of the line power supply voltage and the standard deviation of the line current. Or detecting the whole direct current arc fault based on wavelet transformation, decomposing the direct current electric signal into high-frequency and low-frequency signals by utilizing the wavelet, simultaneously carrying out singular value decomposition to construct a characteristic matrix, and finally judging the direct current arc fault by the average value and the standard deviation. Fault arc detection may also be performed using surface features of the current or voltage, such as the maximum magnitude of the current and voltage, the frequency of the current and voltage, the average current and voltage, etc.
According to the scheme, the secondary exponential smoothing algorithm is used for extracting the characteristics, the weight is reduced from near to far according to the exponential rule, and when the data mode changes, the change of the data mode is automatically identified for adjustment, so that the prediction deviation is effectively reduced, and the accuracy of equipment arc fault detection is improved.
Referring to figure 2 for a second embodiment of the method for detecting arc faults in a device according to embodiments of the present invention,
referring to fig. 3, a flow chart of an arc fault detection method for a device according to an embodiment of the present invention is defined, first, if an arc fault exists in a circuit, the output is 1, if an arc fault does not exist in the circuit, the output is 0, first, sampling direct current data is performed, that is, the electrical data of the device in the present invention is the current data in the circuit, the current data can be directly obtained from the circuit through an acquisition device, the current data obtained by sampling is used for judging outliers by using a median difference value, then the current data is subjected to outlier rejection, optimization of the current data is realized, then an optimal algorithm parameter is obtained by a particle swarm optimization parameter, the optimal algorithm parameter is input to a secondary exponential smoothing algorithm, and one-dimensional direct current data is obtained, wherein the one-dimensional direct current data obtained based on the optimal algorithm parameter is one-dimensional direct current data with the maximum lifting ratio. And judging whether an arc fault exists in the circuit based on the one-dimensional direct current data, specifically, calculating the lifting ratio of the one-dimensional direct current data and the direct current data, confirming that the arc fault exists in the circuit when the lifting ratio is larger than a preset safety threshold, confirming that the arc fault does not exist in the circuit if the lifting ratio is not larger than the preset safety threshold, training a deep learning detection model based on the one-dimensional direct current data without the arc fault to obtain an equipment arc deep learning detection model, wherein the equipment arc deep learning detection model can be used for judging whether the arc fault exists in the circuit directly according to the one-dimensional direct current characteristic data, outputting 1 if the arc fault exists, and determining that the arc fault exists in the currently detected direct current circuit.
The method specifically comprises the following steps:
201. and acquiring the equipment electrical data in a specific time period, carrying out absolute deviation calculation on the equipment electrical data, and removing outliers from the equipment electrical data to obtain a data point sequence.
In addition to calculating absolute deviations from outliers, there are methods to identify and process outliers, extract valuable information from the data and optimize the data, specifically: normalization and normalization, converting the data into a form with similar dimensions and distribution, i.e. subtracting the mean and dividing by the standard deviation, whereas normalization typically scales the data to a range of 0 to 1; missing value processing, deleting a row containing missing values, filling the missing values by using an interpolation method, or predicting the missing values by using a machine learning model; outlier detection and handling, in addition to absolute bias, other outlier detection methods, such as Z-score, box plot, DBSCAN clustering, etc., which may be selected to be deleted, replaced or adjusted once outliers are detected to reduce the impact on analysis; for high-dimensional datasets, selecting the most relevant features can improve the performance and interpretation ability of the model, including variance threshold, mutual information, recursive feature elimination, etc., in addition, dimension reduction techniques such as Principal Component Analysis (PCA) can help reduce the number of features while preserving important information in the data; for time series data or data containing noise, smoothing and filtering techniques may be used to remove noise, revealing trends and periodicity, including moving average, exponential smoothing, and low pass filtering.
202. And sequentially calculating the data point sequence, and calculating a primary smooth value and a secondary smooth value.
In the embodiment of the invention, the current data in the circuit is acquired every one hour, the acquisition frequency can be acquired once per minute or once per second, and it is noted that the acquisition period and the acquisition frequency are not limited, and only the change trend of the historical monitoring data is ensured to accurately judge the trend of the future monitoring data, and the value smoothing coefficient alpha is sequentially acquired from 0 to 1 by taking 0.1 as a unit based on the value range of [0,1], and is substituted into a secondary exponential smoothing algorithm formula, wherein the specific secondary exponential smoothing algorithm formula is as follows:
wherein: s is S t (1) A smoothed value that is a primary index; s is(s) t (2) A smoothed value that is a quadratic index; alpha is a smoothing coefficient (the smaller the smoothing coefficient, the stronger the smoothing effect, but the slower the response to the variation of the data); y is t For the observed input, here raw experimental current data, a primary smoothed value and a secondary smoothed value of the current data over a period of time in the present time period are calculated.
203. And inputting the primary smooth value and the secondary smooth value into a time sequence prediction model, and calculating a predicted value in any time period.
At s t (1) Sum s t (2) In the known case, the prediction model of the secondary exponential smoothing method is:
y t+m ＝a t +b t m,m＝1,2,…
wherein: m is the predicted lead period number; a, a t 、b t Variable values for parameters of each period; y is t+m The feature data obtained by the secondary smoothing is used as a predicted value. m is the value of the predicted lead-in number meaning that it is desired to predict m time points in the future, and in the time series analysis the lead-in number represents the number of time steps that it is desired to predict, typically denoted by h, e.g. if m=3, i.e. the value of 3 time points in the future, y t+m Representing predicted values after the m-period, the level, trend and seasonal components are taken into account to predict future values as accurately as possible. The choice of the parameter variable values affects the accuracy of the prediction and usually requires adjustment according to the characteristics of the actual data. Three parameters are involved in the quadratic exponential smoothing algorithm: the initial value of the smoothing algorithm, the smoothing coefficient and the predicted lead period number have a certain calculated amount due to more data, so that a Particle Swarm Optimization (PSO) algorithm is used for finding the most suitable parameters of each waveform. The PSO does not need the adjustment of a plurality of parameters, is simple and easy to realize, can share the individual extremum with other particles, adjusts the position and the speed of the particles through continuous iterative updating, and finally obtains the optimal individual value as the current global optimal solution of the whole particle swarm. After the optimal parameter value is obtained, a secondary exponential smoothing algorithm is input for feature extraction, a predicted value is obtained by using a smoothing parameter and a recent smoothing value, the square of the predicted value is defined as a feature quantity, and then the obtained deep learning model is used for detection, so that the accuracy of the adopted method for detecting the direct current arc faults is obtained. And comparing and analyzing an arc fault detection result with the existing wavelet algorithm, firstly obtaining parameters of a secondary exponential averaging algorithm by using PSO, and obtaining characteristic quantity by using smoothing parameters and a recent smoothing value. The characteristic extraction is carried out after the secondary exponential smoothing algorithm is carried out on the third group of original data characteristics of the graphite Taking the extracted characteristic average value of the fault state and the average value of the normal state, and obtaining the corresponding lifting ratio of the group of data, namely, comparing the average value of the characteristic quantity obtained after secondary exponential smoothing of the electrical data of the equipment in the fault state with the average value of the characteristic quantity obtained after secondary exponential smoothing of the electrical data of the equipment in the normal state, referring to fig. 4, the characteristic lifting ratio comparison schematic diagram of the original data and the secondary exponential smoothing algorithm under the graphite material provided by the embodiment of the invention is shown in the figure, wherein the lifting ratio of the wavelet data characteristic (WPD) and the secondary exponential smoothing algorithm (DS) to the characteristic extraction of the arc fault data is shown in the figure, the first stage is the normal circuit stage, the second stage is the arc fault generation stage, and the third stage is the arc fault ending stage. In the arc fault occurrence stage, the wavelet data feature ratio of the electrode material graphite is close to 1, and the improvement ratio of the extracted graphite feature of the secondary exponential smoothing algorithm is far greater than 1. Under the characteristic of wavelet analysis data, the corresponding characteristic ratio of arc fault data is nearly 1, and whether arc faults exist is difficult to truly judge. After the secondary exponential smoothing mining algorithm, the improvement ratio of the arc fault characteristics is greatly increased, so that the judgment of the arc fault is obviously more accurate, and the detection of the direct current arc fault is facilitated. In practical application, for a large amount of data, a plurality of periods may be sequentially calculated, specifically, from the first period to the tenth period, a predicted value of the secondary exponential smoothing may be obtained to be substantially identical to the actual value, and trend is the same, so that the predicted result and the predicted trend of the period after the tenth period may also highlight the trend in a certain period after the tenth period.
And generating a secondary exponential smoothing prediction model based on the primary smoothing value and the secondary smoothing value of the data points of each time sequence, and determining an initial value of a smoothing algorithm, a smoothing coefficient and an optimal solution for predicting the lead period based on a particle swarm optimization algorithm to obtain the secondary exponential smoothing prediction model based on the current electrode material.
The primary smoothed value is a simple average of data, which is used to smooth random fluctuations and noise of the data, and can be calculated by, in addition to the above method of calculating the primary smoothed value: determining the data set to be smoothed, assuming that the data are a continuous time series, selecting a smoothing window size (generally denoted by n) representing the number of data points for which an average is to be calculated, starting from the beginning of the data, calculating the average of the data points within the window, and taking this average as a primary smoothed value, then moving the window forward, continuing to calculate the next average until all data points have been processed, the primary smoothed value being calculated as: a first order smoothed value= (data point 1+data point 2+, +data point n)/n; for the second smoothed value: the secondary smoothed value is used for smoothing trend and seasonal components of the data, the primary smoothed value is calculated first, then the primary smoothing is applied to the primary smoothed value again to obtain a secondary smoothed value, and besides the method for calculating the secondary smoothed value, the method can also be used for calculating the secondary smoothed value by the following steps: first, a primary smoothed value is calculated, as described above, and then the result of the primary smoothed value is used as a new data set, and the size of a smoothing window, generally denoted by n, is again selected, and starting from the beginning of the primary smoothed value, the average of the data points within the window is calculated and taken as a secondary smoothed value, and then the window is moved forward to continue calculating the next average until all the data points of the primary smoothed value are processed, and the calculation formula of the secondary smoothed value is as follows: secondary smoothed value= (primary smoothed value 1+primary smoothed value 2+) + primary smoothed value n)/n. By way of example, assume the following 5 consecutive times of current data: current data: [50, 55, 60, 65, 70], first calculating a primary smoothed value, assuming that the selection window size is 2, the primary smoothed value 1= (50+55)/2=525; a first order smoothed value of 2= (55+60)/2=575; a first order smoothed value of 3= (60+65)/2=625; the first smoothed value 4= (65+70)/2=675, then the second smoothed value is calculated, and the window size is selected again to be 2: a second order smoothed value of 1= (525+575)/2=550; a second smoothed value 2= (575+625)/2=600; a second smoothed value of 3= (625+675)/2=650; this results in a set of quadratic smoothed values for smoothing the trend and time components of the current data and further analyzing or predicting the task.
204. And obtaining an arc fault detection result based on the predicted value, comparing the detection result with an actual circuit fault result, and training a deep learning detection model to obtain a preset deep learning model.
And detecting the data with the remarkable characteristics in a trained deep learning model, and obtaining the direct current arc fault detection accuracy of several different circuit materials. Referring to fig. 5, comparing the secondary exponential smoothing with the improvement result of the wavelet scheme on the accuracy of dc arc fault detection for different electrode materials provided by the embodiment of the present invention, it can be obtained that the wavelet analysis detection rate of dc arc fault detection and the accuracy of secondary arc fault detection based on the secondary exponential smoothing data mining are higher than the accuracy of arc fault detection under wavelet analysis, where the wavelet analysis means that only one wavelet check primitive function is used to make hadamard product in the same time window. For the graphite electrode material, compared with wavelet analysis, the accuracy of arc fault detection is improved by 34.9% by using a secondary exponential smoothing algorithm; for the nodular cast iron electrode material, the accuracy of arc fault detection is improved by 31.8% by using the secondary exponential smoothing ratio wavelet analysis; for the aluminum electrode material, the accuracy of arc fault detection is improved by 13.9% by using the secondary exponential smoothing ratio wavelet analysis; for the brass electrode material, the accuracy of arc fault detection is improved by 29.8% by using the secondary exponential smoothing than wavelet analysis; for the stainless steel electrode material, the accuracy of arc fault detection is improved by 9.3% by using the secondary exponential smoothing than wavelet analysis. In general, the arc fault detection accuracy is higher than that of wavelet analysis by using a method of secondary exponential smoothing only, and the accuracy is approximately between 77% and 97%. The detection accuracy of the secondary exponential smoothing data mining arc faults is obviously and greatly improved, and the secondary exponential smoothing data mining method is more advantageous than wavelet analysis. The secondary exponential smoothing method is used for mining data, the accuracy of detection of the DC arc faults of the comparison wavelet is obviously improved when the electrode materials are graphite, spheroidal graphite cast iron and brass, the accuracy of detection of the DC arc faults is improved to different degrees when the electrode materials are aluminum and stainless steel, and the accuracy of detection of the DC arc faults is improved by about 24%, and in the scheme, the secondary exponential smoothing algorithm is comprehensively superior to the wavelet.
205. And based on the predicted value in any time period and the output result of a preset deep learning model, arc fault detection is realized.
In this embodiment, whether an arc fault exists may be directly determined according to a predicted value in any time period, specifically, the predicted value is compared with actual data to obtain a lifting ratio of the electrical data of the device in any time period, the lifting ratio amplifies a characteristic of the electrical data of the device, and the arc fault exists in a circuit corresponding to the electrical data of the device and corresponding to a lifting ratio with a large lifting ratio difference in a normal state. And the output result of the detection model is obtained, and if the accuracy rate in the output result is smaller than a preset value, the arc fault exists in a circuit corresponding to the equipment electrical data corresponding to the one-dimensional direct current data.
According to the scheme, the secondary exponential smoothing algorithm is used for carrying out feature extraction, the primary exponential smoothing is carried out again, the operation is convenient, prediction can be carried out only by selecting one model parameter in an actual direct current arc application scene, and the machine learning model is trained based on one-dimensional direct current feature data, so that the accuracy of direct current arc fault detection is obtained, the prediction deviation is effectively reduced, and the accuracy of equipment arc fault detection is improved.
The method for detecting an arc fault of a device according to the embodiment of the present invention is described above, and the arc fault detection device of the embodiment of the present invention is described in detail from the perspective of a modularized functional entity, referring to fig. 6, a schematic structural diagram of the arc fault detection device of a device according to the embodiment of the present invention includes:
the preprocessing module 610 is configured to acquire device electrical data in a specific time period, and preprocess the device electrical data to obtain a data point sequence;
the prediction module 620 is configured to perform feature extraction on each data point in the data point sequence based on a preset time sequence prediction model, so as to obtain a feature quantity sequence of the data point sequence;
And the determining module 630 is used for realizing arc fault detection of the equipment by using a preset deep learning model and the lifting ratio of the characteristic quantity sequence.
According to the scheme, the secondary exponential smoothing algorithm is used for extracting the characteristics, the weight is reduced from near to far according to the exponential rule, and when the data mode changes, the change of the data mode is automatically identified for adjustment, so that the prediction deviation is effectively reduced, and the accuracy of equipment arc fault detection is improved.
Referring to fig. 7, another schematic structural diagram of an arc fault detection apparatus for a device according to an embodiment of the present invention includes:
the preprocessing module 710 is configured to acquire device electrical data in a specific time period, and preprocess the device electrical data to obtain a data point sequence;
the prediction module 720 is configured to perform feature extraction on each data point in the data point sequence based on a preset time sequence prediction model, so as to obtain a feature quantity sequence of the data point sequence;
and a determining module 730, configured to implement arc fault detection of the device by using a preset deep learning model and the lifting ratio of the feature quantity sequence.
In this embodiment, the preprocessing module 710 includes:
An arrangement unit 711 for arranging the device electrical data according to the time stamp based on the device electrical data acquired by the acquisition device from the circuit to be detected in a specific time period;
and the rejecting unit 712 is configured to perform absolute deviation calculation on the electrical data of each device to obtain an outlier, and reject the outlier from each data point to obtain a data point sequence.
In this embodiment, the culling unit 712 is specifically configured to: sorting based on the numerical value of each piece of equipment electrical data to obtain the median of the equipment electrical data, and calculating the difference value between each piece of equipment electrical data and the median to obtain the absolute deviation from each piece of equipment electrical data to the median; comparing each absolute deviation with a preset safety deviation, and judging whether the absolute difference is larger than the preset safety deviation or not; and if the absolute deviation is larger than the preset value, determining the equipment electrical data corresponding to the absolute deviation as an outlier.
In this embodiment, the prediction module 720 includes:
a first operation unit 721, configured to perform an operation on the data point sequence based on an exponential smoothing algorithm, so as to obtain a primary smoothed value of the data point sequence;
A second operation unit 722 for obtaining a second smoothed value of the data point sequence based on the first smoothed value and the exponential smoothing algorithm;
and an input unit 723, configured to input the primary smoothed value and the secondary smoothed value to a preset time sequence prediction model, so as to obtain a feature quantity sequence of the data point sequence.
In this embodiment, the input unit 723 is specifically configured to: and determining a parameter variable value of the current time period based on the primary smoothing value and the secondary smoothing value, taking the parameter variable value as the input of a secondary exponential smoothing model, and outputting a predicted value of the current time period to obtain a characteristic quantity sequence of the data point sequence.
In this embodiment, the input unit 723 is further configured to: sequentially acquiring algorithm parameters from corresponding preset numerical value intervals based on preset units, and sequentially combining the algorithm parameters, wherein the algorithm parameters comprise a smoothing coefficient, a smoothing algorithm initial value and a predicted lead period number; based on the combined algorithm parameters, generating a corresponding characteristic quantity waveform curve, and determining an optimal algorithm parameter corresponding to the characteristic quantity waveform curve through an optimization algorithm; and taking the optimal algorithm parameter as a target algorithm parameter of a preset time sequence prediction model.
In this embodiment, the determining module 730 includes:
an obtaining unit 731, configured to obtain electrical data of the device in a normal state collected in the current circuit, and calculate an average value in the normal state;
a comparing unit 732, configured to compare an absolute value of the feature data sequence with the absolute value to obtain a lifting ratio of the feature sequence, and determine whether the lifting ratio exceeds a preset safety interval;
the result judging unit 733 is configured to input the feature sequence to a preset deep learning model if the feature sequence is not exceeded, so as to obtain a model output result; and if the output result is a fault, confirming the arc fault of the equipment.
In this embodiment, the result determining unit 733 is specifically configured to: acquiring the lifting ratio of the characteristic quantity sequences in at least two electrode materials, and judging whether the lifting ratio exceeds a preset safety interval or not to obtain a judging result; and comparing the judging result with an actual circuit fault result, and training a deep learning detection model to obtain a preset deep learning model.
The scheme has strong adaptability, the prediction model can automatically identify the change of the data mode to adjust, the accuracy of system detection is improved, the scheme can adapt to various data modes, and the popularization is strong.
Fig. 6 to 7 above describe the arc fault detection apparatus for a medium device in the embodiment of the present invention in detail from the point of view of modularized functional entities, and the arc fault detection apparatus for a medium device in the embodiment of the present invention is described in detail from the point of view of hardware processing.
Fig. 8 is a schematic structural diagram of a device arc fault detection device provided in an embodiment of the present invention, where the device arc fault detection device 800 may have a relatively large difference due to different configurations or performances, and may include one or more processors (CPU) 810 (e.g., one or more processors) and a memory 820, and one or more storage media 830 (e.g., one or more mass storage devices) storing application programs 833 or data 832. Wherein memory 820 and storage medium 830 can be transitory or persistent. The program stored on the storage medium 830 may include one or more modules (not shown), each of which may include a series of instruction operations for the apparatus arc fault detection apparatus 800. Still further, the processor 810 may be configured to communicate with the storage medium 830 and execute a series of instruction operations in the storage medium 830 on the device arc fault detection device 800 to implement the methods provided by the implementations described above.
The device arc fault detection device 800 may also include one or more power supplies 840, one or more wired or wireless network interfaces 850, one or more input/output interfaces 860, and/or one or more operating devices 831, such as WindowsServe, macOSX, unix, linux, freeBSD, and the like. It will be appreciated by those skilled in the art that the apparatus arc fault detection apparatus structure shown in fig. 8 is not limiting of the apparatus arc fault detection apparatus provided by the present invention and may include more or fewer components than shown, or may be a combination of certain components, or a different arrangement of components.
The present invention also provides a computer readable storage medium, which may be a non-volatile computer readable storage medium, and may also be a volatile computer readable storage medium, where instructions are stored in the computer readable storage medium, where the instructions, when executed on a computer, cause the computer to perform the steps of the arc fault detection method for an apparatus provided in the foregoing embodiments.
It will be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working process of the apparatus or device, unit described above may refer to the corresponding process in the foregoing method embodiment, which is not repeated herein.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a read-only memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, or other various media capable of storing program codes.
The above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (11)
1. The equipment arc fault detection method based on the time sequence prediction model is characterized by comprising the following steps of:
acquiring equipment electrical data in a specific time period, and preprocessing the equipment electrical data to obtain a data point sequence;
based on a preset time sequence prediction model, extracting features of each data point in the data point sequence to obtain a feature quantity sequence of the data point sequence;
and detecting the arc faults of the equipment by using a preset deep learning model and the lifting ratio of the characteristic quantity sequence.
2. The method for detecting an arc fault in a device according to claim 1, wherein the steps of acquiring electrical data of the device in a specific period of time, and preprocessing the electrical data of the device to obtain a data point sequence include:
based on equipment electrical data in a specific time period, which is acquired from a circuit to be detected by acquisition equipment, arranging the equipment electrical data according to a time mark;
and carrying out absolute deviation calculation on the electrical data of each device to obtain outliers, and removing the outliers from each data point to obtain a data point sequence.
3. The method for detecting arc faults of equipment according to claim 2, wherein the step of calculating absolute deviation of electrical data of each piece of equipment to obtain outliers comprises:
sorting based on the numerical value of each piece of equipment electrical data to obtain the median of the equipment electrical data, and calculating the difference value between each piece of equipment electrical data and the median to obtain the absolute deviation from each piece of equipment electrical data to the median;
comparing each absolute deviation with a preset safety deviation, and judging whether the absolute difference is larger than the preset safety deviation or not;
and if the absolute deviation is larger than the preset value, determining the equipment electrical data corresponding to the absolute deviation as an outlier.
4. The method for detecting an arc fault of a device according to claim 1, wherein the feature extraction is performed on each data point in the data point sequence based on a preset time sequence prediction model, so as to obtain a feature quantity sequence of the data point sequence, and the method comprises the following steps:
calculating the data point sequence based on an exponential smoothing algorithm to obtain a primary smoothing value of the data point sequence;
obtaining a secondary smoothed value of the sequence of data points based on the primary smoothed value and the exponential smoothing algorithm;
And inputting the primary smoothing value and the secondary smoothing value into a preset time sequence prediction model to obtain a characteristic quantity sequence of the data point sequence.
5. The method for detecting an arc fault in an apparatus according to claim 4, wherein the step of inputting the primary smoothed value and the secondary smoothed value into a predetermined time series prediction model to obtain the characteristic quantity series of the data point series comprises:
and determining a parameter variable value of the current time period based on the primary smoothing value and the secondary smoothing value, taking the parameter variable value as the input of a secondary exponential smoothing model, and outputting a predicted value of the current time period to obtain a characteristic quantity sequence of the data point sequence.
6. The apparatus arc fault detection method according to claim 4, further comprising, after the inputting the primary smoothed value and the secondary smoothed value to a preset time series prediction model, obtaining a feature quantity series of the data point series:
sequentially acquiring algorithm parameters from corresponding preset numerical value intervals based on preset units, and sequentially combining the algorithm parameters, wherein the algorithm parameters comprise a smoothing coefficient, a smoothing algorithm initial value and a predicted lead period number;
Based on the combined algorithm parameters, generating a corresponding characteristic quantity waveform curve, and determining an optimal algorithm parameter corresponding to the characteristic quantity waveform curve through an optimization algorithm;
and taking the optimal algorithm parameter as a target algorithm parameter of a preset time sequence prediction model.
7. The apparatus arc fault detection method according to claim 1, wherein the arc fault detection of the apparatus is realized by using a preset deep learning model and a lifting ratio of the feature quantity sequence, comprising:
acquiring equipment electrical data in a normal state acquired in a current circuit, and calculating an average value in the normal state;
comparing the absolute value of the characteristic quantity data sequence with the absolute value to obtain a lifting ratio of the characteristic quantity sequence, and judging whether the lifting ratio exceeds a preset safety interval or not;
if the feature quantity sequence is not exceeded, inputting the feature quantity sequence into a preset deep learning model to obtain a model output result;
and if the output result is a fault, confirming the arc fault of the equipment.
8. The apparatus arc fault detection method according to claim 1, further comprising, before the inputting the feature sequence to a preset deep learning model, obtaining a model output result:
Acquiring the lifting ratio of the characteristic quantity sequences in at least two electrode materials, and judging whether the lifting ratio exceeds a preset safety interval or not to obtain a judging result;
and comparing the judging result with an actual circuit fault result, and training a deep learning detection model to obtain a preset deep learning model.
9. A device arc fault detection apparatus based on a time series prediction model, the device arc fault detection apparatus comprising:
the preprocessing module is used for acquiring equipment electrical data in a specific time period, and preprocessing the equipment electrical data to obtain a data point sequence;
the prediction module is used for extracting the characteristics of each data point in the data point sequence based on a preset time sequence prediction model to obtain a characteristic quantity sequence of the data point sequence;
and the determining module is used for detecting the arc faults of the equipment by utilizing a preset deep learning model and the lifting ratio of the characteristic quantity sequence.
10. A device arc fault detection device based on a time series prediction model, wherein the device arc fault detection device comprises a memory and at least one processor, the memory having instructions stored therein; the at least one processor invokes the instructions in the memory to cause the device arc fault detection device to perform the steps of the device arc fault detection method of any of claims 1-8.
11. A computer readable storage medium having instructions stored thereon, which when executed by a processor, implement the steps of the method for arc fault detection of a device based on a time series prediction model as claimed in any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311178746.7A CN117330906A (en) | 2023-09-08 | 2023-09-08 | Equipment arc fault detection method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311178746.7A CN117330906A (en) | 2023-09-08 | 2023-09-08 | Equipment arc fault detection method, device, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117330906A true CN117330906A (en) | 2024-01-02 |
Family
ID=89289350
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311178746.7A Pending CN117330906A (en) | 2023-09-08 | 2023-09-08 | Equipment arc fault detection method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117330906A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117743808A (en) * | 2024-02-20 | 2024-03-22 | 中铁西南科学研究院有限公司 | Tunnel deformation prediction method, system, equipment and medium |
| CN117743808B (en) * | 2024-02-20 | 2024-05-14 | 中铁西南科学研究院有限公司 | Tunnel deformation prediction method, system, equipment and medium |
-
2023
- 2023-09-08 CN CN202311178746.7A patent/CN117330906A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117743808A (en) * | 2024-02-20 | 2024-03-22 | 中铁西南科学研究院有限公司 | Tunnel deformation prediction method, system, equipment and medium |
| CN117743808B (en) * | 2024-02-20 | 2024-05-14 | 中铁西南科学研究院有限公司 | Tunnel deformation prediction method, system, equipment and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108429649B (en) | System for comprehensive abnormity judgment based on multiple single-type acquisition results | |
| JP5827425B1 (en) | Predictive diagnosis system and predictive diagnosis method | |
| JP5827426B1 (en) | Predictive diagnosis system and predictive diagnosis method | |
| CN111783696A (en) | Edge calculation method for low-voltage branch topology real-time analysis based on PV relation | |
| CN110991527B (en) | Similarity threshold determining method considering average fluctuation rate of voltage curve | |
| CN112213687B (en) | Gateway electric energy meter data anomaly detection method and system based on pseudo-anomaly point identification | |
| CN115664038B (en) | Intelligent power distribution operation and maintenance monitoring system for electrical safety management | |
| CN111898637B (en) | Feature selection algorithm based on Relieff-DDC | |
| CN116865269A (en) | Wind turbine generator system high harmonic compensation method and system | |
| CN117330906A (en) | Equipment arc fault detection method, device, equipment and storage medium | |
| CN113219333B (en) | Spectrum parameter processing method during motor fault diagnosis | |
| CN108171425B (en) | Power quality partitioning method and device and storage medium | |
| EP4001936A1 (en) | Automatic partial discharge and noise signals separation using arithmetic coding in time domain and magnitude distributions in frequency domain | |
| CN115963350A (en) | Fault positioning method and device for direct-current power distribution network | |
| CN115935285A (en) | Multi-element time series anomaly detection method and system based on mask map neural network model | |
| JP5771318B1 (en) | Abnormality diagnosis apparatus and abnormality diagnosis method | |
| CN114548173A (en) | Accurate periodicity detection method, system and storage medium | |
| CN117540325B (en) | Business database anomaly detection method and system based on data variation capture | |
| CN111460005B (en) | JSD-based outlier detection method for time sequence data | |
| CN116150666B (en) | Energy storage system fault detection method and device and intelligent terminal | |
| CN117273547B (en) | Production equipment operation data processing method based on edge calculation | |
| CN118013443A (en) | Online real-time vacuum dry pump abnormality detection method based on generation model algorithm | |
| Shilpa et al. | Application of Hilbert–Huang Transform and SVM Classifier to Monitor the Power Quality Disturbances | |
| Furlani Bastos et al. | Optimization Techniques for Mining Power Quality Data and Processing Unbalanced Datasets in Machine Learning Applications. Energies 2021, 14, 463 | |
| CN117215887A (en) | Data prediction method based on ARCH model and Kalman filter |
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 |