Misoperation identification method for high-voltage switch cabinetTechnical Field
The invention relates to the technical field of misoperation identification of high-voltage switch cabinets, in particular to an misoperation identification method of a high-voltage switch cabinet.
Background
In high-dust operation scenes such as mines, cement factories and the like, a high-voltage switch cabinet is exposed to coal ash and metal dust for a long time, micron-sized particles continuously permeate into the interior of the cabinet body to form conductive scale, the surfaces of key parts such as a displacement sensor and a photoelectric sensor which are depended on by the existing misoperation prevention technology are easily covered by dust, for example, when dust is accumulated to the thickness of 0.1mm in an induction window, the scattering rate of infrared light beams is increased rapidly by a handcart position detection device, so that the real displacement of the handcart cannot be identified by a system, and more difficult is that after dust and water vapor in a wet mine hole are mixed, the dust is hardened at a rotating shaft of an interlocking mechanism, so that mechanical clamping stagnation is caused.
In recent years, non-contact monitoring schemes such as millimeter wave radar and laser ranging can reduce physical contact, but the scattered dust in a mine still causes serious attenuation of signals, for example, an intelligent switch cabinet in a publication No. CN216043169U adopts acousto-optic dual alarm, in practical application, voice reminding can be submerged in a strong noise environment, and a flickering indicator lamp is insufficient in the recognition degree of a dim roadway.
In the prior art, when the sensor readings drift, false alarms are directly triggered or data are discarded to cause protection failure, but the nonlinear relation between dust concentration and signal attenuation is not modeled although a part of self-adaptive filtering algorithm can be dynamically calibrated, and the key is that dust is regarded as a pure interference factor, and equipment state information carried by the physical deposition process is ignored, so that a high-voltage switch cabinet misoperation identification method is needed to solve the problems.
Disclosure of Invention
The present invention has been made in view of the above-described problems occurring in the prior art.
The invention provides a method for identifying misoperation of a high-voltage switch cabinet, which solves the problems that sensors fail due to mine dust, false alarm is frequently found due to electromagnetic interference, and the robustness of a severe environment cannot be considered in the existing scheme.
In order to solve the technical problems, the invention provides the following technical scheme:
the embodiment of the invention provides a method for identifying misoperation of a high-voltage switch cabinet, which comprises the following steps,
Step S1, multi-source sensor data are collected, wherein the multi-source sensor data comprise cabinet door displacement signals, motor driving current waveforms, cabinet vibration spectrums and environmental dust concentration data;
Step S2, dynamically compensating and processing the multi-source sensor data to generate anti-interference operation time sequence data;
Step S3, extracting an operation intention feature vector, wherein the operation intention feature vector comprises current ripple envelope line features analyzed from a motor driving current waveform and predetermined frequency band energy features analyzed from a cabinet vibration frequency spectrum;
s4, inputting the operation intention feature vector into a pre-trained operation behavior analysis model, and outputting a misoperation risk level;
And S5, executing a grading interlocking action according to the misoperation risk level.
As a preferable scheme of the method for identifying misoperation of the high-voltage switch cabinet, the step S2 of the dynamic compensation processing comprises the following steps:
Constructing a nonlinear mapping relation between dust concentration and displacement signal attenuation amplitude;
when a sudden change in the attenuation amplitude of the displacement signal is detected, the vibration reconstruction unit is activated.
As a preferable scheme of the misoperation identifying method of the high-voltage switch cabinet, the method comprises the following steps of performing physical conversion from vibration spectrum energy to displacement in a vibration reconstruction unit in the step S2:
energy aggregation is carried out on continuous frequency bands subjected to relevant screening:
Wherein, theThe characteristic frequency band aggregate energy, in dB,The time of sampling, unit s,The characteristic band lower limit frequency, in Hz,The upper frequency limit of the characteristic frequency band, in Hz,The gain compensation coefficient of the passband,The frequency of the integral argument, in Hz,Vibration spectrum energy density, unit dB;
the aggregate energy is converted into a sequence that is linearly related to the mechanical displacement using an exponentially weighted model:
Wherein, theA virtual displacement sequence, in mm,The time of sampling, unit s,The index of the frequency component, dimensionless,The total number of frequency components involved in the mapping, dimensionless,First, theComponent energy displacement conversion weight, unit,First, theThe instantaneous energy density of the components, in dB,First, theThe center frequency of the components, in Hz,Background noise power spectrum, unit dB,Nonlinear energy index, dimensionless, experience range 0.8-1.2;
the weight vector adopts ridge regression closed solving and gives a regularization factor self-adaptive expression:
,
,
Wherein, theWeight vectorUnits of,Sample energy matrix consisting ofThe composition, in dB,Indicating the time at which the sample sequence number corresponds,The index of the sample is represented and,Is to be used in the present invention,The ridge regression is used for regularizing factors, dimensionless,An identity matrix, dimensionless,The actual displacement observation vector, in mm,Regular proportionality coefficient, experience range of 0.01-0.05, dimensionless,Is thatIs set in dB.
As a preferable scheme of the misoperation identifying method of the high-voltage switch cabinet, the execution logic of the vibration reconstruction unit comprises the following steps:
Extracting a characteristic frequency band which is strongly related to mechanical displacement in a vibration frequency spectrum of the cabinet body;
based on the phase information of the motor driving current waveform, carrying out time domain alignment on the energy of the characteristic frequency band;
generating a virtual displacement sequence which is linearly related to the actual displacement of the cabinet door;
The determination mode of the characteristic frequency band which is strongly related to the mechanical displacement is as follows:
collecting a vibration spectrum sample of the switch cabinet handcart in a standard displacement interval;
Calculating the pearson correlation coefficient of the energy and the displacement distance of each frequency component;
And screening continuous frequency bands with correlation coefficients exceeding a first threshold and signal-to-noise ratios exceeding a second threshold.
As an optimal scheme of the misoperation identifying method of the high-voltage switch cabinet, the extracting of the current ripple envelope curve characteristics in the step S3 comprises the following steps:
Carrying out band-pass filtering on the waveform of the motor driving current, and separating out ripple components;
and calculating the rising edge slope of the ripple component and the number of zero crossings in unit time.
In step S3, the anti-interference characteristic calculation of the motor driving current ripple envelope curve is carried out, and the method comprises the following steps:
constructing a band-pass impulse response with the center frequency adjusted along with the load:
,
Wherein, theIs the firstThe impulse response of the point filter is set,For discrete time index, dimensionless,Is the center frequency of the ripple, in Hz,Is a half-bandwidth, unit Hz,For the sampling frequency, in Hz,Is Kaiser window coefficient;
At the filtered outputPerforming Hilbert transform on the signal to obtain an analytic signal:
,
Wherein, theFor the ripple envelope magnitude, unit a,Is the waveform of the current after the band pass, the unit A,In the case of a hilbert transform operator,Is the unit of imaginary number, has no dimension,The unit s is the sampling time;
In the observation windowThe internal calculation of two anti-interference features comprises the following calculation formulas:
,
,
Wherein, theFor envelope energy to be normalized to a proportion, dimensionless,For the envelope rising edge coefficient of variation, dimensionless,For the length of the observation window, units s,Is the firstThe number of sampling instants, units s,For sample indexing, dimensionless,Is the firstThe point envelope magnitude, unit a,As the median absolute deviation operator,Is a median operator, and the final anti-interference feature vector is recorded as。
As an optimal scheme of the misoperation identifying method of the high-voltage switch cabinet, the training of the operation behavior analysis model in the step S4 comprises the following steps:
injecting a standard electromagnetic pulse interference sample into the historical operation data;
the robustness of the vibration spectrum characteristic is enhanced by adopting a frequency domain masking technology;
Convolutional neural network decision boundaries are optimized by countermeasure training.
As a preferable scheme of the misoperation identifying method of the high-voltage switch cabinet, the invention further comprises the following steps:
And (3) switching an environment anti-interference mode, and when the intensity of an environment electromagnetic field exceeds the standard:
Suppressing current ripple envelope features in the operational intent feature vector;
the decision weight of the vibration spectrum characteristics of the cabinet body is improved to a preset value;
The triggering conditions of the environment anti-interference mode switching comprise:
Monitoring the frequency spectrum distribution of the environmental electromagnetic field intensity in real time;
When the continuous exceeding of the energy of the 15kHz-100kHz frequency band is detected, determining an electromagnetic interference scene;
If the exceeding duration exceeds the set value, starting a frequency domain feature priority decision mechanism.
As a preferable scheme of the misoperation identifying method of the high-voltage switch cabinet, the step S5 comprises the following steps:
Triggering the electromagnetic locking device of the cabinet door when the misoperation risk level is one level;
When the risk level of misoperation is two-level, synchronously executing the opening of the breaker and the closing of the grounding switch;
A security event log is generated that contains the feature vector hash value.
As an optimal scheme of the high-voltage switch cabinet misoperation identification method, the method for determining the characteristic frequency band comprises the following steps:
analyzing vibration spectrum distribution in the moving process of the switch cabinet handcart;
Selecting a frequency band with the correlation coefficient of energy change and displacement distance larger than 0.8;
excluding the band dominated by ambient background noise.
The invention has the advantages that the invention aims at sensing failure caused by dust coverage and misjudgment caused by electromagnetic interference, rebuilds a high-robustness displacement track by utilizing cabinet vibration energy when the dust shields a displacement sensor through a physical conversion model of vibration spectrum energy to displacement, breaks through the physical limitation of a traditional optical/magnetic induction scheme, designs a load self-adaptive ripple envelope extraction algorithm aiming at an electromagnetic interference scene, combines anti-interference feature quantification and frequency domain mask training, keeps stable under strong pulse by combining current features, solves the problem of time domain signal drift through motor current phase alignment vibration spectrum, combines ridge regression weight optimization with regularization self-adaption, improves virtual displacement mapping accuracy, dynamically switches decision weights by an environment sensing module, realizes dust-electromagnetic double-scene seamless protection, and changes passive alarm into active interception by millisecond-level risk classification interlocking and electromagnetic locking/breaker switching-off, thereby remarkably reducing the risk of malignant accidents.
The invention multiplexes the existing industrial sensor and the edge computing unit, does not need to modify the structure of the switch cabinet, is especially suitable for deployment in severe environments such as mines, substations and the like, and provides intelligent guarantee for the operation safety of high-voltage equipment.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flow chart of a method for identifying misoperation of a high-voltage switch cabinet in embodiment 1.
Detailed Description
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to specific embodiments thereof which are illustrated in the appended drawings.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways other than those described herein, and persons skilled in the art will readily appreciate that the present invention is not limited to the specific embodiments disclosed below.
Further, reference herein to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic can be included in at least one implementation of the invention. The appearances of the phrase "in one embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments.
Embodiment 1, referring to fig. 1, the embodiment provides a method for identifying misoperation of a high-voltage switch cabinet, which comprises the following steps:
Step S1, multi-source sensor data are collected, wherein the multi-source sensor data comprise cabinet door displacement signals, motor driving current waveforms, cabinet vibration spectrums and environmental dust concentration data;
step S2, dynamically compensating and processing multi-source sensor data to generate anti-interference operation time sequence data;
the dynamic compensation process in step S2 includes:
Constructing a nonlinear mapping relation between dust concentration and displacement signal attenuation amplitude;
when the sudden change of the attenuation amplitude of the displacement signal is detected, activating a vibration reconstruction unit;
In the vibration reconstruction unit of step S2, a physical conversion from vibration spectrum energy to displacement is performed, including:
energy aggregation is carried out on continuous frequency bands subjected to relevant screening:
,
Wherein, theThe characteristic frequency band aggregate energy, in dB,The time of sampling, unit s,The characteristic band lower limit frequency, in Hz,The upper frequency limit of the characteristic frequency band, in Hz,The gain compensation coefficient of the passband,The frequency of the integral argument, in Hz,The vibration spectrum energy density is in dB, the integration compresses Gao Weipu into a single energy track, noise is dissipated in an integration window, the gain compensation corrects the difference of a sensing link, the obtained track is monotonously changed along with the stroke of a door body, smooth trend is still kept under a pulse interference scene, the influence of dust shielding on an energy curve is negligible, and a stable reference is laid for the subsequent displacement mapping;
the aggregate energy is converted into a sequence that is linearly related to the mechanical displacement using an exponentially weighted model:
,
Wherein, theA virtual displacement sequence, in mm,The time of sampling, unit s,The index of the frequency component, dimensionless,The total number of frequency components involved in the mapping, dimensionless,First, theComponent energy displacement conversion weight, unit,First, theThe instantaneous energy density of the components, in dB,First, theThe center frequency of the components, in Hz,Background noise power spectrum, unit dB,The nonlinear energy index, dimensionless and experience range is 0.8-1.2, the index weighting model keeps gradient continuity in a small-amplitude stroke and an acceleration section, and solves the problem of bending of an energy displacement relation under different working conditions;
the weight vector adopts ridge regression closed solving and gives a regularization factor self-adaptive expression:
,
,
Wherein, theWeight vectorUnits of,Sample energy matrix consisting ofThe composition, in dB,Indicating the time at which the sample sequence number corresponds,The index of the sample is represented and,Is to be used in the present invention,The ridge regression is used for regularizing factors, dimensionless,An identity matrix, dimensionless,The actual displacement observation vector, in mm,Regular proportionality coefficient, experience range of 0.01-0.05, dimensionless,Is thatMaximum singular value of (a), unit dB; the closed solution avoids the oscillation of the iterative process, keeps stable convergence when the sample is sparse or the frequency components are highly correlated, and the regularization factor is adaptively adjusted along with the matrix spectrum norm, so that the weight is limited to be too large and the underfitting is prevented;
The execution logic of the vibration reconstruction unit includes:
Extracting a characteristic frequency band which is strongly related to mechanical displacement in a vibration frequency spectrum of the cabinet body;
based on the phase information of the motor driving current waveform, carrying out time domain alignment on the energy of the characteristic frequency band;
generating a virtual displacement sequence which is linearly related to the actual displacement of the cabinet door;
the characteristic frequency band strongly related to the mechanical displacement is determined in the following manner:
collecting a vibration spectrum sample of the switch cabinet handcart in a standard displacement interval;
Calculating the pearson correlation coefficient of the energy and the displacement distance of each frequency component;
screening continuous frequency bands with correlation coefficients exceeding a first threshold and signal-to-noise ratios exceeding a second threshold;
The method for determining the characteristic frequency band comprises the following steps:
analyzing vibration spectrum distribution in the moving process of the switch cabinet handcart;
Selecting a frequency band with the correlation coefficient of energy change and displacement distance larger than 0.8;
Excluding a frequency band dominated by environmental background noise;
Step S3, extracting an operation intention feature vector, wherein the operation intention feature vector comprises current ripple envelope line features analyzed from a motor driving current waveform and predetermined frequency band energy features analyzed from a cabinet vibration frequency spectrum;
the extracting of the current ripple envelope curve in the step S3 includes:
Carrying out band-pass filtering on the waveform of the motor driving current, and separating out ripple components;
calculating the rising edge slope of the ripple component and the zero crossing times in unit time;
In step S3, performing a calculation of an anti-interference characteristic of a ripple envelope of the motor driving current, including:
constructing a band-pass impulse response with the center frequency adjusted along with the load:
,
Wherein, theIs the firstThe impulse response of the point filter is set,For discrete time index, dimensionless,Is the center frequency of the ripple, in Hz,Is a half-bandwidth, unit Hz,For the sampling frequency, in Hz,The self-adaptive center frequency moves synchronously along with the drift of load current to avoid the error filtering of actual ripple wave components, the Kaiser window sidelobe is fast in attenuation, adjacent harmonic wave leakage is restrained, half bandwidth is dynamically converged according to the statistical result of ripple wave energy distribution, stable passband gain can be kept in the electromagnetic interference rising period, the filtered waveform only keeps current reversing ripple waves, the attenuation of fundamental waves and peak pulse exceeds 30dB, and additional notch compensation is not needed in the follow-up envelope solving;
At the filtered outputPerforming Hilbert transform on the signal to obtain an analytic signal:
,
Wherein, theFor the ripple envelope magnitude, unit a,Is the waveform of the current after the band pass, the unit A,In the case of a hilbert transform operator,Is the unit of imaginary number, has no dimension,The method is characterized in that the method comprises the steps of sampling time, analyzing a signal method to give instantaneous amplitude and phase once, avoiding double-channel errors caused by upper envelope splitting and lower envelope splitting, enabling Hilbert phase shift to be kept at 90 degrees by symmetrical filtering, inhibiting boundary effects by zero-phase continuation, enabling envelope track and mechanical impact period to have high coupling degree, enabling sporadic pulses to be converted into isolated sharp points, facilitating statistics and filtering, and improving the credibility of subsequent robust indexes;
In the observation windowThe internal calculation of two anti-interference features comprises the following calculation formulas:
,
,
Wherein, theFor envelope energy to be normalized to a proportion, dimensionless,For the envelope rising edge coefficient of variation, dimensionless,For the length of the observation window, units s,Is the firstThe number of sampling instants, units s,For sample indexing, dimensionless,Is the firstThe point envelope magnitude, unit a,As the median absolute deviation operator,Is a median operator, and the final anti-interference feature vector is recorded asWhere (1)Comparing the mean square amplitude with the arithmetic mean amplitude, the ripple distortion caused by the armature commutation abnormality can be rapidly identified,MAD is adopted to describe rising edge jitter, so that the pulse interference and sporadic burrs are natural and stable, and the two indexes are dimensionless and are convenient for cross-equipment normalization;
s4, inputting the operation intention feature vector into a pre-trained operation behavior analysis model, and outputting a misoperation risk level;
the training of the operation behavior analysis model in step S4 includes:
injecting a standard electromagnetic pulse interference sample into the historical operation data;
the robustness of the vibration spectrum characteristic is enhanced by adopting a frequency domain masking technology;
Optimizing a convolutional neural network decision boundary through countermeasure training;
Step S5, performing a grading interlocking action according to the misoperation risk level;
The step S5 of step interlocking action includes:
Triggering the electromagnetic locking device of the cabinet door when the misoperation risk level is one level;
When the risk level of misoperation is two-level, synchronously executing the opening of the breaker and the closing of the grounding switch;
Generating a security event log containing the feature vector hash value;
The method further comprises the steps of:
And (3) switching an environment anti-interference mode, and when the intensity of an environment electromagnetic field exceeds the standard:
Suppressing current ripple envelope features in the operational intent feature vector;
the decision weight of the vibration spectrum characteristics of the cabinet body is improved to a preset value;
the triggering conditions of the environment anti-interference mode switching include:
Monitoring the frequency spectrum distribution of the environmental electromagnetic field intensity in real time;
When the continuous exceeding of the energy of the 15kHz-100kHz frequency band is detected, determining an electromagnetic interference scene;
If the exceeding duration exceeds the set value, starting a frequency domain feature priority decision mechanism.
It should be noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention may be modified or substituted without departing from the spirit and scope of the technical solution of the present invention, which is intended to be covered in the scope of the claims of the present invention.