Fire-fighting equipment power supply detection method and system

Abstract

The invention relates to the technical field of power supply detection, and discloses a fire-fighting equipment power supply detection method and system.The detection method comprises the following steps that voltage fluctuation, current phase and ripple coefficients are collected through a multi-channel sensor group, a feature sequence is generated through sliding window noise reduction, the Euclidean distance is calculated through Z-score standardization, and a calibration instruction is generated when a threshold value is exceeded; kalman filtering adjusts frequency output detection parameters, fast Fourier transform is executed to extract harmonic amplitude to generate a suppression vector, and a least square reconstruction model outputs a state identifier; according to the method, voltage, current and ripples are collected through multiple channels, anti-interference performance and data integrity are enhanced through sliding window noise reduction, a threshold value is adjusted through Z-score and Euclidean distance self-matching, sampling frequency and detection parameters are optimized through Kalman filtering, harmonic characteristics are extracted through FFT, and the state is evaluated through a least square reconstruction model. And a multi-dimensional classification system is constructed by fusing the distortion rate and the phase deviation, so that the detection precision, the sensitivity and the response real-time performance are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power supply detection, and in particular relates to a method and system for detecting power supply of fire-fighting equipment. Background Art

[0002] The field of power supply detection technology mainly involves methods and devices for real-time monitoring and diagnosis of the power supply status of various types of equipment. Its core issues include the collection, data transmission, analysis and judgment of power supply parameters such as voltage, current, and frequency, as well as the identification and early warning mechanism of abnormal power supply status. In the power supply detection application of fire-fighting equipment, this technical field pays particular attention to whether the equipment is in a normal power supply state and whether the emergency power supply can be connected in time in the event of a power outage to ensure the reliable operation of the fire-fighting system. This field also covers the long-term monitoring and management of key performance parameters such as power supply stability and backup power switching response time, and is widely used in multiple scenarios such as building fire protection, power operation and maintenance, and safety supervision.

[0003] Among them, the traditional fire-fighting equipment power supply detection method refers to a method of judging the on and off status of the main power supply and the backup power supply through regular manual inspections or the use of relay detection devices. This type of method judges whether the power supply is normal based on whether the voltmeter measurement value is within the preset threshold range and whether the backup power supply can be successfully switched by the mechanical switching mechanism. Some methods trigger periodic detection of the power supply status by setting a timed inspection program.

[0004] Traditional detection relies on manual inspections and mechanical switching devices, using fixed thresholds to determine the voltage on / off status. The fixed detection interval makes it impossible to capture sudden voltage fluctuations in a timely manner. The mechanical switching mechanism is constrained by physical inertia, and the backup power supply switching response is delayed. The scheduled inspection program cannot match the natural drift of power supply parameters, and false alarms occur frequently. The voltmeter only monitors instantaneous values ​​and lacks continuous tracking of parameters such as ripple coefficient and phase offset. It is impossible to identify hidden faults caused by harmonic distortion. Long-term stability assessment relies on manual experience, and it is difficult to quantify the power supply performance degradation trend, affecting the power supply reliability of the fire protection system. Summary of the Invention

[0005] The present invention can solve the technical problems raised in the background technology and provide a fire-fighting equipment power supply detection method and system. Through multi-channel acquisition of voltage, current and ripple, sliding window noise reduction enhances anti-interference and data integrity, Z-score and Euclidean distance self-matching adjust threshold, Kalman filter optimizes sampling frequency and detection parameters, fast Fourier transform extracts harmonic features, least squares reconstruction model evaluates status, and distortion rate and phase deviation are integrated to construct a multidimensional classification system, thereby improving detection accuracy, sensitivity and real-time response.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: A method for detecting power supply of fire-fighting equipment comprises the following steps: S1: The voltage fluctuation sequence, current phase offset, and ripple coefficient of the power supply are collected through a multi-channel sensor group, and a sliding window mechanism is used for noise reduction to generate a reference power supply characteristic sequence; S2: Call the benchmark power feature sequence, perform Z-score normalization on the voltage fluctuation sequence, calculate the Euclidean distance with the feature sequence, and generate a power state calibration instruction when the three distance values ​​simultaneously exceed the voltage difference judgment threshold; S3: Based on the power state calibration instruction, the Kalman filter algorithm is used to dynamically adjust the sampling frequency. According to the current phase offset change rate and the second-order derivative of the ripple coefficient, a power parameter time-varying model is constructed. The power state evolution is dynamically fed back and a self-matching detection parameter group is output. S4: performing a fast Fourier transform on the main frequency component of the ripple spectrum in the reference power supply characteristic sequence according to the self-matching detection parameter group to extract the harmonic amplitude. When the harmonic amplitude exceeds a set ratio of the fundamental amplitude, a harmonic suppression strategy vector is generated. S5: Based on the harmonic suppression strategy vector, the least squares method is used to reconstruct the power supply evaluation model, input voltage fluctuation range, current phase cumulative deviation, total harmonic distortion rate, and output power status classification identification.

[0007] The following is a further optimization of the above technical solution by the present invention: The benchmark power supply feature sequence includes the Z-score standardized voltage sequence, the phase offset mean sequence and the ripple coefficient smoothing sequence; the power supply state calibration instructions include the voltage fluctuation threshold, the phase offset tolerance and the ripple tolerance; the self-matching detection parameter group includes the dynamic sampling frequency parameter, the phase change rate threshold and the ripple second-order derivative parameter; the harmonic suppression strategy vector includes the harmonic attenuation coefficient group, the fundamental amplitude protection threshold and the spectrum weight factor; the power supply state classification identification includes the voltage extreme difference level, the phase cumulative deviation range and the total harmonic distortion rate level.

[0008] Further optimization: The specific steps of S1 include: S101: Using a multi-channel sensor group, collect the real-time voltage value of the power supply, the phase offset of the current signal, and the ripple data of the corresponding frequency band in the voltage signal. Perform synchronization timestamp correction on the channel data under a unified time base, construct a channel-by-channel data structure, perform continuity check and missing value marking, and obtain an integrated synchronous data sequence. S102: Constructing a sliding window of equal width based on the integrated synchronous data sequence, setting the step size and window length, performing interpolation correction, deviation elimination, and smoothing processing, eliminating interference segments with sudden changes and short-term high-frequency spike data, and obtaining characteristic sequence data after noise reduction processing; S103: Call the characteristic sequence data after noise reduction processing, extract the phase difference, fluctuation interval change ratio and ripple amplitude ratio between multi-channel signals in the same time window, perform linear normalization processing and then perform vector recombination to generate a reference power supply characteristic sequence.

[0009] Further optimization: The specific steps of S2 include: S201: Calling a reference power supply characteristic sequence, synchronously taking values ​​of the voltage values ​​at the same position in the current voltage fluctuation sequence based on the reference voltage data in the time axis, constructing a corresponding relationship sequence, and sequentially comparing the amplitudes of the two sequences to establish a voltage characteristic response sequence; S202: Based on the window data of the voltage characteristic response sequence, call the medium-length segment data in the reference power characteristic sequence, process the synchronization point values ​​of the three dimensions in sequence, calculate the degree of difference between the three groups of values ​​using the Euclidean distance formula, and generate a three-dimensional voltage difference value; S203: Based on the three distance values ​​in the three-dimensional voltage difference value, compare the preset voltage difference judgment threshold to determine whether they exceed the corresponding threshold at the same time. When all three judgment conditions are met, mark the window voltage state and generate a power state calibration instruction.

[0010] Further optimization: The specific steps for S3 include: S301: Based on the power state calibration instruction, the current and voltage waveform sequences within the continuous cycle are collected, the sampling and prediction error covariance matrix is ​​constructed, the weights are calculated according to the observation and prediction errors, the frequency is corrected using the Kalman filter algorithm, and a dynamic frequency adjustment coefficient is generated; S302: Calculate the phase difference change rate based on the dynamic frequency adjustment coefficient and perform second-order difference, extract the phase offset trend and construct a ripple correlation data set, compare the trend node with the ripple response analysis, and generate a phase change fitting coefficient set; S303: Calling the phase change fitting coefficient group, constructing the cycle-to-cycle difference weight function and updating the fitting sequence, performing iterative calculation to obtain the smoothing coefficient sequence and matching the detection parameter combination to obtain the self-matching detection parameter group.

[0011] Further optimization: The specific steps of S4 include: S401: Obtain the main frequency component of the ripple spectrum in the reference power supply characteristic sequence based on the self-matching detection parameter group, perform fast Fourier transform to extract the energy concentration frequency point, fit the harmonic frequency distribution vector, calculate the frequency distribution deviation characteristic value, and obtain the frequency distribution value group; S402: Based on the frequency distribution value group, detect the amplitude component of the harmonic frequency position, calculate the amplitude ratio of the harmonic frequency component to the fundamental wave amplitude, filter the frequency bands above the ratio threshold, and obtain the harmonic amplitude exceeding the threshold frequency band value; S403: calling the frequency band value where the harmonic amplitude exceeds the threshold, matching the response rule template in the self-matching detection parameter group, selecting the suppression strategy unit according to the mapping relationship between the frequency band and the suppression strategy, and generating a harmonic suppression strategy vector.

[0012] Further optimization: The formula used to calculate the frequency distribution deviation eigenvalue is: ; in, Representative The frequency distribution deviation characteristic value of the frequency point is a dimensionless parameter. Represents the number of energy-concentrated frequency points corresponding to the frequency point, which is a dimensionless parameter. Representative The frequency point is The spectral weight factor at the frequency position is a dimensionless parameter. Representative The frequency point is The spectrum power value of the frequency position, in W, Representative The average value of the spectrum power value of the frequency position frequency point, the unit is W, Representative The variance of the spectrum power value of the frequency position frequency point, the unit is W², Representative The average value of the ripple frequency amplitude corresponding to the frequency position frequency point, the unit is V, Represents equivalent impedance, in Ω.

[0013] Further optimization: The specific steps of S5 include: S501: Based on the harmonic suppression strategy vector, the voltage fluctuation range and the current phase cumulative deviation are extracted. The voltage-current correlation equation is established using the least squares method. The residual sum of squares and goodness of fit are calculated to generate the power supply evaluation model parameters. S502: Call the power supply evaluation model parameters, input the total harmonic distortion rate to expand the model, reconstruct the weight matrix through the least squares method, verify the sensitivity coefficient of the harmonic component to the model output, and obtain the optimized power supply evaluation model; S503: Match the output value of the optimized power supply evaluation model with the preset state threshold interval, perform numerical interval membership calculation and discrete coding mapping, and obtain a power supply state classification identifier.

[0014] Further optimization: The specific calculation formula of the sensitivity coefficient is: ; in, Representative The model output sensitivity coefficient corresponding to the quasi-harmonic component is a dimensionless parameter. Representative Class model The weight coefficient of each harmonic frequency band is a dimensionless parameter. Representative The amplitude of each harmonic frequency band, in V, Represents the average value of the harmonic frequency band amplitude, the unit is V, Represents the standard deviation of voltage amplitude, in V, Representative Stable offset parameter for amplitude fluctuation regulation in class models, Representative The power spectral density of each frequency band, in W / Hz, Represents the average value of the power spectral density of the frequency band, in W / Hz, Represents the standard deviation of power spectral density, unit is W / Hz, Represents the total number of harmonic frequency bands involved in the model and is a dimensionless parameter.

[0015] The present invention also provides a fire-fighting equipment power supply detection system, which is used to implement the above-mentioned fire-fighting equipment power supply detection method, and the system includes: The signal acquisition module is used to obtain the voltage fluctuation sequence, current phase offset and ripple coefficient through a multi-channel sensor group, and use a sliding window mechanism to perform noise reduction on the original signal to generate a reference power supply feature sequence, which is then passed to the feature calibration module; The feature calibration module is used to call the benchmark power feature sequence to perform Z-score normalization on the voltage fluctuation sequence, calculate the Euclidean distance between the current window voltage value, current phase offset, ripple coefficient and the feature sequence, and generate a power state calibration instruction when the three distance values ​​simultaneously exceed the set threshold and pass it to the parameter adaptation module; The parameter adaptation module is used to call the power state calibration instruction, dynamically adjust the sampling frequency using the Kalman filter algorithm, build a time-varying power parameter model based on the current phase offset change rate and the second-order derivative of the ripple coefficient, and iteratively calculate and output a self-matching detection parameter group, which is passed to the harmonic analysis module; The harmonic analysis module is used to call the self-matching detection parameter group, perform fast Fourier transform on the main frequency component of the ripple spectrum in the reference power supply characteristic sequence, extract the harmonic amplitude and compare it with the fundamental amplitude. When the harmonic amplitude exceeds the set ratio of the fundamental amplitude, a harmonic suppression strategy vector is generated and passed to the state assessment module; The state assessment module is used to call the harmonic suppression strategy vector, reconstruct the power supply assessment model using the least squares method, input the voltage fluctuation range, current phase cumulative deviation, and total harmonic distortion rate parameters, calculate them through the model, and output the power supply state classification label.

[0016] The present invention adopts the above technical solution, which has at least the following beneficial effects: The present invention uses a multi-channel sensor group to synchronously collect voltage fluctuation sequences, current phase offsets and ripple coefficients, combines sliding window noise reduction to generate a benchmark feature sequence, improves data anti-interference and integrity, Z-score standardization and Euclidean distance dynamic calibration mechanism to achieve threshold self-matching adjustment, Kalman filter algorithm to build a time-varying model to optimize sampling frequency and detection parameters, match the nonlinear change characteristics of the power supply, fast Fourier transform to extract harmonic spectrum components, least squares method to reconstruct the evaluation model, and comprehensively establish a classification system based on multi-dimensional parameters such as harmonic distortion rate and phase cumulative deviation, thereby enhancing the sensitivity of abnormal state detection and classification accuracy, reducing the risk of misjudgment caused by environmental noise interference, and improving diagnostic stability and real-time response under complex working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the workflow of the present invention; Figure 2 It is a system flow chart of the present invention. DETAILED DESCRIPTION

[0018] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations; any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs; to be precise, the use of the word "example" is intended to present concepts in a concrete way; in addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0019] In the embodiments of the present invention, “image” and “picture” may sometimes be used interchangeably. It should be noted that when the distinction between them is not emphasized, the meanings they intend to express are consistent; “of”, “corresponding” and “corresponding” may sometimes be used interchangeably. It should be noted that when the distinction between them is not emphasized, the meanings they intend to express are consistent.

[0020] In the embodiments of the present invention, sometimes a subscript such as W1 may be written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are the same.

[0021] In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0022] See also Figure 1 A fire-fighting equipment power supply detection method, the processing flow of the method includes the following steps: S1: The voltage fluctuation sequence, current phase offset, and ripple coefficient of the power supply are collected through a multi-channel sensor group, and a sliding window mechanism is used for noise reduction to generate a reference power supply characteristic sequence; S2: Call the benchmark power feature sequence, perform Z-score normalization on the voltage fluctuation sequence, calculate the Euclidean distance with the feature sequence, and generate a power state calibration instruction when the three distance values ​​simultaneously exceed the voltage difference judgment threshold; S3: Based on the power state calibration instruction, the Kalman filter algorithm is used to dynamically adjust the sampling frequency. According to the current phase offset change rate and the second-order derivative of the ripple coefficient, a power parameter time-varying model is constructed. The power state evolution is dynamically fed back and a self-matching detection parameter group is output. S4: performing a fast Fourier transform on the main frequency component of the ripple spectrum in the reference power supply characteristic sequence according to the self-matching detection parameter group to extract the harmonic amplitude. When the harmonic amplitude exceeds a set ratio of the fundamental amplitude, a harmonic suppression strategy vector is generated. S5: Based on the harmonic suppression strategy vector, the least squares method is used to reconstruct the power supply evaluation model, input voltage fluctuation range, current phase cumulative deviation, total harmonic distortion rate, and output power status classification identification.

[0023] The benchmark power supply feature sequence includes the Z-score standardized voltage sequence, the phase offset mean sequence and the ripple coefficient smoothing sequence; the power supply state calibration instructions include the voltage fluctuation threshold, the phase offset tolerance and the ripple tolerance; the self-matching detection parameter group includes the dynamic sampling frequency parameter, the phase change rate threshold and the ripple second-order derivative parameter; the harmonic suppression strategy vector includes the harmonic attenuation coefficient group, the fundamental amplitude protection threshold and the spectrum weight factor; the power supply state classification identification includes the voltage extreme difference level, the phase cumulative deviation range and the total harmonic distortion rate level.

[0024] In this embodiment, the specific steps of S1 include: S101: Collect the real-time voltage value of the power supply, the phase offset of the current signal, and the ripple data of the corresponding frequency band in the voltage signal through a multi-channel sensor group, perform synchronous timestamp correction on the channel data under a unified time base, construct a channel-by-channel data structure, perform continuity check and missing value marking, and obtain an integrated synchronous data sequence.

[0025] In the process of collecting the real-time voltage value of the power supply, the phase offset of the current signal, and the ripple data of the corresponding frequency band in the voltage signal through the multi-channel sensor group, it is necessary to first connect each power supply circuit to the corresponding collection node. For example, in the fire power supply system of a high-rise building, three power supply channels A, B, and C are set up. Each channel is connected to the three-phase power supply circuit, and the real-time voltage value of each phase is collected separately using a multi-function electrical parameter sensor with a Hall sensor module. (unit is V) and current signal (Unit is A), and at the same time, the voltage waveform signal is collected and the ripple amplitude in the ripple frequency band (such as 120Hz±10Hz) is extracted through a digital bandpass filter. The ripple data needs to be Fourier transformed to extract the amplitude of a specific frequency band and use it as the ripple feature. The sampling frequency is set to 10kHz to ensure the accuracy of capturing short-term anomalies. During the acquisition process, the time synchronization of the channel acquisition module must be ensured. The high-precision real-time clock (RTC) signal of the unified timing module can be used as the reference signal source to uniformly calibrate and correct the channel timestamp, so that each group Data corresponds to the same collection time When constructing the channel data structure, use the timestamp as the primary key and establish a nested dictionary structure or matrix data index for the A, B, and C channels respectively. When performing continuity check, check whether there is a sampling data gap in each channel within a fixed time interval (such as 100ms). If so, record it as a missing point and mark it with interpolation. Set the identification field in the missing point mark. Indicates missing data. Indicates that the data is complete; ultimately completed under a unified time base Subsampling, forming ( is the number of channels) of the matrix data structure.

[0026] In order to more intuitively reflect the sampling instance, some sampling point data can be listed in the following data structure: Table 1: Multi-channel power supply parameter sampling table (sampling interval is 100ms) Sampling time (ms) Channel A voltage (V) A channel current (A) A ripple amplitude (mV) B channel voltage (V) B channel current (A) B ripple amplitude (mV) C channel voltage (V) C channel current (A) C ripple amplitude (mV) 0 220.5 2.15 28.3 220.2 2.10 29.1 219.8 2.18 30.0 100 220.4 2.16 27.9 220.0 2.11 28.8 219.7 2.19 29.7 200 220.6 2.15 28.5 220.3 2.12 28.9 219.9 2.20 30.2 As shown in Table 1, by sampling the voltage, current and ripple amplitude of different channels at regular intervals, a multi-channel parameter data set at the same time can be constructed; when performing missing value marking, for example, if the data of channel A is lost at time 300ms, it will be recorded as , and set the subsequent interpolation or missing processing process; in the process of synchronous timestamp correction, if the sampling time error of a channel exceeds the ±5ms threshold, it will be adjusted by comparing the RTC benchmark to make the channel sampling data unified to ms time point.

[0027] In addition, the acquisition of phase offset requires the introduction of the time difference between the current phase current and the zero crossing point of the corresponding voltage waveform. , which is then converted into phase difference ,in is the signal period. If the frequency is 50Hz, then ms, if ms, then , this angle is the phase offset between the current channel voltage and current. The phase offset collected by the channel is also uniformly added to the data structure.

[0028] The results show that by implementing unified and synchronous sampling of indicators such as voltage, current, phase, and ripple during the acquisition phase, and performing integrity verification and missing annotation, a channel-oriented, time-consistent data infrastructure can be constructed to provide accurate data support for subsequent data window processing and feature extraction stages.

[0029] S102: Construct an equal-width sliding window based on the integrated synchronous data sequence, set the step size and window length, perform interpolation correction, deviation elimination and smoothing processing, eliminate interference segments with sudden changes and short-term high-frequency spike data, and obtain characteristic sequence data after noise reduction processing.

[0030] To build a sliding window of equal width based on the integrated synchronous data sequence, we first need to set the step size and length of the window; assuming the window length is , the step size is ,in Represents the time range covered by each sliding window, is the time interval when the window moves, for example: ms, ms, each window will cover 200ms of sampled data, and the window sliding step is 100ms; each sliding window covers the sampling time from the sampling moment arrive The data within the time period is continuously sliding until the traversal of the synchronous data sequence is completed.

[0031] In the process of interpolation correction, if the data at a certain moment is missing or discontinuous, the linear interpolation method can be used to repair the missing data; , the calculation formula of linear interpolation is: ; This formula means that at the missing data point, the missing data is filled by the average of the two previous and next data points; for example, if a channel is at time The sampling value of ms is missing, and the sampling data before and after this moment are and , can be obtained by interpolation .

[0032] The purpose of deviation elimination and smoothing is to remove extreme mutation values ​​and short-term high-frequency spike data to ensure the smoothness of the data. The data can be filtered by setting a threshold. For example, if the amplitude change of a data point exceeds 10% of the fluctuation range (i.e. ), the data is considered to be an outlier and needs to be removed or smoothed; through smoothing, such as the moving average method, each data point is smoothed using the average value of a window size of 5, for example, for the moment ms sampling data , can be replaced by the average of the five data points before and after to obtain the smoothed data: ; For example, if at time The sampling data of ms is , and the data of the 5 moments before and after are , then the smoothed result is: ; After the above-mentioned interpolation correction, deviation elimination and smoothing processing, the noise-reduced characteristic sequence data is finally obtained, which can more accurately reflect the characteristic changes of the power supply.

[0033] S103: Call the characteristic sequence data after noise reduction processing, extract the phase difference, fluctuation interval change ratio and ripple amplitude ratio between multi-channel signals in the same time window, perform linear normalization processing and then perform vector recombination to generate a reference power supply characteristic sequence.

[0034] After obtaining the denoised characteristic sequence data, it is necessary to extract the phase difference, fluctuation interval change ratio and ripple amplitude ratio between the multi-channel signals. First, for the extraction of the phase difference, it is necessary to calculate the phase of the voltage and current signals of the channel. Assuming that the voltage signal and the current signal in the same time window are Fourier transformed, the phases obtained are and , the phase difference can be calculated by the following formula: ; If the phase difference between the voltage and current signals of a channel is , the other channel is , then the phase difference between them is .

[0035] The calculation of the fluctuation range change ratio requires measuring the degree of change of the signal fluctuation range within a certain time window, which can be expressed as the ratio between the maximum and minimum values. In a certain time window, assuming that the maximum value of the signal is , the minimum value is , fluctuation range change ratio Then: ; For example, the maximum value of a signal in a time window is , the minimum value is , then the fluctuation range change ratio is: ; The calculation of the ripple amplitude ratio depends on the ripple amplitude of the channel ; Assume that the ripple amplitudes of channels A, B, and C are mV, mV, mV, ripple amplitude ratio It can be calculated by the relative ratio between the channel ripple amplitudes; for example, the ripple amplitude ratio between A and B is: ; After feature extraction is completed, linear normalization is used to map the eigenvalues ​​to the range of [0, 1]. For example, through normalization, the phase difference, fluctuation range change ratio, and ripple amplitude ratio are combined into a new vector feature sequence, and a reference power supply feature sequence is obtained through vector recombination. This reference power supply feature sequence can be used for subsequent analysis tasks such as power supply status monitoring and fault diagnosis.

[0036] In this embodiment, the specific steps of S2 include: S201: Call the reference power supply characteristic sequence, synchronously obtain the voltage value at the same position in the current voltage fluctuation sequence according to the reference voltage data in the time axis, build a corresponding relationship sequence, and compare the amplitudes of the two sequences in turn to establish a voltage characteristic response sequence.

[0037] When calling the benchmark power characteristic sequence, you must first determine the position index interval of the voltage fluctuation sequence to be analyzed in the current time axis, such as selecting the sampling period ms, the corresponding sampling frequency is 1kHz, then it is necessary to take out the voltage values ​​from the 200th to 400th sampling points as the reference point index, and then extract the reference voltage data at the same position from the reference power characteristic sequence to construct a set of corresponding voltage value pairs. In the actual implementation process, for example, the current voltage fluctuation sequence is taken as V, the voltage at the same position in the reference power sequence is V, then the corresponding relationship sequence constructed is: .

[0038] On this basis, amplitude comparison is performed, and the difference calculation of multiple groups of voltage values ​​is performed in sequence. Specifically, the current voltage value in each group is subtracted from the reference voltage value one by one, and the absolute difference is recorded as the comparison result. For example, the difference of the first group is V, the second group is V, and so on, to get a set of difference sequences V, constitutes a voltage characteristic response sequence.

[0039] In order to more intuitively reflect the sampling data, the actual sampling data is as shown in Table 2 below: Table 2: Voltage characteristic response comparison data table Sampling time (ms) Current voltage (V) Reference voltage (V) Amplitude difference (V) 200 220.5 220.4 0.1 201 220.7 220.6 0.1 202 220.6 220.5 0.1 203 220.9 220.8 0.1 204 221.0 220.9 0.1 As shown in Table 2, this voltage characteristic response sequence is used for subsequent characteristic difference measurement analysis.

[0040] S202: Based on the window data of the voltage characteristic response sequence, call the medium-length segment data in the benchmark power characteristic sequence, process the synchronization point values ​​of the three dimensions in sequence, calculate the degree of difference between the three groups of values ​​through the Euclidean distance formula, and generate a three-dimensional voltage difference value.

[0041] Based on the window data of the constructed voltage characteristic response sequence, the length of each sliding window must be defined first. , for example, sampling points, select 5 groups of response differences in each window as the samples for this round of comparison. For example, , extract the corresponding amplitude difference sequence, phase difference sequence and slope difference sequence from the reference power characteristic sequence, each dimension corresponds to the same time index segment, and on this basis, extract the synchronization point values ​​of the three dimensions at each group of time points to form a three-dimensional vector group, for example, ms, let the amplitude difference sequence be , the phase difference sequence is (unit is °), the slope difference sequence is , which correspond to the instantaneous incremental difference between the current data and the benchmark data. After constructing the three-dimensional difference vector, the Euclidean distance between the three sets of values ​​is calculated using the following formula: ; in, represents the amplitude difference, Indicates the phase difference value, Represents the slope difference. The above operation is performed once for each time point to obtain the difference value of the point in three-dimensional space. For example, ms, set the three-dimensional value of the current point to , the reference point is ,but: The differences are: , , ; The Euclidean distance is: ; Similarly, the Euclidean distance is calculated for each of the five points in the entire window interval to obtain a set of three-dimensional voltage difference values, such as , which can be used as input parameters for subsequent judgments, 、 、 Obtained through synchronous comparison, such as x i =Current amplitude response-reference amplitude response, y i = Current phase difference - reference phase difference, z i =Current slope difference-reference slope difference. The innovation of the above formula is that by synchronously quantifying the three key voltage indicators of amplitude, phase, and slope and combining them into a unified spatial measurement system, the accuracy of distinguishing abnormal voltage responses is enhanced.

[0042] S203: Based on the three distance values ​​in the three-dimensional voltage difference value, compare the preset voltage difference judgment threshold to determine whether they exceed the corresponding threshold at the same time. When all three judgment conditions are met, mark the window voltage state and generate a power state calibration instruction.

[0043] The three distance values ​​are the Euclidean distance differences between the voltage characteristic response sequence and the reference sequence in the three dimensions of amplitude, phase and slope.

[0044] The voltage difference judgment threshold is based on 500 sets of statistical analysis of typical load sample collection data, determining the 95% confidence interval of the Euclidean distance distribution of abnormal samples, and statistically obtaining the Euclidean distance range of the abnormal voltage response sequence.

[0045] After obtaining the above three-dimensional voltage difference values, they are judged item by item, and the amplitude distance, phase distance and slope distance are compared to see whether they exceed the corresponding preset threshold range. The threshold settings of the three dimensions are all based on 500 groups of typical load samples for Euclidean distance statistical analysis, and the maximum distance of the dimension under the 95% confidence interval is obtained as the upper limit of the judgment. For example, after analyzing 500 groups of samples, it is found that the upper limit of the Euclidean distance of the amplitude difference within the 95% confidence interval is 0.15V, and the upper limit of the phase difference is , the upper limit of the slope difference is 0.04V / ms, then in the actual judgment, the current point difference value is compared with the corresponding threshold one by one, for example, for the above The calculation results of the ms point are amplitude difference 0.05V, phase difference 0.1°, and slope difference 0.01V / ms. All three items do not exceed the threshold, so no state mark is made. However, if three consecutive points among the five sampling points in a certain window exceed the threshold in three dimensions, the voltage state mark operation is performed on the window to generate a power state calibration instruction. For example, if the Euclidean distance result in another interval is It is obvious that the first three items are respectively higher than the amplitude threshold of 0.15. If the phase difference and slope difference of the corresponding points also exceed the limit, the simultaneous exceeding condition is met, and the status marking operation is triggered. The marking data and instructions can be recorded as "abnormal status-high fluctuation section (t=300~304ms)" to facilitate the subsequent execution of the status adjustment logic.

[0046] In this embodiment, the specific steps of S3 include: S301: Based on the power state calibration instruction, collect the current and voltage waveform sequences in continuous cycles, construct the sampling and prediction error covariance matrix, calculate the weights according to the observation and prediction errors, use the Kalman filter algorithm to correct the frequency, and generate a dynamic frequency adjustment coefficient.

[0047] The dynamic frequency adjustment coefficient dynamically corrects the frequency based on the Kalman filter to generate an adjustment coefficient for real-time compensation of the system frequency offset.

[0048] Based on the power state calibration instruction, the operation of collecting the current and voltage waveform sequence in continuous cycles is refined. First, the voltage and current values ​​of each sampling point in 5 consecutive grid cycles are collected from the power system. The synchronous sampling frequency of this process needs to be set to 10kHz to ensure that 200 sampling points are collected in each cycle, that is, a total of 1000 points are collected to form the waveform sequence; then, the points need to be segmented and a group of 200 points of data are extracted separately for each cycle. By comparing the changes in the amplitude and phase difference of each point in adjacent cycles, the observation error sequence and the prediction error sequence are preliminarily constructed; in actual operation, it is assumed that the voltage collected in a certain cycle is an array , current is an array , then the observation error is obtained by calculating the difference between the corresponding sampling points of this period and the previous period and prediction error , if the voltage difference at a certain point is , the error at this point is ; Calculate the mean and variance at the sampling points and construct the error covariance matrix ; Based on the variance analysis of the elements in this matrix, if the error value variance , then set the corresponding weight Otherwise, set it to , from which we get the weight vector ; Then, in the frequency correction, the predicted value of the current cycle frequency is used Fitting frequency with actual waveform Compare and calculate the prediction error , and bring it into the Kalman filter correction formula: ; in, , we can get , and finally the dynamic frequency adjustment coefficient is , used to correct the current frequency offset of the system.

[0049] S302: Calculate the phase difference change rate and perform second-order difference based on the dynamic frequency adjustment coefficient, extract the phase offset trend and construct a ripple correlation data group, compare the trend node with the ripple response analysis, and generate a phase change fitting coefficient group.

[0050] According to the above dynamic frequency adjustment coefficient, the phase difference of the voltage signal at each corresponding sampling point in the current cycle and the previous cycle is first extracted, and the phase difference change rate is calculated. , assuming that the phase of a sampling point in the current cycle is , the corresponding sampling point of the previous cycle is , time interval , then the rate of change is ; Then, based on the change rate sequence, the adjacent difference values ​​are subjected to the second-order difference operation. For example, if the difference between the current, previous, and previous two frames is 100, 95, and 90 respectively, the second-order difference is , which is used to judge the stability of the phase change trend; in the trend extraction process, for any continuous sampling period, a ripple correlation data group is constructed to record the amplitude disturbance amplitude and phase disturbance amplitude within the period. For example, if the voltage ripple amplitude is , the phase perturbation is , store the data into a vector sequence And compare this set of data with the fitted trend sequence one by one; set the offset judgment threshold for the mutation point of the trend node, if the trend change value exceeds When , it is defined as a significant deviation and determined as a new trend node; the phase change of the area where the node is located is fitted into a linear function , and extract the fitting parameter set ,in is the trend slope, is the offset starting point, and finally, The combination is used as the phase change fitting coefficient group for this stage.

[0051] S303: Calling the phase change fitting coefficient group, constructing the cycle-to-cycle difference weight function and updating the fitting sequence, performing iterative calculation to obtain the smoothing coefficient sequence and matching the detection parameter combination to obtain the self-matching detection parameter group.

[0052] The self-matching detection parameter group is a set of optimal characteristic parameters dynamically extracted through frequency correction and phase fitting iteration based on the current power waveform state. It is used to accurately reflect the system frequency and phase offset characteristics and support subsequent detection decisions.

[0053] The difference weight function is an important function used to measure the difference in phase change trends between cycles. It is weighted according to the fitted phase change coefficient group and the phase difference characteristics between cycles to reflect the trend consistency and the severity of the offset.

[0054] After calling the above phase change fitting coefficient group, first construct the cycle-to-cycle difference weight function, and extract the phase change coefficient difference values ​​between multiple cycles. , and compare the difference value with the reference value after processing the absolute value. Assuming that the reference difference is set to ,like , then the corresponding period is assigned a weight of 0.9, otherwise it is assigned 0.3; the weights constitute the weight function array , which is used to measure the consistency of phase trend between cycles; then the weight function is applied to the previous fitting sequence, its weight is iteratively updated, a new fitting sequence is constructed, and the iterative operation is performed at least 3 times to stabilize; the phase slope change in each iteration is used As a basis for measurement, the trend segment with the best convergence is extracted and its corresponding smoothing coefficient sequence is defined, such as the smoothing coefficient , calculate the sequence Match the disturbance pattern in the smooth sequence with the original ripple response, compare the set ripple pattern group in the sample library, select the parameter combination that best matches the smoothing coefficient and phase change trend, and construct a self-matching detection parameter group; this parameter group consists of frequency adjustment value, phase slope, and ripple disturbance value, such as , specific parameters are shown in Table 3.

[0055] Table 3: Self-matching detection parameter table Cycle Number Frequency adjustment coefficient Phase change slope (° / s) Voltage ripple amplitude (V) P1 1.000 0.22 1.2 P2 0.998 0.20 1.1 P3 1.001 0.21 1.3 As shown in Table 3, the self-matching detection parameter groups extracted under different cycles are listed. Each group of parameters is used to further describe the system state characteristics within the corresponding cycle and is used for subsequent matching and judgment analysis. The obtained parameters are all derived from actual calculations and meet the normal range of power system operation.

[0056] In this embodiment, the specific steps of S4 include: S401: According to the self-matching detection parameter group, obtain the main frequency component of the ripple spectrum in the reference power supply characteristic sequence, perform fast Fourier transform to extract the energy concentration frequency point, fit the harmonic frequency distribution vector, calculate the frequency distribution deviation characteristic value, and obtain the frequency distribution value group.

[0057] Frequency distribution deviation eigenvalue: measures the degree of deviation of the harmonic frequency distribution from the ideal equally spaced distribution and is used to identify spectral structure anomalies.

[0058] Based on the self-matching detection parameter set, the primary frequency component of the ripple spectrum in the benchmark power supply characteristic sequence is first extracted. This extraction process requires obtaining power supply current / voltage waveform data from actual power equipment operation records and converting it into frequency domain signals. The benchmark sequence length is set to 10 seconds, and the sampling frequency is set to 25kHz, resulting in 250,000 sampling points. Fast Fourier transform (FFT) is used to process the time domain signal, generating spectrum data from 0Hz to half the sampling frequency (12.5kHz). The primary frequency component of the ripple spectrum can be extracted by identifying frequency points in the spectrum curve where the power density is greater than twice the average power density. The average power density is set to 0. 8W / Hz, then the main frequency threshold is 1.6W / Hz, select the frequency point with power greater than this value, record it as the energy concentration frequency point group, and then count the frequency point distribution every 50Hz as a frequency window to form a harmonic frequency distribution vector. At each frequency position, summarize the corresponding frequency point power value and calculate its mean and variance. Suppose there are 5 frequency points at a certain frequency position (such as 300Hz), and the spectrum power values ​​are 1.5W, 1.8W, 2.1W, 2.4W, and 2.6W respectively, then the mean is 2.08W and the variance is 0.14W². For each frequency point, further call its ripple frequency amplitude average value. For example, the corresponding amplitude measured at this position is 0.34V, which is used as , introduce the equivalent impedance Z (for example, set it to 50Ω) and substitute it into the formula to calculate the deviation characteristic value.

[0059] The calculation formula used for the frequency distribution deviation characteristic value is: ; in, Representative The frequency distribution deviation characteristic value of the frequency point is a dimensionless parameter. Represents the number of energy-concentrated frequency points corresponding to the frequency point, which is a dimensionless parameter. Representative The frequency point is The spectral weight factor at the frequency position is a dimensionless parameter. Representative The frequency point is The spectrum power value of the frequency position, in W, Representative The average value of the spectrum power value of the frequency position frequency point, the unit is W, Representative The variance of the spectrum power value of the frequency position frequency point, the unit is W², Representative The average value of the ripple frequency amplitude corresponding to the frequency position frequency point, the unit is V, Represents equivalent impedance, in Ω.

[0060] Taking the frequency position k = 1 (i.e. 300 Hz) as an example, assume that the number of frequency points M = 5, the five collected spectrum power values ​​are shown above, and the spectrum weight factors are 0.18, 0.20, 0.22, 0.19, and 0.21, respectively. The calculation is as follows: ; ; ; ; Denominator: ; Bring in the multi-point calculation numerator: First point: ; Second point: ; Third point: ; Fourth point: ; Fifth point: ; ; The results show that at the frequency position of 300Hz, the frequency distribution deviation characteristic value is 0.00203, which can be used as a basis for constructing subsequent distribution value groups.

[0061] S402: Based on the frequency distribution value group, detect the amplitude component of the harmonic frequency position, calculate the amplitude ratio of the harmonic frequency component to the fundamental wave amplitude, filter the frequency bands above the ratio threshold, and obtain the harmonic amplitude exceeding the threshold frequency band value.

[0062] The harmonic amplitude exceeding threshold frequency band value refers to the harmonic frequency band whose amplitude exceeds the fundamental wave ratio threshold, which is used to locate high-energy harmonics that need to be suppressed.

[0063] The ratio threshold is set to 7% based on a harmonic distribution analysis of 500 sets of typical fire-fighting equipment power supply environment sampling data. This is to balance the detection sensitivity and false alarm rate in a complex power environment and ensure the stable operation of the fire-fighting equipment power supply.

[0064] Based on the above frequency distribution value group, the amplitude component of the harmonic frequency position is first detected. During the extraction process, each frequency window (such as a frequency band divided by 50Hz) needs to be traversed, and the maximum amplitude of the frequency point in the window is counted as the representative amplitude of the frequency band. Assuming that the current frequency band is the 6th band (frequency range 250Hz~300Hz), the spectrum power values ​​of the collected frequency points in this band are 1.9W, 2.3W, 2.5W, 1.8W, and 2.0W respectively. The corresponding voltage amplitude can be calculated by the power calculation formula Deducing the voltage amplitude, assuming Z = 50Ω, the multi-point voltage amplitude is 、 、 、 、 , the maximum amplitude of this frequency band is 11.18V. Then we need to obtain the fundamental amplitude as the reference value. Assuming the fundamental frequency is 50Hz, the measured amplitude is 220V. The ratio of the maximum amplitude of the frequency band to the fundamental amplitude is calculated, that is, the amplitude ratio is , and then compared with the set ratio threshold, the ratio threshold is 7%. According to the previous statistics of 500 groups of fire equipment power supply data, the distribution of the proportion of harmonic amplitude exceeding the fundamental amplitude in the frequency band is recorded during the statistical process. According to its distribution law, the amplitude ratio corresponding to the upper 7% frequency points is extracted as the harmonic boundary value to ensure that the selected frequency band is representative and has actual interference risk, forming the data summary shown in Table 4: Table 4: Amplitude ratio statistics Serial number Band center frequency (Hz) Maximum amplitude (V) Fundamental wave amplitude (V) ratio(%) 1 150 13.45 220 6.11 2 200 14.20 220 6.45 3 250 16.70 220 7.59 4 300 11.18 220 5.08 5 350 18.20 220 8.27 As shown in Table 4, the frequency bands with amplitude ratios exceeding 7% include 250 Hz and 350 Hz. The frequency bands are screened, and their frequency ranges and corresponding amplitude information are retained to form an array of harmonic amplitude over-threshold frequency band values. The array format is {250 Hz: 7.59%, 350 Hz: 8.27%}. The record of each frequency band value will be used as an input for subsequent response rule matching and harmonic suppression strategy formulation. During the entire screening process, the calculation result of the frequency distribution vector and the fundamental wave amplitude ratio is called, and then a judgment is made with the set threshold of 7%. The judgment method is "If the ratio value is > 0.07, then record the current frequency band." An over-threshold mapping index table is established for the recorded frequency bands, and finally the harmonic amplitude over-threshold frequency band value is obtained.

[0065] S403: calling the frequency band value where the harmonic amplitude exceeds the threshold, matching the response rule template in the self-matching detection parameter group, selecting the suppression strategy unit according to the mapping relationship between the frequency band and the suppression strategy, and generating a harmonic suppression strategy vector.

[0066] Call the above-mentioned harmonic amplitude over-threshold frequency band value array, first extract the frequency range of each frequency band, such as the previous result is {250Hz, 350Hz}, and then match the mapping relationship between the frequency band and the suppression strategy recorded in the template according to the response rule template set in the self-matching detection parameter group. For example, for the frequency band 250Hz, the corresponding response rule template sets the frequency range to 240Hz~260Hz, and the matching strategy unit is "injection reverse current amplitude 0.4V", while for the frequency band 350Hz, the set response strategy is "filter bandwidth setting 50Hz, center frequency 350Hz". After the matching is completed, the suppression strategy parameters corresponding to each frequency band are extracted and encoded to form a harmonic suppression strategy vector. The vector format is as follows: ; During the generation process, it is necessary to determine whether the frequency band hits the template setting range in turn. The judgment logic is: Set the center frequency of the frequency band to , if there is a frequency interval in the template ,satisfy , then extract its response strategy parameter group; encode the extracted parameter items, directly enter the frequency item in Hz, convert the response voltage or bandwidth into vector elements in V or Hz notation, and record the response mechanism with the strategy action identifier (such as reverse injection, filtering suppression). Finally, combine them into a complete harmonic suppression strategy vector as the output of this detection step.

[0067] In this embodiment, the specific steps of S5 include: S501: Based on the harmonic suppression strategy vector, the voltage fluctuation range and the current phase cumulative deviation are extracted, the voltage-current correlation equation is established using the least squares method, the residual sum of squares and the goodness of fit are calculated, and the power supply evaluation model parameters are generated.

[0068] In the process of extracting the voltage fluctuation range and the current phase cumulative deviation based on the harmonic suppression strategy vector, the voltage response value and strategy category corresponding to each frequency band are first extracted from the previously obtained harmonic suppression strategy vector, the voltage response value is arranged according to the frequency time window, the maximum and minimum voltage values ​​in each time window are recorded, and the difference between the two is calculated as the voltage fluctuation value of the window, and then the overall voltage fluctuation range is obtained by superposition window by window; for example, the reverse current injection strategy is applied at the frequency band of 250Hz, and the voltage response value in 10 time windows is recorded. The response values ​​are 0.36V, 0.40V, 0.38V, 0.42V, 0.39V, 0.41V, 0.37V, 0.43V, 0.38V, and 0.40V, respectively. The voltage fluctuation range is 0.07V, which is the maximum value 0.43V minus the minimum value 0.36V. At the same time, the response current value recorded in the frequency band is extracted. By sampling the current phase data, the phase difference between consecutive sampling points is calculated, and then the phase difference is accumulated at each time point to obtain the current phase cumulative deviation. Assume that the current phase of adjacent sampling points is 10°. , 12°, 15°, 17°, 20°, the phase differences are 2°, 3°, 2°, 3° respectively, and the cumulative deviation is 10°; the above two parameters are used as voltage and current response characteristic inputs. According to the least squares fitting principle, when constructing the voltage-current correlation equation, the voltage fluctuation range is used as the dependent variable and the current phase cumulative deviation is used as the independent variable. The calculation is performed as follows: Assuming that in the 5 groups of samples, the voltage fluctuation range is 0.05V, 0.06V, 0.08V, 0.07V, 0.09V respectively, and the current phase The cumulative deviations are 6°, 8°, 9°, 10°, and 12° respectively. By constructing a straight line fit with the minimum residual sum of squares, the slope and intercept of the regression model are obtained. The residual sum of squares is further calculated as the sum of the squares of the differences between the predicted values ​​and the true values ​​of each group of points. For example, the predicted value of sample 1 is 0.054V, the residual is 0.004V, and the square is 0.000016V². This is repeated for the remaining samples in turn, and the total residual sum of squares is 0.00023V². The goodness of fit is then calculated based on the closeness between the fitted value and the true value. , using the formula: ; in is the residual sum of squares, is the sum of the squares of the true value deviation from the mean. Here, the sample mean of 0.07V is used as a reference, and the final result is , indicating the degree of association; the slope, intercept and The values ​​are recorded together as power supply evaluation model parameters.

[0069] S502: Call the power supply evaluation model parameters, input the total harmonic distortion rate to expand the model, reconstruct the weight matrix through the least square method, verify the sensitivity coefficient of the harmonic component to the model output, and obtain the optimized power supply evaluation model.

[0070] Call the power supply evaluation model parameters and input the total harmonic distortion rate to expand the model. First, set the input total harmonic distortion rate THD to 6.8%, quantize it into the amplitude and power spectrum density of multiple frequency bands, and construct the harmonic frequency band amplitude vectors respectively. With the power spectral density vector For example, the frequency bands 250Hz, 300Hz, and 350Hz have amplitudes of 16.2V, 14.5V, and 18.0V, and the corresponding power spectrum densities are 0.75W / Hz, 0.68W / Hz, and 0.83W / Hz. Combining the average value and the standard deviation, let the average amplitude be V, standard deviation V, the average spectral density is W / Hz, standard deviation W / Hz, call the original weight coefficient matrix in the model , for example, , stable offset parameter , calculate the sensitivity coefficient using the formula: ; in, Representative The model output sensitivity coefficient corresponding to the quasi-harmonic component is a dimensionless parameter. Representative Class model The weight coefficient of each harmonic frequency band is a dimensionless parameter. Representative The amplitude of each harmonic frequency band, in V, Represents the average value of the harmonic frequency band amplitude, the unit is V, Represents the standard deviation of voltage amplitude, in V. Representative Stable offset parameter for amplitude fluctuation regulation in class models, Representative The power spectral density of each frequency band, in W / Hz, Represents the average value of the power spectral density of the frequency band, in W / Hz, Represents the standard deviation of power spectral density, unit is W / Hz, Represents the total number of harmonic frequency bands involved in the model and is a dimensionless parameter.

[0071] Substitute the parameters for calculation: ; First, the amplitude part is normalized and weighted: ; ; Similarly, the remaining two items are calculated, and the final weighted sum of the amplitude is approximately 0.0113. Then calculate the power spectral density term: ; ; The other two results are 0.0134 and 0.0216 respectively.

[0072] ; = ; The above sensitivity coefficients are recorded as parameters in the expanded optimized power supply assessment model to indicate the degree of influence of the frequency band on the model results.

[0073] S503: Match the output value of the optimized power supply evaluation model with the preset state threshold interval, perform numerical interval membership calculation and discrete coding mapping, and obtain a power supply state classification identifier.

[0074] Match the output value of the optimized power supply evaluation model with the preset state threshold interval. First, it is necessary to set the power supply performance interval for state division. Assume that the power supply state is divided into three categories based on the original data statistics: normal, attention-demanding, and abnormal. The corresponding membership intervals are [0, 0.03], (0.03, 0.07], and (0.07, 1], respectively. The sensitivity coefficient value obtained in the previous section is Substitute this into the judgment process to determine whether it falls into the "normal" state interval. The comparison method is to directly compare the value with the upper and lower limits of the interval. If it is confirmed that it is greater than or equal to 0 and less than or equal to 0.03, it is classified as a normal state. Then, according to the discrete mapping coding rule corresponding to the state interval, the "normal" state is mapped to 0, the state requiring attention is mapped to 1, and the abnormal state is mapped to 2. Therefore, the output of this example is mapped to 0; after completion, the above judgment process is repeated for each frequency band to construct a complete frequency band state code vector. If 250Hz, 300Hz, and 350Hz correspond to state codes 0, 1, and 0 respectively, then the state code vector is [0, 1, 0]. This vector will be used as the output result of the power state classification and recognition module.

[0075] Table 5: Power status determination interval table Status Category Membership interval Encoded value normal [0，0.03] 0 Need attention (0.03，0.07] 1 abnormal (0.07，1] 2 As shown in Table 5, the power status intervals correspond to the corresponding coding values. In the actual implementation process, the sensitivity coefficient of the frequency band is first calculated, and then the interval to which it belongs is determined based on the value and converted into a classification identification code.

[0076] like Figure 2As shown, the present invention also provides a fire-fighting equipment power supply detection system, which is used to implement the above-mentioned fire-fighting equipment power supply detection method, and the system includes: The signal acquisition module is used to obtain the voltage fluctuation sequence, current phase offset and ripple coefficient through a multi-channel sensor group, and use a sliding window mechanism to perform noise reduction on the original signal to generate a reference power supply feature sequence, which is then passed to the feature calibration module; The feature calibration module is used to call the benchmark power feature sequence to perform Z-score normalization on the voltage fluctuation sequence, calculate the Euclidean distance between the current window voltage value, current phase offset, ripple coefficient and the feature sequence, and generate a power state calibration instruction when the three distance values ​​simultaneously exceed the set threshold and pass it to the parameter adaptation module; The parameter adaptation module is used to call the power state calibration instruction, dynamically adjust the sampling frequency using the Kalman filter algorithm, build a time-varying power parameter model based on the current phase offset change rate and the second-order derivative of the ripple coefficient, and iteratively calculate and output a self-matching detection parameter group, which is passed to the harmonic analysis module; The harmonic analysis module is used to call the self-matching detection parameter group, perform fast Fourier transform on the main frequency component of the ripple spectrum in the reference power supply characteristic sequence, extract the harmonic amplitude and compare it with the fundamental amplitude. When the harmonic amplitude exceeds the set ratio of the fundamental amplitude, a harmonic suppression strategy vector is generated and passed to the state assessment module; The state assessment module is used to call the harmonic suppression strategy vector, reconstruct the power supply assessment model using the least squares method, input the voltage fluctuation range, current phase cumulative deviation, and total harmonic distortion rate parameters, calculate them through the model, and output the power supply state classification label.

[0077] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the scope of protection of the present invention; therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for detecting power supply of fire-fighting equipment, characterized in that: The following steps are involved: S1: The voltage fluctuation sequence, current phase offset, and ripple coefficient of the power supply are collected through a multi-channel sensor group, and a sliding window mechanism is used for noise reduction to generate a reference power supply characteristic sequence; S2: Call the benchmark power feature sequence, perform Z-score normalization on the voltage fluctuation sequence, calculate the Euclidean distance with the feature sequence, and generate a power state calibration instruction when the three distance values ​​simultaneously exceed the voltage difference judgment threshold; S3: Based on the power state calibration instruction, the Kalman filter algorithm is used to dynamically adjust the sampling frequency. According to the current phase offset change rate and the second-order derivative of the ripple coefficient, a power parameter time-varying model is constructed. The power state evolution is dynamically fed back and a self-matching detection parameter group is output. S4: performing a fast Fourier transform on the main frequency component of the ripple spectrum in the reference power supply characteristic sequence according to the self-matching detection parameter group to extract the harmonic amplitude. When the harmonic amplitude exceeds a set ratio of the fundamental amplitude, a harmonic suppression strategy vector is generated. S5: Based on the harmonic suppression strategy vector, the least squares method is used to reconstruct the power supply evaluation model, input voltage fluctuation range, current phase cumulative deviation, total harmonic distortion rate, and output power status classification identification.

2. The fire-fighting equipment power supply detection method according to claim 1, characterized in that: The benchmark power supply characteristic sequence includes a Z-score standardized voltage sequence, a phase offset mean sequence, and a ripple coefficient smoothing sequence; the power supply state calibration instructions include a voltage fluctuation threshold, a phase offset tolerance, and a ripple tolerance; the self-matching detection parameter group includes a dynamic sampling frequency parameter, a phase change rate threshold, and a ripple second-order derivative parameter; and the harmonic suppression strategy vector includes a harmonic attenuation coefficient group, a fundamental amplitude protection threshold, and a spectrum weight factor. The power status classification identification includes voltage extreme difference level, phase cumulative deviation range and total harmonic distortion rate level.

3. The fire-fighting equipment power supply detection method according to claim 1, characterized in that: The specific steps of S1 include: S101: Using a multi-channel sensor group, collect the real-time voltage value of the power supply, the phase offset of the current signal, and the ripple data of the corresponding frequency band in the voltage signal. Perform synchronization timestamp correction on the channel data under a unified time base, construct a channel-by-channel data structure, perform continuity check and missing value marking, and obtain an integrated synchronous data sequence. S102: Constructing a sliding window of equal width based on the integrated synchronous data sequence, setting the step size and window length, performing interpolation correction, deviation elimination, and smoothing processing, eliminating interference segments with sudden changes and short-term high-frequency spike data, and obtaining characteristic sequence data after noise reduction processing; S103: Call the characteristic sequence data after noise reduction processing, extract the phase difference, fluctuation interval change ratio and ripple amplitude ratio between multi-channel signals in the same time window, perform linear normalization processing and then perform vector recombination to generate a reference power supply characteristic sequence.

4. The fire-fighting equipment power supply detection method according to claim 3, characterized in that: The specific steps of S2 include: S201: Calling a reference power supply characteristic sequence, synchronously taking values ​​of the voltage values ​​at the same position in the current voltage fluctuation sequence based on the reference voltage data in the time axis, constructing a corresponding relationship sequence, and sequentially comparing the amplitudes of the two sequences to establish a voltage characteristic response sequence; S202: Based on the window data of the voltage characteristic response sequence, call the medium-length segment data in the reference power characteristic sequence, process the synchronization point values ​​of the three dimensions in sequence, calculate the degree of difference between the three groups of values ​​using the Euclidean distance formula, and generate a three-dimensional voltage difference value; S203: Based on the three distance values ​​in the three-dimensional voltage difference value, compare the preset voltage difference judgment threshold to determine whether they exceed the corresponding threshold at the same time. When all three judgment conditions are met, mark the window voltage state and generate a power state calibration instruction.

5. The fire-fighting equipment power supply detection method according to claim 4, characterized in that: The specific steps of S3 include: S301: Based on the power state calibration instruction, the current and voltage waveform sequences within the continuous cycle are collected, the sampling and prediction error covariance matrix is ​​constructed, the weights are calculated according to the observation and prediction errors, the frequency is corrected using the Kalman filter algorithm, and a dynamic frequency adjustment coefficient is generated; S302: Calculate the phase difference change rate based on the dynamic frequency adjustment coefficient and perform second-order difference, extract the phase offset trend and construct a ripple correlation data set, compare the trend node with the ripple response analysis, and generate a phase change fitting coefficient set; S303: Calling the phase change fitting coefficient group, constructing the cycle-to-cycle difference weight function and updating the fitting sequence, performing iterative calculation to obtain the smoothing coefficient sequence and matching the detection parameter combination to obtain the self-matching detection parameter group.

6. The fire-fighting equipment power supply detection method according to claim 5, characterized in that: The specific steps of S4 include: S401: Obtain the main frequency component of the ripple spectrum in the reference power supply characteristic sequence based on the self-matching detection parameter group, perform fast Fourier transform to extract the energy concentration frequency point, fit the harmonic frequency distribution vector, calculate the frequency distribution deviation characteristic value, and obtain the frequency distribution value group; S402: Based on the frequency distribution value group, detect the amplitude component of the harmonic frequency position, calculate the amplitude ratio of the harmonic frequency component to the fundamental wave amplitude, filter the frequency bands above the ratio threshold, and obtain the harmonic amplitude exceeding the threshold frequency band value; S403: calling the frequency band value where the harmonic amplitude exceeds the threshold, matching the response rule template in the self-matching detection parameter group, selecting the suppression strategy unit according to the mapping relationship between the frequency band and the suppression strategy, and generating a harmonic suppression strategy vector.

7. The fire-fighting equipment power supply detection method according to claim 6, characterized in that: The formula used to calculate the characteristic value of the frequency distribution deviation is: ； in, Representative The frequency distribution deviation characteristic value of the frequency point is a dimensionless parameter. Represents the number of energy-concentrated frequency points corresponding to the frequency point, which is a dimensionless parameter. Representative The frequency point is The spectral weight factor at the frequency position is a dimensionless parameter. Representative The frequency point is The spectrum power value of the frequency position, in W, Representative The average value of the spectrum power value of the frequency position frequency point, the unit is W, Representative The variance of the spectrum power value of the frequency position frequency point, the unit is W², Representative The average value of the ripple frequency amplitude corresponding to the frequency position frequency point, the unit is V, Represents equivalent impedance, in Ω.

8. The fire-fighting equipment power supply detection method according to claim 6, characterized in that: The specific steps of S5 include: S501: Based on the harmonic suppression strategy vector, the voltage fluctuation range and the current phase cumulative deviation are extracted. The voltage-current correlation equation is established using the least squares method. The residual sum of squares and goodness of fit are calculated to generate the power supply evaluation model parameters. S502: Call the power supply evaluation model parameters, input the total harmonic distortion rate to expand the model, reconstruct the weight matrix through the least squares method, verify the sensitivity coefficient of the harmonic component to the model output, and obtain the optimized power supply evaluation model; S503: Match the output value of the optimized power supply evaluation model with the preset state threshold interval, perform numerical interval membership calculation and discrete coding mapping, and obtain a power supply state classification identifier.

9. The fire-fighting equipment power supply detection method according to claim 8, characterized in that: The specific calculation formula of the sensitivity coefficient is: ； in, Representative The model output sensitivity coefficient corresponding to the quasi-harmonic component is a dimensionless parameter. Representative Class model The weight coefficient of each harmonic frequency band is a dimensionless parameter. Representative The amplitude of each harmonic frequency band, in V, Represents the average value of the harmonic frequency band amplitude, the unit is V, Represents the standard deviation of voltage amplitude, in V. Representative Stable offset parameter for amplitude fluctuation regulation in class models, Representative The power spectral density of each frequency band, in W / Hz, Represents the average value of the power spectral density of the frequency band, in W / Hz, Represents the standard deviation of power spectral density, unit is W / Hz, Represents the total number of harmonic frequency bands involved in the model and is a dimensionless parameter.

10. A fire-fighting equipment power supply detection system, characterized by: The system is used to implement the fire-fighting equipment power supply detection method according to any one of claims 1 to 9, and the system includes: The signal acquisition module is used to obtain the voltage fluctuation sequence, current phase offset and ripple coefficient through a multi-channel sensor group, and use a sliding window mechanism to perform noise reduction on the original signal to generate a reference power supply feature sequence, which is then passed to the feature calibration module; The feature calibration module is used to call the benchmark power feature sequence to perform Z-score normalization on the voltage fluctuation sequence, calculate the Euclidean distance between the current window voltage value, current phase offset, ripple coefficient and the feature sequence, and generate a power state calibration instruction when the three distance values ​​simultaneously exceed the set threshold and pass it to the parameter adaptation module; The parameter adaptation module is used to call the power state calibration instruction, dynamically adjust the sampling frequency using the Kalman filter algorithm, build a time-varying power parameter model based on the current phase offset change rate and the second-order derivative of the ripple coefficient, and iteratively calculate and output a self-matching detection parameter group, which is passed to the harmonic analysis module; The harmonic analysis module is used to call the self-matching detection parameter group, perform fast Fourier transform on the main frequency component of the ripple spectrum in the reference power supply characteristic sequence, extract the harmonic amplitude and compare it with the fundamental amplitude. When the harmonic amplitude exceeds the set ratio of the fundamental amplitude, a harmonic suppression strategy vector is generated and passed to the state assessment module; The state assessment module is used to call the harmonic suppression strategy vector, reconstruct the power supply assessment model using the least squares method, input the voltage fluctuation range, current phase cumulative deviation, and total harmonic distortion rate parameters, calculate them through the model, and output the power supply state classification label.

Images