A power quality detection method of a smart meter

Abstract

The present application relates to the technical field of electric energy measurement, in particular to a power quality detection method of smart meter, comprising the following steps: calculating channel energy performance and calibrating abnormal channels by smart meter, analyzing spectral shape drift to generate harmonic combination results, detecting abnormal events and classifying, evaluating power quality level, screening consistent trend period segments and predicting quality change, and generating trend modeling data.In the present application, the period energy performance is calculated and the sampling abnormal channel is marked, the spectral shape drift is identified in combination with the harmonic amplitude inter-frequency band combination characteristics, the period overlap behavior is further analyzed in the abnormal event detection, and the abnormal degree of the period segment is quantified, the multi-parameter classification results are constructed by event type aggregation, the repeated abnormal rules in the time sequence are identified and the dense score is constructed accordingly, the input period segments with consistent trend are extracted in combination with the voltage fluctuation, frequency change and harmonic amplitude change trend, and the power quality trend prediction is realized.

Description

Technical Field

[0001] This invention relates to the field of electrical energy measurement technology, and in particular to a method for detecting the power quality of a smart meter. Background Technology

[0002] The field of power measurement technology encompasses various measurement technologies and methods for sensing, recording, analyzing, and managing power parameters in power systems. This includes real-time detection and acquisition of power quality parameters such as voltage, current, active power, reactive power, electrical energy, harmonics, frequency, and voltage fluctuations. Power parameters are sensed through embedded measurement circuits, sensors, power acquisition chips, and multi-channel acquisition circuits. Data processing devices, such as microcontrollers or application-specific integrated circuits, perform real-time processing and analysis of the acquired data for power supply status assessment, equipment operation monitoring, and power usage statistics and optimization. This covers multiple aspects, including the structural design of power measurement devices, data acquisition accuracy improvement, signal conditioning circuits, electromagnetic compatibility processing, and communication interface protocol implementation. Applications include smart grids, industrial power systems, and commercial and residential power monitoring scenarios. Specifically, the power quality detection method for smart meters refers to a method for collecting various power quality parameters from smart meters. The specific implementation of the processing and judgment method targets power quality issues such as voltage deviation, frequency deviation, voltage fluctuation, flicker, and harmonic distortion. Specifically, it uses a power quality monitoring circuit that communicates with the main chip to collect signals through a front-end circuit with voltage and current sampling channels. The embedded Fourier transform circuit calculates the harmonic content and frequency fluctuation, and the voltage fluctuation judgment circuit monitors the voltage change trend in real time. The collected data is integrated and processed through a specific time window and then transmitted to the microprocessor for preliminary classification and judgment of power quality events. The quality level is evaluated and recorded based on the set technical parameter thresholds. The method covers a synchronous voltage and current acquisition mechanism based on an integrated signal acquisition chip, a frequency domain analysis process with a transformation calculation circuit as the core, and a judgment rule execution path based on fixed thresholds.

[0003] Traditional power quality detection technologies use fixed thresholds to process frequency, voltage, and harmonic information within a specific time window in a single operation. This makes it difficult to track the evolution trend of power quality events over a continuous period. When multiple anomalies occur alternately with similar frequency amplitudes, it is impossible to effectively determine the overlapping characteristics between events. Furthermore, it does not introduce a cross-period analysis structure to identify repetitive anomalies. In actual operation, multiple intermittent interferences may occur but are only identified as isolated anomalies, leading to assessment bias. In scenarios with frequent voltage fluctuations or dynamic changes in the spectrum, it is impossible to establish a model for predicting the intensity and trend of event clustering, affecting the stability of long-term quality assessment. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a power quality detection method for smart meters.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a power quality detection method for smart meters, comprising the following steps:

[0006] S1: Using smart meters, calculate the instantaneous power of each channel and obtain the periodic energy performance by accumulating it. Compare the differences between channels, mark channels with abnormal sampling status, determine the offset direction and correct the sampling start point, and generate an energy abnormal channel identifier.

[0007] S2: Using the energy anomaly channel identifier, compare the harmonic amplitude of each order in multiple cycles, analyze the frequency band combination relationship of the maximum and second largest frequency points, determine the change characteristics of the combination in the periodic sequence based on the change direction and frequency of the combination, screen the spectral drift frequency band combination, and generate harmonic combination analysis results.

[0008] S3: Using the harmonic combination analysis results, detect and analyze multiple abnormal events in the circuit, analyze the start and end periods of multiple abnormal segments, identify the overlapping behavior of multiple abnormalities within the period, calculate the degree of abnormality of multiple period segments, classify the abnormal state of the circuit, and generate multi-parameter aggregated classification information.

[0009] S4: Based on the multi-parameter aggregated classification information, analyze the location of abnormal events, calculate the interval structure between adjacent cycles in each abnormality type, identify the clustering characteristics of abnormal events, construct the repetitive dense scoring results for each type of abnormality, evaluate the power quality level, and generate power quality assessment results.

[0010] As a further embodiment of the present invention, the energy anomaly channel identifier specifically includes a channel number, an energy difference direction identifier, and a sampling start point correction parameter; the harmonic combination analysis result includes a primary and secondary frequency band combination sequence label, a spectral variation trend parameter, and a drift frequency band identifier number; the multi-parameter aggregation classification information specifically includes anomaly type code, a period overlap degree label, and anomaly segment classification number; and the power quality assessment result includes anomaly type score items, a scoring path label, and a circuit level assessment label.

[0011] As a further aspect of the present invention, the step of obtaining the energy anomaly channel identifier specifically includes:

[0012] S111: Using a smart meter, periodic waveform sampling data of three-phase voltage and current signals are obtained. The instantaneous power of the channel is calculated based on the voltage and current at each sampling point. Then, the instantaneous power of each sampling point within the channel period is accumulated and integrated to obtain the periodic energy integral value of each channel and normalized to generate periodic energy data.

[0013] S112: Based on the periodic energy data, compare the energy performance differences of each channel, calculate the energy offset difference rate, and generate an energy structure offset index.

[0014] S113: Based on the energy structure offset index, identify abnormal sampling channels, determine the sampling offset direction, and perform periodic offset correction on the sampling start point position of the channel to obtain an energy abnormal channel identifier.

[0015] As a further aspect of the present invention, the steps for obtaining the harmonic combination analysis results are specifically as follows:

[0016] S211: Based on the energy anomaly channel identifier, collect amplitude data of each order harmonic within a continuous period, select the frequency point with the largest amplitude and the second largest amplitude frequency point for each period, and establish primary and secondary frequency band combination data by combining the corresponding frequency band width and time span.

[0017] S212: Based on the primary and secondary frequency band combination data, analyze the transformation direction and frequency of the target frequency band combination in a continuous period. Based on the primary and secondary frequency difference change and combination existence time parameter of each primary and secondary frequency band combination in a continuous period, calculate and obtain the spectral combination fluctuation coefficient of each frequency band combination, and establish a spectral fluctuation combination parameter set.

[0018] S213: Based on the spectral wave combination parameter set, by analyzing the variation characteristics of multiple frequency band combinations in the periodic sequence, the spectral drift frequency band combination is screened, and the harmonic combination analysis results are generated.

[0019] As a further aspect of the present invention, the step of obtaining the multi-parameter aggregated classification information specifically includes:

[0020] S311: Using the harmonic combination analysis results, detect and analyze multiple abnormal events in the circuit, including voltage drop, current change, and frequency shift, record the interval distribution of multiple abnormal events within the period, and generate abnormal event interval distribution parameters;

[0021] S312: Based on the abnormal event interval distribution parameters, compare the start and end intervals of multiple abnormal event types, filter the overlapping abnormal event intervals within the periodic segment, count the number and duration ratio of overlapping segments within each periodic segment, and calculate the abnormal clustering intensity index of each periodic segment in combination with the frequency of occurrence of event types.

[0022] S313: Call the aforementioned abnormal clustering intensity index, combine the abnormal clustering intensity of the periodic segment with the occurrence frequency of each type of event, identify and classify the circuit abnormal state of multiple periods, and obtain multi-parameter aggregated classification information.

[0023] As a further aspect of the present invention, the steps for obtaining the power quality assessment results are specifically as follows:

[0024] S411: Obtain the multi-parameter aggregated classification information, count the occurrence position of each type of abnormal event in the periodic sequence, record the occurrence period and order of each type of abnormal event, and generate abnormal event periodic distribution data;

[0025] S412: Based on the periodic distribution data of the abnormal events, analyze the distribution interval of each type of abnormal event in a continuous period, compare the time distribution characteristics of the same type of abnormal events in each week, identify the distribution clustering phenomenon, and obtain the abnormal event distribution clustering parameters.

[0026] S413: Call the abnormal event distribution aggregation parameters, analyze the periodic aggregation characteristics of each type of abnormal event, construct the repetitive dense scoring results of each type of abnormality, evaluate the power quality status of the target circuit in the periodic sequence, and generate power quality assessment results.

[0027] As a further aspect of the present invention, the method further includes:

[0028] S5: Using the power quality assessment results, analyze the change trajectory of voltage fluctuation, frequency change and harmonic amplitude parameters, select candidate segments with consistent trends according to the direction and amplitude of periodic changes, use them as input periodic segments for trend modeling, analyze and predict the power quality changes of the target circuit, and generate quality trend modeling data.

[0029] The quality trend modeling data specifically refers to the trend direction vector, periodic input structure, and modeling trigger identifier.

[0030] As a further aspect of the present invention, the steps for obtaining the quality trend modeling data are specifically as follows:

[0031] S511: Call the power quality assessment results to obtain the change trajectory of voltage fluctuation parameters, frequency change parameters and harmonic amplitude parameters in a continuous cycle, record the change direction and amplitude of the three parameters in each cycle, and generate a cycle parameter change sequence.

[0032] S512: Based on the cycle parameter change sequence, compare the change direction of the three parameters in each cycle, and combine the amplitude change to screen candidate segments with consistent trends to obtain cycle parameters with consistent trends.

[0033] S513: Using the trend-consistent periodic parameter, the target candidate segment is used as the trend modeling input periodic segment to analyze and predict the power quality changes of the target circuit and generate quality trend modeling data.

[0034] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0035] In this invention, by calculating the periodic energy performance and marking the sampling abnormal channels, and combining the frequency band combination features between harmonic amplitudes to identify spectral drift, the periodic overlap behavior is further analyzed in abnormal event detection, and the degree of abnormality of periodic segments is quantified. Multi-parameter classification results are constructed by aggregating event types, thereby identifying recurring abnormal patterns in the time series and constructing dense scoring accordingly. By combining the trends of voltage fluctuations, frequency changes, and harmonic amplitude changes, input periodic segments with consistent trends are extracted to achieve power quality trend prediction. This invention can integrate information from multiple dimensions to obtain potential fluctuation patterns from time series evolution, avoid misjudging the trend of power quality level changes, and improve the overall time series sensitivity and abnormal continuity judgment ability of the periodic series. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the main steps of the present invention;

[0037] Figure 2 This is a flowchart of the energy anomaly channel identifier acquisition process of the present invention;

[0038] Figure 3 This is a flowchart of the process for obtaining the harmonic combination analysis results of the present invention;

[0039] Figure 4 This is a flowchart of the multi-parameter aggregation classification information acquisition process of the present invention;

[0040] Figure 5 This is a flowchart of the process for obtaining power quality assessment results according to the present invention;

[0041] Figure 6 The flowchart illustrates the data acquisition process for quality trend modeling in this invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0043] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0044] Please see Figure 1 This invention provides a technical solution: a method for detecting the power quality of a smart meter, comprising the following steps:

[0045] S1: Using smart meters, calculate the instantaneous power of each channel and obtain the periodic energy performance by accumulating it. Compare the differences between channels, mark channels with abnormal sampling status, determine the offset direction and correct the sampling start point, and generate an energy abnormal channel identifier.

[0046] S2: Using energy anomaly channel identifiers, compare the harmonic amplitude of each order in multiple cycles, analyze the frequency band combination relationship of the maximum and second largest frequency points, determine the change characteristics of the combination in the periodic sequence based on the change direction and frequency of the combination, screen the spectral drift frequency band combination, and generate harmonic combination analysis results;

[0047] S3: Utilize the results of harmonic combination analysis to detect and analyze multiple abnormal events in the circuit, analyze the start and end cycles of multiple abnormal segments, identify the overlapping behavior of multiple types of abnormalities within the cycle, calculate the degree of abnormality of multiple cycle segments, classify the abnormal state of the circuit, and generate multi-parameter aggregated classification information.

[0048] S4: Based on multi-parameter aggregated classification information, analyze the location of abnormal events, calculate the interval structure between adjacent cycles in each abnormality type, identify the clustering characteristics of abnormal events, construct the repetitive dense scoring results for each type of abnormality, evaluate the power quality level, and generate power quality assessment results.

[0049] S5: Utilize power quality assessment results to analyze the trajectory of voltage fluctuations, frequency changes, and harmonic amplitude parameters. Select candidate segments with consistent trends based on the direction and amplitude of periodic changes, and use them as input periodic segments for trend modeling. Analyze and predict power quality changes in the target circuit, and generate quality trend modeling data.

[0050] The energy anomaly channel identifier specifically includes the channel number, energy difference direction identifier, and sampling start point correction parameter. The harmonic combination analysis results include the main and secondary frequency band combination sequence label, spectral variation trend parameter, and drift frequency band identifier number. The multi-parameter aggregation classification information specifically includes the anomaly type code, period overlap degree label, and anomaly segment classification number. The power quality assessment results include the anomaly type score item, scoring path label, and circuit level assessment label. The quality trend modeling data specifically refers to the trend direction vector, periodic input structure, and modeling trigger identifier.

[0051] Please see Figure 2 The specific steps for obtaining the energy anomaly channel identifier are as follows:

[0052] S111: Using a smart meter, periodic waveform sampling data of three-phase voltage and current signals are obtained. The instantaneous power of the channel is calculated based on the voltage and current at each sampling point. Then, the instantaneous power of each sampling point within the channel period is accumulated and integrated to obtain the periodic energy integral value of each channel and normalized to generate periodic energy data.

[0053] Based on the high-speed signal acquisition module embedded in the smart meter, voltage transformers and current transformers are connected to the three-phase voltage and current signal input terminals respectively to acquire the periodic waveform data of the three-phase AC input signals. The sampling frequency is set to 5kHz, the number of sampling points per cycle is 64, and the corresponding cycle duration is 12.8ms. In practice, the initial voltage sampling values ​​of phases A, B, and C are set to 220V, 218V, and 219V respectively, and the corresponding initial current sampling values ​​are 1.5A, 1.6A, and 1.4A. After the system starts, the voltage and current values ​​of the corresponding channels are extracted sequentially starting from the first sampling point, and the instantaneous power value at a single point is calculated. For example, the power at the first point of phase A is 220 × 10⁻¹⁰. =330W. This calculation continues until the 64th sampling point. Then, the power array for each phase is called, and each item is multiplied by the time interval Δt=0.0002s and accumulated to complete the integration operation. The periodic energy of phase A is obtained as follows: Σ(330,328,331,…,322)×0.0002s≈4.24J, phase B is 4.11J, and phase C is 4.07J. Thus, the periodic energy integral data of the three-phase channels are obtained. Then, based on the maximum value of 4.24J, the periodic energy of each channel is divided by the maximum value to obtain the normalized energy array {1.000,0.969,0.960}. This set of data is the periodic energy data.

[0054] S112: Based on periodic energy data, compare the differences in energy performance for each channel using the following formula:

[0055] ;

[0056] Calculate the energy offset difference rate and generate the energy structure offset index;

[0057] in, For channel Normalized energy shift difference rate, For channel The periodic energy normalization value, through the channel The instantaneous power integral is obtained by dividing by the periodic maximum energy integral. This refers to the channel index number currently being processed in the three-phase system. The normalized mean of the three-channel periodic energy is obtained through arithmetic averaging. For channel The sum of squares of the normalized energy differences from the other channels, where the subscripts are... For channel index, not equal to , For the first The normalized instantaneous power value at each sampling point is obtained by multiplying the instantaneous voltage and instantaneous current at that sampling point and then dividing by the maximum instantaneous power of the period. For the first The phase reference factor for each sampling point is obtained through the phase offset characteristics of the corresponding voltage signal. This represents the sampling point index, with a value range of [value range missing]. arrive , The total number of sampling points in each period. The normalized periodic time constant is obtained by dividing the total periodic sampling duration by the maximum period duration.

[0058] Based on the normalized periodic energy data, let the current processing channel be phase A, and its normalized energy be... The average energy of the three channels is calculated as follows: Take the absolute value of the difference Next, we process the first term in the denominator, the sum of squares of energy differences. Then, process the second term in the denominator by summing the weighted product of the normalized instantaneous power and the phase factor. Some data are shown in Table 1:

[0059] Table 1. Normalized power values ​​and phase factors at sampling points:

[0060] ;

[0061] The estimated sum of the 64 items in the period is:

[0062] ;

[0063] Combine the denominators:

[0064] ;

[0065] The final normalized energy shift difference rate is:

[0066] ;

[0067] Among them, the energy structure shift difference rate characterizes the degree of structural shift in the normalized energy performance of a certain channel in a three-phase current system relative to the system average state within a unit period. Its calculation considers not only the energy difference between this channel and other channels, but also the disturbance effect of sampling point power and phase shift on the system power structure. The larger the value of this parameter, the more prominent the energy and phase structure difference of this channel compared to other channels in the system. The specific effect of this parameter is that it can help accurately locate structurally abnormal channels in the sampling sequence caused by phase drift, providing a core basis for subsequent sampling point synchronization correction and power balance reconstruction, significantly enhancing the accuracy of anomaly detection at the system level. Based on the set shift judgment threshold of 0.008, because... If phase A is the normal channel, the same calculation process is continued for phases B and C to obtain a complete array of offset difference rates as the energy structure offset index. The formula introduces a weighted structure in the denominator, multiplying the normalized instantaneous power value by the phase factor, and combines this with the sum of squared energy differences to determine the energy offset difference rate. This reflects both the overall structural deviation between the current channel and other channels, and the microscopic power characteristics of the sampling points to reflect the fluctuations within the period. This allows the index to simultaneously consider both the lateral comparison between channels and the longitudinal power characteristics within the period, enhancing the accuracy and rationality of offset identification.

[0068] S113: Based on the energy structure offset index, identify abnormal sampling channels, determine the sampling offset direction, and perform periodic offset correction on the sampling start point position of the channel to obtain the energy abnormal channel identifier.

[0069] After obtaining the complete three-channel energy structure offset index array, the system sequentially processes each channel... The value is compared with a threshold of 0.008. For example, if the B-phase offset difference rate is 0.011, it is judged as an abnormal channel because 0.011 > 0.008, and is marked into the abnormal list. Then, the offset direction of the channel is further analyzed by comparison. and The magnitude of the phase shift is determined by the relationship between the phase values. For example, if the normalized value of phase B is 0.969, which is less than the mean of 0.976, it is judged as a low offset. Then, the offset trend of the starting point is judged by the distribution of its phase factor. The mean of the phase factor of the first 32 sampling points is compared with the mean of the last 32 sampling points. If the mean of the first segment is 1.02 and the mean of the last segment is 1.05, then because the phase of the last segment is higher, the channel sampling offset direction is considered to be forward. The starting sampling point needs to be adjusted to be shifted backward. The number of shifts is determined according to the estimated phase offset amplitude. In the example, the shift is 3 points, and the original starting point is adjusted from point 1 to point 4. At the same time, the correction mark is recorded to form the final energy anomaly channel mark set for input in subsequent steps.

[0070] Please see Figure 3The specific steps for obtaining the harmonic combination analysis results are as follows:

[0071] S211: Based on the energy anomaly channel identifier, collect amplitude data of each order harmonic within a continuous period, select the frequency point with the largest amplitude and the second largest amplitude frequency point for each period, and establish the main and secondary frequency band combination data by combining the corresponding frequency band width and time span.

[0072] Based on the energy anomaly channel identification, channels identified as anomaly are extracted from the sampling channels for continuous periodic frequency domain analysis. The analysis period is set to 10 complete power grid cycles. The harmonic frequency points collected in each cycle range from 0 to 2 kHz. The FFT transform method is used to perform spectral analysis on the voltage signal of each cycle, extracting the amplitude data of all harmonics and their corresponding frequency components within each cycle. A frequency-amplitude pair mapping structure is constructed. In each cycle, the frequency corresponding to the maximum amplitude is found from the frequency distribution results and recorded as the dominant frequency point, and the frequency corresponding to the second largest amplitude is recorded as the secondary frequency point. For example, in cycle 1, the dominant frequency point is 500 Hz with an amplitude of 0.75 V, and the secondary frequency point is 600 Hz with an amplitude of 0.68 V; these are then marked as the dominant and secondary frequency band combination F. F1=500Hz, F2=600Hz, the recorded frequency band width Δf=F2−F1=100Hz, and the continuous existence duration of this combination in the current cycle T1=6ms are recorded. A normalization strategy is adopted, with the maximum possible frequency of 2000Hz as the benchmark, F1 and F2 are normalized to 0.25 and 0.3 respectively, and the time parameter is normalized to 0.3 with the maximum cycle length of 20ms. Finally, the combination data vector (0.25,0.3,0.3) in this cycle is formed. All abnormal channels are processed in 10 cycles to form multiple combination data records. Each group of primary and secondary frequency band combinations is tracked in subsequent cycles. Finally, a primary and secondary frequency band combination dataset containing combination frequency, amplitude, frequency band width and duration is established.

[0073] S212: Based on the primary and secondary frequency band combination data, analyze the transformation direction and frequency of the target frequency band combination in a continuous period. Based on the change in the primary and secondary frequency difference of each primary and secondary frequency band combination in a continuous period and the combination existence time parameter, the following formula is used:

[0074] ;

[0075] The spectral combination fluctuation coefficients for each frequency band combination are obtained through calculation, and a spectral combination fluctuation parameter set is established.

[0076] in, Let be the spectral combination fluctuation coefficient, representing the th The intensity of the normalized frequency difference variation in the combination of primary and secondary frequency bands over a continuous period. For the first Within the cycle The normalized frequency value of the group's main frequency point is obtained by acquiring the maximum amplitude point frequency of each cycle and then normalizing it. For the first Within the cycle The normalized frequency values ​​of the grouped frequencies are obtained by acquiring the frequencies of the second largest amplitude points in each cycle and then normalizing them. For the first Within the cycle The normalized duration of a combination is obtained by the ratio of the duration of the combination's occurrence within a period to the total duration of the maximum period. The number of cycles continuously involved in the calculation is obtained by counting the total number of cycles in which the primary and secondary frequency bands combine within a continuous period segment. This represents the index number in the periodic sequence, indicating the position of the consecutive period being compared. Indicates at index Based on the current period, it advances one period forward to establish a comparison relationship between adjacent periods, meaning that the first period... The 1st period, that is, the next period in the periodic sequence, is used to connect with the 1st period. The period constitutes the difference comparison term. The combination index number represents different combinations of primary and secondary frequency bands, used to distinguish multiple combination objects;

[0077] Based on the primary and secondary frequency band combination dataset, the difference and fluctuation trend analysis of the continuous occurrence behavior of each combination in the periodic sequence is performed, and the processing group is set as the [number missing]. A group, specifically a combination with a primary frequency of 0.25 (corresponding to 500Hz) and a secondary frequency of 0.3 (corresponding to 600Hz), is defined by the number of consecutive cycles. This indicates that the combination exists in all 5 periods. The normalized main frequency and secondary frequency data of the combination in periods 1 to 5 are extracted as shown in the table below.

[0078] Table 2: Normalized data of primary and secondary frequency points:

[0079] ;

[0080] As shown in Table 2, the number of records is as follows: The normalized frequency values ​​and durations of the main and secondary frequency band combinations over five consecutive periods provide basic parameter support for the calculation of the spectral combination fluctuation coefficient.

[0081] Calculate in sequence Divide by For example, item 1:

[0082] ;

[0083] ;

[0084] ;

[0085] And so on. arrive Therefore, we can conclude that:

[0086] ;

[0087] The spectral pattern combination fluctuation coefficient is a composite index used to measure the rate of change of normalized frequency difference and the degree of temporal rhythm fluctuation of a certain type of primary and secondary frequency band combination in a continuous period. Specifically, it represents a joint variability of the rate of change of the frequency band combination pattern and its temporal structure under normalized dimensions, reflecting the instability of the frequency band combination in the overall spectral pattern evolution process. A larger value indicates more drastic and faster changes in the primary and secondary frequency differences of the frequency band combination, and a more unstable structure. Its calculation effect is mainly used to identify which primary and secondary frequency band combinations exhibit continuous variability or trend-based structural drift characteristics, serving as an important basis for judging spectral pattern drift combinations and supporting the detection of harmonic anomalies and the location of structural disturbances. The formula extracts the frequency instability intensity of the spectral pattern combination in the time series by weighted fusion of the frequency difference variation between the primary and secondary frequencies and the combination duration, quantifying the fluctuation intensity and spectral structure drift behavior of the combination in the periodic sequence. The results indicate that the frequency difference fluctuation amplitude of the first group of frequency bands within a continuous period is 0.00306, which is below the set spectral stability threshold of 0.005, and is preliminarily judged to be a spectrally stable combination.

[0088] S213: Based on the spectral wave combination parameter set, by analyzing the variation characteristics of multiple frequency band combinations in the periodic sequence, the spectral drift frequency band combination is screened, and the harmonic combination analysis results are generated.

[0089] Spectral fluctuation coefficients for all frequency band combinations After calculation, the system iterates through each combination coefficient and compares it with a preset drift threshold of 0.005. If the combination fluctuation coefficient is greater than the threshold, it is marked as a drift frequency band combination; otherwise, it is marked as a stable combination. For example, the coefficient of combination 2 is 0.00678. Since 0.00678 > 0.005, it is marked as a drift frequency band combination. Then, it records the start and end period indices and labels the corresponding primary and secondary frequency points. It counts all combinations with fluctuation coefficients greater than the threshold among multiple combinations and summarizes the frequency difference trend sequence of each combination. For example, in combination 2, the frequency difference changes from -0.049, -0.046, -0. The changes of .050 and -0.054 indicate that the first-order difference sequence is alternating between positive and negative values, indicating a high-frequency change structure. The proportion of the combination within the entire 10 cycles is then statistically analyzed. For example, if it exists in cycles 3 to 8 (6 cycles in total), compared to the total number of cycles (10), it accounts for 60%. Since the screening threshold is set to 50%, the screening conditions are met. Finally, the frequency band combinations that meet the conditions of fluctuation intensity and time proportion are output as a set of spectral drift frequency band combinations. At the same time, the information on the primary and secondary frequency points, frequency band width, and occurrence frequency in the drift combination are compiled into harmonic combination analysis results, which serve as the input basis for the correlation of abnormal events in the next stage.

[0090] Please see Figure 4 The specific steps for obtaining multi-parameter aggregated classification information are as follows:

[0091] S311: Using the results of harmonic combination analysis, detect and analyze multiple abnormal events in the circuit, including voltage drop, current change, and frequency shift, record the interval distribution of various abnormal events within the period, and generate abnormal event interval distribution parameters;

[0092] Using harmonic combination analysis results, an event detection task was performed on the sampled data over 10 consecutive cycles. Three types of abnormal events were defined: voltage drop, current surge, and frequency shift. The minimum amplitude of the voltage waveform within a single cycle was extracted and compared to a reference voltage of 220V. If the amplitude was below 198V (a threshold of 220V × 0.9 was set), it was marked as a voltage drop event. The start and end sampling points of the cycle containing this abnormal point were also recorded. For example, in cycle 3, the voltage from point 10 to point 24 was below 198V, corresponding to the abnormal interval... In the detection of sudden current events, the rate of change between sampling points before and after is used to determine anomalies. If the rate of change... If the threshold is exceeded by 10%, it is identified as a mutation point, and an abnormal segment is constructed by extending the sampling points forward and backward by 3 points. For example, the mutation occurs at point 17 in period 5, forming an interval. Frequency drift detection is based on the analysis of the amplitude of harmonic dominant frequency changes. If the frequency drift exceeds the set threshold of ±3Hz, the corresponding period is considered an abnormal period, with the range being [missing information]. After all abnormal events have been identified, each type of abnormality is recorded as interval information according to the start and end sampling points, and then normalized. For example, if the total sampling point for a period is 64 points, the event interval is... Normalized length is Finally, an anomaly distribution table for the three types of events in each cycle was constructed.

[0093] S312: Based on the distribution parameters of abnormal event intervals, compare the start and end intervals of multiple abnormal event types, filter out overlapping abnormal event intervals within the periodic segment, count the number and duration ratio of overlapping segments within each periodic segment, and combine this with the frequency of event type occurrences using the formula:

[0094] ;

[0095] Calculate the abnormal clustering intensity index for each periodic segment;

[0096] in, The anomaly clustering intensity within a periodic segment represents the degree to which different anomalous events co-cluster in time and frequency within that segment. It is obtained by statistically analyzing the length of overlapping segments and the normalized frequency fluctuations of multiple event types. The first in the periodic segment The normalized length of each overlapping segment represents the degree of overlap between different abnormal events within the same time period. It is obtained by comparing the overlap of normalized intervals for the start and end periods of each type of abnormal event. The normalized total length of the periodic segments represents the proportion of the periodic segment length used for aggregation analysis in the total periodic sequence. It is obtained by normalizing the original period lengths. The first in the periodic segment The occurrence count of a certain type of abnormal event indicates the frequency with which a particular type of event is identified within a given period. This is obtained by detecting the boundaries of abnormal event intervals and accumulating the statistical count of that type of event within the interval. This represents the average number of occurrences of all abnormal events within a periodic segment, serving as a reference benchmark value for measuring the frequency fluctuations of different events. The arithmetic mean of the sums of the frequencies of the events is obtained. This represents the number of overlapping segments within a periodic interval, indicating the number of segments in which all types of abnormal events intersect within that periodic interval. It is obtained by comparing the intersection of the start and end intervals of each type of abnormal event. The total number of abnormal event types involved in the calculation represents the number of all independent abnormal types involved in this round of calculation, such as voltage drop, current surge, and frequency shift. This is a fixed, preset total number of types. Indicates the period segment number of the current analysis. This indicates the statistical result within the periodic interval. Number of overlapping segments Indicates the type number of the exception event;

[0097] Based on the abnormal event intervals, the start and end positions of various events are extracted and normalized. Event intervals within the same period are compared pairwise to determine if they overlap. If the start and end ranges of the intervals intersect, they are recorded as a single overlapping segment, and its normalized length is calculated. For example, in cycle 7, the voltage drops to The current suddenly becomes The overlapping interval is ,but After completing the overlapping segment detection for all cycles in sequence, the total length of the cycle segment is calculated. For example, a segment consisting of periods 6 to 8 covers 3 × 20 ms = 60 ms, with a maximum duration of 100 ms after normalization. Then extract the number of occurrences of each event within that segment. For example, if there are two voltage drops, three current abrupt changes, and one frequency shift, calculate the average value of these three types of events. Substitute into the following formula:

[0098] ;

[0099] Let the number of overlapping segments be... Its corresponding , ,but:

[0100] ;

[0101] Then calculate the frequency deviation term:

[0102] ;

[0103] Final calculation:

[0104] ;

[0105] Among them, the anomaly clustering intensity index is a core indicator used to quantify whether multiple abnormal events occur simultaneously and their degree of clustering within a circuit cycle segment. A higher value indicates that abnormal events (such as voltage drops, current surges, and frequency shifts) occur more concentratedly and simultaneously within that cycle segment, with smaller differences in their distribution types. Therefore, this index can not only be used to determine the existence of compound abnormal behavior, but also for subsequent key analytical tasks such as anomaly type attribution, power quality level assessment, and trend prediction. It is an important parameter reflecting the anomaly clustering state of a cycle segment. By comparing this index across multiple cycle segments, dense anomaly areas can be effectively identified, providing an efficient basis for circuit fault early warning and classification. The formula combines the degree of overlap of abnormal events with frequency fluctuations, assessing both the synchronicity of events on the time axis and the consistency of their type distribution, thus comprehensively characterizing the spatial density and temporal consistency intensity of circuit abnormal events. The results show that the anomaly clustering intensity in the 3rd cycle segment is 0.4154, combined with the preset clustering intensity judgment threshold of 0.35, therefore... It was identified as a region with high abnormal aggregation.

[0106] S313: Call the abnormal clustering intensity index, combine the abnormal clustering intensity of the periodic segment with the occurrence frequency of each type of event, identify and classify the circuit abnormal state of multiple periods, and obtain multi-parameter aggregated classification information.

[0107] The abnormal clustering intensity index of each period segment According to the segment number, the regions are divided into two categories: strong clustering and weak clustering. The clustering threshold is set to 0.35. Segments with an anomaly intensity index greater than this value are marked as high-clustering segments. The frequency distribution of various abnormal events is analyzed in high-clustering segments, and event type distribution vectors are constructed. For example, the event frequency in the periodic segment q=3 is vector [2,3,1]. It is compared with other segments such as q=4 [1,1,0] to extract the differences in event occurrence patterns. After clustering according to the event frequency matrix, it is classified into three abnormal state types, named voltage-dominated, current-dominated, and mixed drift type. Among them, the concentration of voltage-dominated events accounts for more than 50% in the first category. In the current periodic segment q=3, the voltage drop frequency is 2, the total frequency is 6, accounting for 33%, and it is classified as mixed drift type. Finally, all periodic segments are classified according to the frequency vector of various events to obtain a complete multi-parameter aggregated classification information list, which serves as the input basis for subsequent modeling.

[0108] Please see Figure 5 The specific steps for obtaining power quality assessment results are as follows:

[0109] S411: Obtain multi-parameter aggregated classification information, count the occurrence position of each type of abnormal event in the periodic sequence, record the occurrence period and order of each type of abnormal event, and generate abnormal event periodic distribution data;

[0110] After constructing multi-parameter aggregated classification information, the system first stores the classified abnormal events by type and establishes a periodic index mapping table. Then, it retrieves the occurrence of each type of abnormal event in chronological order, recording the period number of its first occurrence for each type of abnormal event, and continuously adding records of the position of the event recurring in subsequent periods, forming a periodic distribution set of the event in the periodic sequence. For example, if a voltage drop event is detected in periods 1, 3, 4, 6, and 9, it is written into the event distribution table as 1, 3, 4, 6, 9; a current surge event occurs in periods 2, 5, 6, and 8, and is recorded as 2, 5, 6, 8; a frequency shift event occurs in periods 3, 4, and 7, and is recorded as 3, 4, 7. After recording, each set of periodic position data is associated with the corresponding event type, and a sequence number is added to indicate the first, second, and subsequent occurrence order of the event. For example, the "3rd occurrence" of a voltage drop is located in period 4. Finally, a periodic distribution dataset containing fields such as event type, occurrence period, and event order is obtained and used for subsequent distribution interval analysis processing.

[0111] S412: Based on the periodic distribution data of abnormal events, analyze the distribution interval of each type of abnormal event in a continuous period, compare the time distribution characteristics of the same type of abnormal events in each week, identify the distribution clustering phenomenon, and obtain the abnormal event distribution clustering parameters.

[0112] Based on the position of various abnormal events in the periodic sequence, the periodic interval information between events of the same type is extracted sequentially. The difference in the period numbers of adjacent events is used as the interval value. The calculated interval sequence is used to characterize the density of the periodic distribution. For example, the periodic distribution of voltage drop events is 1, 3, 4, 6, 9, with corresponding intervals of 2, 1, 2, 3; the periodic distribution of current surge events is 2, 5, 6, 8, with intervals of 3, 1, 2; and the periodic distribution of frequency shift events is 3, 4, 7, with intervals of 1, 3. Each set of interval data is statistically analyzed into four indicators: average value, minimum value, maximum value, and fluctuation range. The average value is used to reflect whether abnormal events are periodically concentrated. For example, the average interval for voltage drop is 2.0, the average interval for frequency shift is 2.0, and the average interval for current surge is 2.0. The interval characteristics of the events are initially similar; the minimum interval is used to determine whether there is a concentrated outbreak phenomenon. For example, if the minimum interval is 1, it indicates that the events occur continuously in at least some periods; the fluctuation range (maximum value minus minimum value) can measure the stability of the distribution. For example, if the frequency offset fluctuation range is 2, it is a medium fluctuation type. Based on the above information system, the clustering level of each type of abnormal event is classified. The characteristics of clustered events are low average interval and low fluctuation value. In the current data, the voltage drop event is marked as a high clustering event when the interval is consistent and the fluctuation range is low. The frequency offset is marked as a medium clustering event because the fluctuation value is slightly higher. The current change is a stable clustering type. All classification results and statistical data are recorded together as abnormal event distribution clustering parameters.

[0113] S413: Call the abnormal event distribution aggregation parameters, analyze the periodic aggregation characteristics of each type of abnormal event, construct the repetitive dense scoring results of each type of abnormality, evaluate the power quality status of the target circuit in the periodic sequence, and generate power quality assessment results.

[0114] After completing the aggregation parameter analysis of various anomalies, the system scores and summarizes the recurrence of anomalies. Combining the average interval and distribution fluctuation of each type of anomaly, and referring to the total length of the periodic sequence and the number of event types, the system assigns a recurrence density score to each type of anomaly. The scoring results are shown in the table below:

[0115] Table 3. Scoring Table for Repeated and Dense Abnormal Events:

[0116] ;

[0117] As shown in Table 3, the average occurrence intervals of the three types of events within the 10-cycle range are relatively close, but the standard deviation of the voltage drop event is slightly smaller, indicating that its recurrence location is more concentrated, thus its score is slightly higher; the frequency offset score is the lowest, representing that its event distribution is more loose. The scores of the three types of events are summarized into an overall score, and the average of the three scores is calculated as (0.597+0.565+0.560) / 3≈0.574. This score is in the medium to weak range of the five-level power quality assessment system, and the power quality level is rated as "Level 3," indicating that the current circuit system has frequent mild to moderate abnormal interference. It is recommended to continuously monitor and focus on the frequently occurring event segments. Finally, the clustering of each type of event, the score value, and the level label are combined into a complete power quality assessment result output, which serves as a direct input for subsequent trend prediction modeling.

[0118] Please see Figure 6 The specific steps for obtaining quality trend modeling data are as follows:

[0119] S511: Call the power quality assessment results, obtain the change trajectory of voltage fluctuation parameters, frequency change parameters and harmonic amplitude parameters in a continuous cycle, record the change direction and amplitude of the three parameters in each cycle, and generate a cycle parameter change sequence.

[0120] After obtaining the power quality assessment results, the system further analyzes the key parameters of each cycle, extracting three types of power quality indicators for each cycle: voltage fluctuation parameters, frequency variation parameters, and harmonic amplitude parameters. The voltage fluctuation parameter refers to the normalized value of the difference between the maximum and minimum voltage values ​​within a cycle. For example, in cycle 1, the maximum voltage is 230V and the minimum voltage is 210V; the normalized fluctuation is... The frequency variation parameter is the percentage difference between the dominant frequency drift value and the standard 50Hz during the period. For example, if the dominant frequency for period 3 is 49.3Hz, the normalized variation is... The harmonic amplitude parameter refers to the proportion of the sum of the amplitudes of the primary and secondary harmonics in a period to the fundamental frequency amplitude. For example, in period 5, the combined amplitudes of the 3rd and 5th harmonics are 18V, and the fundamental frequency is 220V. Normalized to... For each cycle, the changes in these three parameters are recorded, and their direction of change compared to the previous cycle is calculated: an increase is marked as "+1", a decrease as "-1", and remaining essentially unchanged as "0". For example, if the frequency change value in cycle 4 is lower than that in cycle 3, the direction of frequency change in that cycle is marked as "-1". Ultimately, each cycle forms a "direction + amplitude" structure for the three parameters, which are sequentially arranged to generate a sequence of cycle parameter changes, used to identify trend correlation segments.

[0121] S512: Based on the cycle parameter change sequence, compare the change direction of the three parameters in each cycle, and combine the amplitude change to screen candidate segments with consistent trends to obtain cycle parameters with consistent trends.

[0122] Based on the cyclical parameter change sequence, the system compares the consistency of the three parameter change directions for each cycle, using a consistency criterion: if at least two of the three change direction values ​​are the same, the cycle is marked as a "consistent trend cycle"; if all three directions are completely consistent, it is marked as a "strongly consistent trend cycle". For example, in cycle 6, the voltage direction is "+1", the frequency is "+1", and the harmonic is "+1", which is strongly consistent; if in cycle 7, the voltage is "+1", the frequency is "-1", and the harmonic is "+1", it is generally consistent. These consistent trend cycles are further compared to see if their parameter amplitudes are within the same variation level range. The variation level range is divided by actual values: less than 0.01 is "slight variation", 0.01–0.05 is "moderate variation", and greater than 0.05 is "drastic variation". If at least two of the three parameters are within the same level range, the cycle can be included in the candidate segment for consistent trend. In the example, cycle 6 has amplitudes of 0.02, 0.025, and 0.03, all belonging to the moderate variation level, and is judged to be a consistent trend segment. The system then filters out three or more consecutive trend-consistent periodic segments from the periodic sequence. For example, if period 5–7 consecutively meets the condition, it is recorded as a candidate segment. Finally, a set of all trend periodic segments that satisfy both direction consistency and amplitude level consistency is compiled as the trend-consistent periodic parameters required for subsequent modeling.

[0123] S513: Using trend-consistent periodic parameters, the target candidate segment is used as the input periodic segment for trend modeling to analyze and predict the power quality changes of the target circuit and generate quality trend modeling data.

[0124] After extracting the consistent trend periodic parameters, the system constructs the input structure of the power quality trend prediction model using candidate trend segments as input. Each candidate segment includes a period number, a sequence of changes in three indicators, a label indicating the direction of change, and a numerical range. The system performs sequence fitting on the parameter change curves within the trend segment, recording the overall increase / decrease trend, fluctuation range, and rhythm information of each indicator within the trend segment. For example, in period 5–7, the voltage fluctuation parameter increases from 0.06 to 0.09 periodically, the frequency changes from 0.015 to 0.011, and the harmonic amplitude oscillates between 0.07 and 0.08. Based on this, structured modeling data items are generated: [period start point = 5, number of periods = 3, voltage trend = increasing, frequency trend = decreasing, harmonic trend = oscillating]. Average, maximum, and minimum values ​​are added as data feature sets for modeling. Finally, the trend description results of all candidate segments are output and stored as quality trend modeling data in the subsequent prediction module. This result data is used to support the judgment and prediction of the future cyclical evolution trend of the target circuit's power quality in subsequent stages.

[0125] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for detecting power quality in a smart meter, characterized in that, Includes the following steps: S1: Using smart meters, calculate the instantaneous power of each channel and obtain the periodic energy performance by accumulating it. Compare the differences between channels, mark channels with abnormal sampling status, determine the offset direction and correct the sampling start point, and generate an energy abnormal channel identifier. S2: Using the energy anomaly channel identifier, compare the harmonic amplitude of each order in multiple cycles, analyze the frequency band combination relationship of the maximum and second largest frequency points, determine the change characteristics of the combination in the periodic sequence based on the change direction and frequency of the combination, screen the spectral drift frequency band combination, and generate harmonic combination analysis results. S3: Using the harmonic combination analysis results, detect and analyze multiple abnormal events in the circuit, analyze the start and end periods of multiple abnormal segments, identify the overlapping behavior of multiple abnormalities within the period, calculate the degree of abnormality of multiple period segments, classify the abnormal state of the circuit, and generate multi-parameter aggregated classification information. The specific steps for obtaining the multi-parameter aggregated classification information are as follows: S311: Using the harmonic combination analysis results, detect and analyze multiple abnormal events in the circuit, including voltage drop, current change, and frequency shift, record the interval distribution of multiple abnormal events within the period, and generate abnormal event interval distribution parameters; S312: Based on the abnormal event interval distribution parameters, compare the start and end intervals of multiple abnormal event types, filter the overlapping abnormal event intervals within the periodic segment, count the number and duration ratio of overlapping segments within each periodic segment, and calculate the abnormal clustering intensity index of each periodic segment in combination with the frequency of occurrence of event types. S313: Call the aforementioned abnormal clustering intensity index, combine the abnormal clustering intensity of the periodic segment with the occurrence frequency of each type of event, identify and classify the circuit abnormal state of multiple periods, and obtain multi-parameter aggregated classification information; S4: Based on the multi-parameter aggregated classification information, analyze the location of abnormal events, calculate the interval structure between adjacent cycles in each abnormality type, identify the clustering characteristics of abnormal events, construct the repetitive dense scoring results for each type of abnormality, evaluate the power quality level, and generate power quality assessment results.

2. The power quality detection method for smart meters according to claim 1, characterized in that, The energy anomaly channel identifier specifically includes channel number, energy difference direction identifier, and sampling start point correction parameter. The harmonic combination analysis result includes primary and secondary frequency band combination sequence label, spectral variation trend parameter, and drift frequency band identifier number. The multi-parameter aggregation classification information specifically includes anomaly type code, period overlap degree label, and anomaly segment classification number. The power quality assessment result includes anomaly type score item, scoring path label, and circuit level assessment label.

3. The power quality detection method for smart meters according to claim 1, characterized in that, The specific steps for obtaining the energy anomaly channel identifier are as follows: S111: Using a smart meter, periodic waveform sampling data of three-phase voltage and current signals are obtained. The instantaneous power of the channel is calculated based on the voltage and current at each sampling point. Then, the instantaneous power of each sampling point within the channel period is accumulated and integrated to obtain the periodic energy integral value of each channel and is normalized to generate periodic energy data. S112: Based on the periodic energy data, compare the energy performance differences of each channel, calculate the energy offset difference rate, and generate an energy structure offset index. S113: Based on the energy structure offset index, identify abnormal sampling channels, determine the sampling offset direction, and perform periodic offset correction on the sampling start point position of the channel to obtain the energy abnormal channel identifier.

4. The power quality detection method for smart meters according to claim 3, characterized in that, The specific steps for obtaining the harmonic combination analysis results are as follows: S211: Based on the energy anomaly channel identifier, collect amplitude data of each order harmonic within a continuous period, select the frequency point with the largest amplitude and the second largest amplitude frequency point for each period, and establish primary and secondary frequency band combination data by combining the corresponding frequency band width and time span. S212: Based on the primary and secondary frequency band combination data, analyze the transformation direction and frequency of the target frequency band combination in a continuous period. Based on the primary and secondary frequency difference change and combination existence time parameter of each primary and secondary frequency band combination in a continuous period, calculate and obtain the spectral combination fluctuation coefficient of each frequency band combination, and establish a spectral fluctuation combination parameter set. S213: Based on the spectral wave combination parameter set, by analyzing the variation characteristics of multiple frequency band combinations in the periodic sequence, the spectral drift frequency band combination is screened, and the harmonic combination analysis results are generated.

5. The power quality detection method for smart meters according to claim 1, characterized in that, The specific steps for obtaining the power quality assessment results are as follows: S411: Obtain the multi-parameter aggregated classification information, count the occurrence position of each type of abnormal event in the periodic sequence, record the occurrence period and order of each type of abnormal event, and generate abnormal event periodic distribution data; S412: Based on the periodic distribution data of the abnormal events, analyze the distribution interval of each type of abnormal event in a continuous period, compare the time distribution characteristics of the same type of abnormal events in each week, identify the distribution clustering phenomenon, and obtain the abnormal event distribution clustering parameters. S413: Call the abnormal event distribution aggregation parameters, analyze the periodic aggregation characteristics of each type of abnormal event, construct the repetitive dense scoring results of each type of abnormality, evaluate the power quality status of the target circuit in the periodic sequence, and generate power quality assessment results.

6. The power quality detection method for smart meters according to claim 1, characterized in that, The method further includes: S5: Using the power quality assessment results, analyze the change trajectory of voltage fluctuation, frequency change and harmonic amplitude parameters, select candidate segments with consistent trends according to the direction and amplitude of periodic changes, use them as input periodic segments for trend modeling, analyze and predict the power quality changes of the target circuit, and generate quality trend modeling data. The quality trend modeling data specifically refers to the trend direction vector, periodic input structure, and modeling trigger identifier.

7. The power quality detection method for smart meters according to claim 6, characterized in that, The specific steps for obtaining the quality trend modeling data are as follows: S511: Call the power quality assessment results to obtain the change trajectory of voltage fluctuation parameters, frequency change parameters and harmonic amplitude parameters in a continuous cycle, record the change direction and amplitude of the three parameters in each cycle, and generate a cycle parameter change sequence. S512: Based on the cycle parameter change sequence, compare the change direction of the three parameters in each cycle, and combine the amplitude change to screen candidate segments with consistent trends to obtain cycle parameters with consistent trends. S513: Using the trend-consistent periodic parameter, the target candidate segment is used as the trend modeling input periodic segment to analyze and predict the power quality changes of the target circuit and generate quality trend modeling data.

Images