A household photovoltaic inverter fault early warning method and system

Abstract

The application discloses a household photovoltaic inverter fault early warning method and system, comprising: collecting a household photovoltaic historical electrical parameter set sequence, extracting the valley value and the peak value respectively, obtaining a historical valley value electrical parameter sequence and a historical peak value electrical parameter sequence, and calculating their approximation degrees with a first fault electrical threshold and a second fault electrical threshold respectively to obtain a first approximation degree sequence and a second approximation degree sequence; individually analyzing the discreteness of the two approximation degree sequences, and collectively analyzing the discreteness of the same district monitoring data to obtain individual approximation degree discrete parameters and district approximation degree discrete parameters; and classifying the fault early warning according to the individual approximation degree discrete parameters and the district approximation degree discrete parameters to obtain a power grid fault early warning parameter and a component fault early warning parameter. The application solves the technical problem that the prior art cannot effectively distinguish whether the household photovoltaic inverter fault is originated from the power grid side or the inverter component side, thereby leading to low operation and maintenance efficiency.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power generation fault diagnosis technology, specifically to a fault early warning method and system for residential photovoltaic inverters. Background Technology

[0002] As the core equipment of distributed photovoltaic (PV) power generation systems, the operational reliability of residential PV inverters directly affects the safety and economy of the PV power generation system. With the rapid growth of residential PV installed capacity, inverter fault early warning technology has become one of the key technologies to ensure the safe operation of residential PV systems.

[0003] Residential photovoltaic inverters generate a large amount of operational data, making it difficult to extract effective information. Furthermore, there is a lack of unified fault quantification standards across different devices. When an inverter malfunctions, existing methods struggle to distinguish whether the fault originates from the grid side or the inverter's own components. Early warning results lack a clear indication of the fault's source, leading to insufficient targeting and inefficiency in maintenance and repair. Summary of the Invention

[0004] This invention provides a fault early warning method and system for residential photovoltaic inverters. This invention solves the technical problem of existing technologies being unable to effectively distinguish whether a fault in a residential photovoltaic inverter originates from the grid side or from the inverter's own components, leading to low operation and maintenance efficiency.

[0005] In view of the above problems, the present invention provides a fault early warning method for residential photovoltaic inverters, the method comprising: Collect the historical electrical parameter set sequence of household photovoltaic in the past preset time range, and extract the valley value and peak value respectively to obtain the historical valley value electrical parameter sequence and the historical peak value electrical parameter sequence; The approximation degree between the historical valley electrical parameter sequence and the first fault electrical threshold, and between the historical peak electrical parameter sequence and the second fault electrical threshold are calculated respectively to obtain the first approximation degree sequence and the second approximation degree sequence; Individual discreteness analysis was performed on the first and second approximation sequences, and collective discreteness analysis was performed on the monitoring data of the same station area to obtain individual approximation discreteness parameters and station area approximation discreteness parameters. Based on the individual approximation degree discrete parameters and the transformer area approximation degree discrete parameters, fault warning classification is performed to obtain power grid fault warning parameters and component fault warning parameters.

[0006] The present invention also provides a fault early warning system for residential photovoltaic inverters, comprising: The data acquisition and feature extraction module is used to collect the historical electrical parameter set sequence of household photovoltaic systems within a preset time range, extract the valley value and peak value respectively, and obtain the historical valley value electrical parameter sequence and the historical peak value electrical parameter sequence. The approximation calculation module is used to calculate the approximation degree between the historical valley electrical parameter sequence and the first fault electrical threshold, and between the historical peak electrical parameter sequence and the second fault electrical threshold, respectively, to obtain the first approximation degree sequence and the second approximation degree sequence; The discreteness analysis module is used to perform individual discreteness analysis on the first approximation sequence and the second approximation sequence, and to perform collective discreteness analysis on the monitoring data of the same station area, so as to obtain individual approximation discreteness parameters and station area approximation discreteness parameters. The fault warning classification module is used to classify fault warnings based on the individual proximity degree discrete parameters and the distribution area proximity degree discrete parameters, and to obtain power grid fault warning parameters and component fault warning parameters.

[0007] One or more technical solutions provided in this invention have at least the following technical effects or advantages: This invention extracts valley and peak values ​​within a unit time window, compressing massive amounts of high-frequency sampling data into ordered feature sequences, significantly reducing the data processing burden. Simultaneously, using a natural day as the statistical period fully aligns with the daily cycle of photovoltaic power generation, naturally separating normal daily variations from abnormal fluctuations, thus reducing false alarms. By acquiring fault electrical thresholds and calculating the approximation degree between the valley and peak value sequences and their respective thresholds, the original voltage data is uniformly transformed into an approximation degree sequence. This quantifies the safety margin of the inverter from faults, providing a clear numerical basis for early warning decisions. It also eliminates dimensional differences between different devices, making the monitoring data of multiple inverters in a distribution area comparable horizontally. By calculating the variance and weighted summing of the valley and peak value approximation degree sequences of a single inverter, a quantitative assessment of the inverter's own operational stability is achieved. Furthermore, by collecting the same family of approximation degree sequences from other inverters in the same distribution area and performing collective discreteness analysis, a reference benchmark for the overall operational status of the distribution area is constructed. The organic combination of individual and collective discreteness analysis effectively distinguishes between grid faults and component faults. Based on this, a classification agent is constructed using historical fault data to classify the discrete parameters of individual approximation. Then, a discrete consistency coefficient is introduced to correct the classification results. Finally, power grid fault warning parameters and component fault warning parameters are output, providing maintenance personnel with clear and quantitative basis for judging the source of faults, which facilitates accurate dispatching and targeted maintenance, and effectively reduces the cost and time of fault investigation.

[0008] In summary, this invention solves the technical problem that existing technologies cannot effectively distinguish whether a fault in a residential photovoltaic inverter originates from the grid side or from the inverter's own components, leading to low operation and maintenance efficiency. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a flowchart illustrating a fault early warning method for a residential photovoltaic inverter provided in an embodiment of the present invention; Figure 2 This is a logical schematic diagram of a fault early warning method for a residential photovoltaic inverter provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a residential photovoltaic inverter fault early warning system provided in an embodiment of the present invention; The components represented by each number in the attached diagram are explained below: The module includes a data acquisition and feature extraction module 11, an approximation degree calculation module 12, a discreteness analysis module 13, and a fault early warning classification module 14. Detailed Implementation

[0011] This invention provides a fault early warning method and system for residential photovoltaic inverters, which specifically solves the technical problem that existing technologies cannot effectively distinguish whether a fault in a residential photovoltaic inverter originates from the grid side or from the inverter's own components, leading to low operation and maintenance efficiency.

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0013] Example 1, as Figure 1 and Figure 2 As shown, the present invention provides a fault early warning method for residential photovoltaic inverters, the method comprising: S100: Collect the historical electrical parameter set sequence of household photovoltaic in the past preset time range, extract the valley value and peak value respectively, and obtain the historical valley value electrical parameter sequence and the historical peak value electrical parameter sequence.

[0014] In existing fault early warning technologies for residential photovoltaic inverters, electrical parameters such as voltage are typically collected continuously at high frequencies of seconds or even milliseconds, resulting in an extremely large amount of data. If all the original sampling points are analyzed comprehensively, the computational burden is too heavy, making it difficult to meet the timeliness requirements of real-time early warning. On the other hand, if only the instantaneous voltage value at a single moment is taken for judgment, it is easily affected by random noise and instantaneous fluctuations, leading to a high false alarm rate.

[0015] Meanwhile, residential photovoltaic (PV) power generation exhibits a distinct diurnal cycle, with voltage varying regularly between day and night. Existing methods often fail to adequately consider this periodicity, making it difficult to effectively extract fault symptoms that accurately reflect the equipment's health status from long-term operational data. Furthermore, they cannot provide differentiated characteristic data for distinguishing between grid-side and component-side faults. Therefore, how to efficiently extract key fault features from massive amounts of high-frequency sampled historical voltage data in a manner consistent with the diurnal cycle characteristics of PV, significantly reducing data dimensionality while ensuring comprehensive information, and forming a differentiated data foundation for two-dimensional fault classification analysis, has become a pressing technical challenge.

[0016] Step S100 in the method provided in this embodiment of the invention includes: The electrical parameters of residential photovoltaic systems are collected over multiple unit time periods within a preset time range to obtain a sequence of historical electrical parameters, where each historical electrical parameter includes voltage. Extract the valley and peak values ​​from each historical electrical parameter set to obtain the historical valley electrical parameter sequence and the historical peak electrical parameter sequence.

[0017] In this embodiment of the invention, electrical parameters of the residential photovoltaic system are first collected over multiple unit time periods within a preset time range. The preset time range can be flexibly set according to actual analysis needs, for example, it can be set to the past week, the past half month, or the past month. In this embodiment of the invention, the example is set to the preset time range of the past week and the unit time length is set to one day, i.e., 24 hours.

[0018] Specifically, during normal operation, the built-in voltage sensor of a residential photovoltaic inverter continuously collects output voltage data at a high sampling frequency. Assuming a sampling frequency of one voltage value per second, 86,400 voltage data points can be collected in a day. These data points reflect the continuous changes in the inverter's output voltage throughout the entire power generation cycle from dawn to late at night, including voltage data during peak photovoltaic power generation periods during the day and voltage data during periods of no sunlight at night.

[0019] Furthermore, all voltage data points collected each day are combined to form a historical electrical parameter set, i.e., the voltage dataset corresponding to that day. In this embodiment, the past week contains seven calendar days, so seven historical electrical parameter sets can be collected, namely the dataset for day one, the dataset for day two, and so on up to the dataset for day seven. These seven datasets are arranged in chronological order to form a sequence of historical electrical parameter sets.

[0020] It should be noted that the sampling frequency and preset time range mentioned above are merely illustrative examples. In practical applications, adjustments can be made based on the device's data storage capacity, communication bandwidth, and the real-time requirements of fault warnings. For instance, when data storage and communication conditions are good, a higher sampling frequency can be used to obtain more detailed voltage change information; in scenarios where only daily trend analysis is required, the sampling frequency can be appropriately reduced to decrease data redundancy.

[0021] Furthermore, the collected voltage can be either the DC input voltage of the inverter or the AC output voltage of the inverter, whichever can be selected based on actual monitoring needs. In this embodiment of the invention, the AC output voltage of the inverter is collected because this voltage is directly related to the power quality at the grid connection point, and grid faults and component faults will both exhibit relatively obvious abnormal characteristics in the AC voltage.

[0022] Furthermore, after obtaining the historical electrical parameter set sequence, the valley and peak values ​​within each historical electrical parameter set are extracted and arranged in chronological order to obtain the historical valley electrical parameter sequence and the historical peak electrical parameter sequence. A valley value refers to the smallest voltage value among all voltage sampling data for that day contained in a historical electrical parameter set, i.e., the lowest voltage of that day. A peak value refers to the largest voltage value in the same historical electrical parameter set, i.e., the highest voltage of that day.

[0023] For example, on a certain natural day, a total of 86,400 voltage data points were collected. By comparing these voltage data points, it was found that the lowest voltage of the day occurred during the nighttime off-peak electricity consumption period, with a value of 218 volts; the highest voltage of the day occurred during the midday period when photovoltaic power generation was at its peak, with a value of 241 volts. Therefore, the valley value of that natural day was 218 volts, and the peak value was 241 volts.

[0024] It should be noted that before extracting valley and peak values, the collected daily voltage data can be preprocessed to remove obvious abnormal data points caused by occasional sensor malfunctions, communication interruptions, or strong external electromagnetic interference. For example, if a voltage value significantly lower than the normal range appears in the collected data for a particular day, this value may be a sampling error rather than the inverter's actual output voltage. Using this as a valley value would severely interfere with the accuracy of subsequent analysis results. Proper preprocessing can effectively improve the accuracy and reliability of valley and peak value extraction.

[0025] This invention, by using a "day" as the unit time window, compresses voltage sampling data from multiple consecutive days into daily valley and daily peak sequences, significantly reducing the data volume and the processing burden of subsequent calculations, thus improving the computing speed of the early warning system. Simultaneously, the daily valley and peak values ​​represent the lowest and highest voltages of the day, effectively preserving information on extreme voltage fluctuations and exhibiting high sensitivity to early fault signs. Furthermore, using a natural day as the statistical period fully aligns with the daily cycle of photovoltaic power generation, naturally separating normal daily variations from abnormal fluctuations, thereby reducing false alarms. In addition, the independent extraction of valley and peak sequences provides a differentiated data foundation for subsequent approximation analysis from the two dimensions of "low voltage" and "high voltage," effectively distinguishing between grid faults and component faults.

[0026] S200: Calculate the approximation degree between the historical valley electrical parameter sequence and the first fault electrical threshold, and between the historical peak electrical parameter sequence and the second fault electrical threshold, respectively, to obtain the first approximation degree sequence and the second approximation degree sequence.

[0027] In existing fault early warning technologies for residential photovoltaic inverters, the extracted valley and peak value sequences often exist only as raw voltage values, lacking a unified quantitative indicator to describe the degree to which the current operating voltage deviates from the normal state. This makes it impossible for analysts to intuitively judge the safety margin of the inverter from a true fault, and the early warning lacks clear quantitative basis. At the same time, existing methods often separate valley and peak value analysis, making it difficult to comprehensively assess and compare the risks of low voltage and high voltage.

[0028] Furthermore, because the rated voltage parameters and fault thresholds of different inverters vary, the raw valley and peak values ​​of different devices lack horizontal comparability, making it difficult to analyze the collective discreteness of multiple inverters within a distribution area. Therefore, how to construct a unified approximation quantification index system, transforming the valley and peak value sequences into approximation sequences with their respective fault thresholds, and providing standardized fault risk quantification data for each inverter, has become an urgent technical problem to be solved.

[0029] Step S200 in the method provided in this embodiment of the invention includes: Obtain the first fault electrical threshold and the second fault electrical threshold for a residential photovoltaic inverter. Calculate the similarity between each valley electrical parameter in the historical valley electrical parameter sequence and the first fault electrical threshold to obtain the first approximation sequence; The similarity between each peak electrical parameter in the historical peak electrical parameter sequence and the second fault electrical threshold is calculated to obtain the second approximation sequence.

[0030] In this embodiment of the invention, a first fault electrical threshold and a second fault electrical threshold are first obtained for a residential photovoltaic inverter. The first fault electrical threshold is a benchmark value used to determine the risk of undervoltage. When the actual operating voltage of the inverter consistently approaches or falls below this threshold, it indicates that the inverter is at risk of failure due to a drop in grid voltage or abnormal internal components. The second fault electrical threshold is a benchmark value used to determine the risk of overvoltage. When the actual operating voltage of the inverter consistently approaches or exceeds this threshold, it indicates that the inverter is at risk of failure due to a rise in grid voltage or aging of internal components.

[0031] The first and second electrical fault thresholds can be obtained by consulting the technical specifications provided with the inverter equipment at the time of manufacture. Inverter manufacturers typically indicate the normal operating voltage range of the equipment in the product manual. The lower limit of this range can be used as the first electrical fault threshold, and the upper limit as the second electrical fault threshold. For example, if a certain model of residential photovoltaic inverter has a rated operating voltage range of 220 volts with a permissible deviation of ±15%, then its normal operating voltage lower limit is 187 volts and the upper limit is 253 volts. When the actual operating voltage is lower than 187 volts or higher than 253 volts, the equipment is at risk of failure. Therefore, the first electrical fault threshold can be set at 187 volts, and the second electrical fault threshold at 253 volts.

[0032] Furthermore, after obtaining the first and second fault electrical thresholds, the similarity between each valley electrical parameter in the historical valley electrical parameter sequence and the first fault electrical threshold is first calculated to obtain the first approximation sequence. The similarity between each valley electrical parameter and the first fault electrical threshold is used to measure how close each valley electrical parameter is to the first fault electrical threshold. The higher the similarity, the closer the actual minimum voltage of that day is to the fault warning line, and the greater the risk of a low voltage fault.

[0033] In this embodiment of the invention, the similarity between each valley electrical parameter and the first fault electrical threshold is calculated as 1 - |(a valley electrical parameter - the first fault electrical threshold) / the first fault electrical threshold|. For example, the first fault electrical threshold is set to 187 volts, and the historical valley electrical parameter sequence for the past seven days is 218 volts, 217 volts, 219 volts, 216 volts, 214 volts, 213 volts, and 211 volts. Following the similarity calculation method described above, and sorting the results chronologically, the first approximation sequence is obtained as follows: 0.834, 0.840, 0.829, 0.845, 0.856, 0.861, and 0.872. It can be clearly seen from this first approximation sequence that as time progresses from day one to day seven, the approximation value gradually increases, rising from 0.834 to 0.872, meaning that the daily valley electrical parameter is gradually approaching the first fault electrical threshold of 187 volts. This means that the inverter's minimum daily voltage is gradually decreasing, getting closer and closer to the fault warning line, and the risk of equipment failures related to low voltage is gradually increasing.

[0034] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0035] Further, the similarity between each peak electrical parameter in the historical peak electrical parameter sequence and the second fault electrical threshold is calculated to obtain the second approximation sequence. Similarly, the similarity between each peak electrical parameter and the second fault electrical threshold is calculated as 1 - |(a peak electrical parameter - the second fault electrical threshold) / the second fault electrical threshold|. For example, the second fault electrical threshold is set to 253 volts, and the historical peak electrical parameter sequence for the past seven days is 241 volts, 242 volts, 240 volts, 243 volts, 245 volts, 246 volts, and 248 volts. Following the similarity calculation method described above and sorting the results chronologically, the second approximation sequence is obtained as follows: 0.953, 0.957, 0.949, 0.960, 0.968, 0.972, and 0.980. The second approximation sequence clearly shows that as time progresses from day one to day seven, the approximation value gradually increases, rising from 0.953 to 0.980. This means that the daily peak electrical parameters are gradually approaching the second fault electrical threshold of 253 volts. This implies that the inverter's daily maximum voltage is gradually increasing, getting closer and closer to the fault warning line, and the risk of overvoltage-related faults in the equipment is increasing accordingly.

[0036] It is important to note that if the peak electrical parameter of a certain day exceeds the second fault electrical threshold, it indicates that the inverter's highest voltage on that day has exceeded the fault warning line, and the equipment is already in an overvoltage fault state. In this case, the approximation degree for that day should be directly assigned a value of one, signifying that the approximation degree has reached its maximum value, and should no longer be calculated according to the aforementioned difference formula. Similarly, if the valley electrical parameter of a certain day is less than the first fault electrical threshold, the approximation degree for that day should also be directly assigned a value of one.

[0037] Combining the trends of the first and second approximation sequences in the two examples above, it can be concluded that over a seven-day period, the inverter's daily minimum voltage continuously decreased while its daily maximum voltage continuously increased, indicating a continuously expanding voltage fluctuation range. This bidirectional approximation trend of decreasing troughs and increasing peaks is a significant indicator of a serious potential fault in the inverter.

[0038] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation of the present invention. Furthermore, the method for calculating the approximation degree described above is merely a preferred example and does not constitute a limitation of the present invention.

[0039] This invention, through obtaining a first fault electrical threshold and a second fault electrical threshold, transforms the valley value sequence and peak value sequence into a unified approximation degree sequence, thereby quantifying the safety margin of the inverter from fault. The closer the approximation degree is to a value of 1, the higher the risk, providing a clear numerical basis for early warning decisions. Simultaneously, the valley value approximation degree and peak value approximation degree are obtained within the same framework, facilitating a comprehensive assessment of both low and high voltage risks. Furthermore, the approximation degree conversion eliminates dimensional differences between different devices, enabling horizontal comparability of monitoring data from multiple inverters in a distribution area, laying a data foundation for subsequent collective dispersion analysis.

[0040] S300: Perform individual discreteness analysis on the first approximation sequence and the second approximation sequence respectively, and perform collective discreteness analysis on the monitoring data of the same station area to obtain individual approximation discreteness parameters and station area approximation discreteness parameters.

[0041] In existing fault early warning technologies for residential photovoltaic inverters, the calculation results of the approximation degree sequence are often used only as static indicators, lacking analysis of the time-varying patterns of the approximation degree sequence, making it difficult to capture the dynamic characteristics of the gradual evolution of faults. Furthermore, existing methods only focus on the operating status of a single inverter, neglecting the synchronous operating data of other inverters within the same grid area. This results in the inability to effectively distinguish systemic voltage anomalies from faults in the inverter's own components.

[0042] Furthermore, the discrete characteristics of the valley and peak approximation sequences were not extracted and comprehensively utilized separately, making it difficult to quantitatively assess the fluctuations of low and high voltage risks. Therefore, constructing a complete analytical framework that moves from individual to collective discreteness analysis—quantifying and comprehensively assessing the fluctuations of the valley and peak approximation sequences for a single inverter, while also introducing group data from multiple inverters in the same area as a reference—to effectively distinguish between grid faults and component faults through comparison of individual and collective discreteness, has become a pressing technical problem to be solved.

[0043] Step S300 in the method provided in this embodiment of the invention includes: Individual discreteness analysis was performed on the first approximation sequence and the second approximation sequence respectively to obtain the individual approximation discreteness parameters; Collect multiple sets of first approximation sequences and multiple sets of second approximation sequences of other household photovoltaic inverters within the same distribution area where the household photovoltaic system is located; Based on the multiple sets of first approximation degree sequences and multiple sets of second approximation degree sequences of the same family, a collective discreteness analysis is performed to obtain the discretization parameters of the transformer area approximation degree.

[0044] In this embodiment of the invention, individual discreteness analysis is first performed on the first approximation sequence and the second approximation sequence. Individual discreteness analysis refers to analyzing the fluctuation degree of the approximation sequence based on the historical data of a single inverter. The greater the discreteness, the more unstable the operating state of the inverter and the higher the risk of failure.

[0045] Specifically, the first approximation sequence is subjected to discrete parameter calculation to obtain the first volume approximation discrete parameter. The discrete parameter can be represented by standard deviation, variance, or mean absolute deviation, etc. In this embodiment of the invention, variance is used as an example. For example, the obtained first approximation sequence is 0.834, 0.840, 0.829, 0.845, 0.856, 0.861, and 0.872. First, the average value of the first approximation sequence is calculated, and the first average approximation is approximately 0.848. Then, the discrete parameter of the first approximation sequence is calculated by calculating the difference between each value of the first approximation sequence and the average value of 0.848, squaring each difference, summing them, and then dividing by the number of data points to obtain the variance, which is the first volume approximation discrete parameter. After calculation, the discrete parameter of the first approximation sequence is approximately 0.000225, which reflects the fluctuation range of the valley approximation sequence around its average value. In this embodiment of the invention, the difference between each value and the average value is calculated and rounded to three decimal places.

[0046] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention. In the embodiments of the present invention, the difference between each value of the first approximation sequence and the average value is rounded to three decimal places.

[0047] Specifically, the second approximation sequence is subjected to discrete parameter calculation to obtain the second volume approximation discrete parameter. For example, the obtained second approximation sequence is: 0.953, 0.957, 0.949, 0.960, 0.968, 0.972, 0.980. First, the mean of this second approximation sequence is calculated, and then its variance is calculated, resulting in a variance of approximately 0.000121, which is the second volume approximation discrete parameter. It reflects the fluctuation range of the peak approximation sequence around its mean.

[0048] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention. In the embodiments of the present invention, the difference between each value of the second approximation sequence and the average value is rounded to three decimal places.

[0049] Furthermore, multiple sets of first-degree-of-approximation sequences and multiple sets of second-degree-of-approximation sequences from other residential PV inverters within the same distribution transformer area were collected. A distribution transformer area refers to the region supplied by the same distribution transformer. Within the same distribution transformer area, all residential PV inverters are connected to the same grid terminal and experience essentially the same grid voltage background. Therefore, when a grid-side fault occurs, the inverters within the distribution transformer area will exhibit similar systemic anomalies; while when only a single inverter exhibits anomalies, it is more likely due to a component failure within that inverter itself.

[0050] For example, assume that in addition to the target inverter, there are four other inverters in the distribution area, numbered Inverter A, Inverter B, Inverter C, and Inverter D respectively. For each inverter, calculate its own first approximation sequence and second approximation sequence, thereby obtaining four sets of first approximation sequences and four sets of second approximation sequences of the same family.

[0051] Furthermore, collective discreteness analysis is performed based on multiple sets of first-degree approximation sequences and multiple sets of second-degree approximation sequences from the same family to obtain the distribution area approximation discrete parameters. Collective discreteness analysis refers to analyzing the degree of dispersion of the approximation data of multiple inverters within a distribution area at the collective level. If the fluctuation of a certain inverter is large but the overall fluctuation of the distribution area is small, it indicates that the anomaly of that inverter is an isolated event, and is more likely a component failure; if the overall fluctuation of the distribution area is also large, it indicates that the anomaly may originate from the grid side.

[0052] Individual discreteness analysis was performed on the first and second approximation sequences respectively to obtain individual approximation discreteness parameters, including: The first volume approximation degree discrete parameters are obtained by performing discrete parameter calculation on the first approximation degree sequence. The second volume approximation degree discrete parameters are obtained by performing discrete parameter calculation on the second approximation degree sequence. Calculate the mean of the first approximation sequence and the second approximation sequence to obtain the first average approximation and the second average approximation, and assign the first approximation weight and the second approximation weight. The first approximation weight and the second approximation weight are used to calculate the first individual approximation degree discrete parameter and the second individual approximation degree discrete parameter by weighting.

[0053] In this embodiment of the invention, the first volume approximation discrete parameter is obtained by performing discrete parameter calculation on the first approximation sequence; the second volume approximation discrete parameter is obtained by performing discrete parameter calculation on the second approximation sequence.

[0054] For example, the obtained first approximation sequence is 0.834, 0.840, 0.829, 0.845, 0.856, 0.861, and 0.872. First, the average value of this first approximation sequence is calculated, resulting in a first average approximation of approximately 0.848. Then, the discrete parameter of this first approximation sequence is calculated by taking the difference between each value and the average value of 0.848, squared each difference, summing the results, and then dividing by the number of data points. The variance obtained is the first volume approximation discrete parameter. After calculation, the discrete parameter of this first volume approximation sequence is approximately 0.000225.

[0055] The obtained second approximation sequence is: 0.953, 0.957, 0.949, 0.960, 0.968, 0.972, 0.980. First, the mean of this second approximation sequence is calculated, and then the variance of this second approximation sequence is calculated. The variance is approximately 0.000121, which is the discrete parameter of the second volume approximation.

[0056] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0057] Furthermore, the mean values ​​of the first approximation sequence and the second approximation sequence are calculated respectively to obtain the first average approximation degree and the second average approximation degree, and the first approximation weight and the second approximation weight are assigned. The higher the first average approximation degree, the closer the daily minimum voltage of the inverter is to the fault threshold, and the higher the risk of excessively low voltage; the higher the second average approximation degree, the closer the daily maximum voltage of the inverter is to the fault threshold, and the higher the risk of excessively high voltage.

[0058] The first and second approximation weights are dynamically determined based on the actual mean values ​​of the first and second approximation degree sequences. Specifically, the first approximation weight equals the first average approximation degree divided by the sum of the first and second average approximation degrees; the second approximation weight equals the second average approximation degree divided by the sum of the first and second average approximation degrees. The calculated first and second approximation weights thus satisfy the condition that their sum equals one.

[0059] For example, if the obtained first approximation sequence is 0.834, 0.840, 0.829, 0.845, 0.856, 0.861, 0.872, then the first average approximation is 0.848; if the obtained second approximation sequence is 0.953, 0.957, 0.949, 0.960, 0.968, 0.972, 0.980, then the second average approximation is 0.963. Therefore, the first approximation weight = 0.848 / (0.848 + 0.963) ≈ 0.468, and the second approximation weight = 0.963 / (0.848 + 0.963) ≈ 0.532.

[0060] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0061] Furthermore, using a first approximation weight and a second approximation weight, the first volume approximation degree discrete parameter and the second individual approximation degree discrete parameter are weighted and calculated to obtain the individual approximation degree discrete parameter. The individual approximation degree discrete parameter comprehensively reflects the overall dispersion of the inverter's valley approximation degree and peak approximation degree sequences, providing a comprehensive quantitative assessment of the inverter's operational stability. For example, if the discrete parameter of the first volume approximation degree sequence for a given inverter is 0.000225 and the discrete parameter of the second volume approximation degree sequence is 0.000121; and the first approximation weight is set to 0.468 and the second approximation weight to 0.532, then the individual approximation degree discrete parameter of the inverter = 0.000225 × 0.468 + 0.000121 × 0.532 ≈ 0.000170.

[0062] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0063] Based on the multiple sets of first approximation degree sequences and multiple sets of second approximation degree sequences of the same family, a collective discreteness analysis is performed to obtain the discrete parameters of the transformer area approximation degree, including: Discrete parameters are calculated for multiple sets of first-degree-of-approximation sequences and multiple sets of second-degree-of-approximation sequences of the same family, respectively, to obtain multiple first-degree-of-approximation discrete parameters and multiple second-degree-of-approximation discrete parameters; Calculate the mean of multiple sets of first approximation degree sequences and multiple sets of second approximation degree sequences of the same family respectively, to obtain multiple first average approximation degree sequences and multiple second average approximation degree sequences of the same family, and assign multiple first approximation degree sequences and multiple second approximation degree sequences of the same family. Multiple first-family approximation weights and multiple second-family approximation weights are used to perform weighted calculations on multiple first-family approximation degree discrete parameters and multiple second-family approximation degree discrete parameters to obtain multiple weighted collective approximation degree discrete parameters, and the station area approximation degree discrete parameters are calculated.

[0064] In this embodiment of the invention, discrete parameters are first calculated for multiple sets of first-degree-of-approximation sequences and multiple sets of second-degree-of-approximation sequences of the same family, respectively, to obtain multiple first-degree-of-approximation discrete parameters and multiple second-degree-of-approximation discrete parameters of the same family.

[0065] For example, in addition to the target inverter, four other inverters within the transformer area were selected as family references, numbered Inverter A, Inverter B, Inverter C, and Inverter D. First, the mean values ​​of the first approximation sequence of the four family inverters were obtained as 0.846, 0.849, 0.847, and 0.852, respectively. Then, the discrete parameters of the first family approximation were obtained as 0.000256, 0.000225, 0.000289, and 0.000196, respectively. The mean values ​​of the second approximation sequence of the four family inverters were obtained as 0.960, 0.961, 0.962, and 0.961, respectively. Then, the discrete parameters of the second family approximation were obtained as 0.000121, 0.000144, 0.000121, and 0.000121, respectively.

[0066] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0067] Furthermore, the average values ​​of multiple sets of first-degree approximation sequences and multiple sets of second-degree approximation sequences within the same family are calculated respectively to obtain multiple first-average approximation degrees and multiple second-average approximation degrees within the same family. Multiple first-family approximation weights and multiple second-family approximation weights are then assigned. For example, in addition to the target inverter, four other inverters within the transformer area are selected as family references, numbered Inverter A, Inverter B, Inverter C, and Inverter D respectively. The mean values ​​of the first approximation sequences of the four inverters in the same family are 0.846, 0.849, 0.847, and 0.852, respectively. The mean values ​​of the second approximation sequences of the four inverters in the same family are 0.960, 0.961, 0.962, and 0.961, respectively. The first family approximation weight of inverter A is 0.846 / (0.846+0.960)≈0.468, and the second family approximation weight of inverter A is 0.960 / (0.846+0.960)≈0.532. Similarly, the first family approximation weights of the other three inverters are calculated to be 0.469, 0.468, and 0.470, respectively, and the second family approximation weights of each inverter are 0.531, 0.532, and 0.530, respectively.

[0068] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0069] Finally, multiple first-family approximation weights and multiple second-family approximation weights are used to weight and calculate the discrete parameters of the first-family approximation degree and the discrete parameters of the second-family approximation degree, resulting in multiple weighted collective approximation degree discrete parameters. The transformer area approximation degree discrete parameters are then calculated using an arithmetic mean, meaning the transformer area approximation degree discrete parameter equals the sum of the multiple weighted collective approximation degree discrete parameters within the transformer area divided by the number of inverters in the same family.

[0070] This invention first calculates and weights the discrete parameters of the valley and peak approximation sequences of a single inverter, thereby achieving a quantitative assessment of the fluctuation amplitude of the inverter's own approximation sequence. A larger discrete parameter indicates a more unstable operating state, effectively revealing the dynamic evolution trend of faults. Simultaneously, by collecting the same family of approximation sequences from other inverters in the same distribution area and performing collective discreteness analysis, a reference benchmark for the overall operating state of the distribution area is constructed. When the individual discrete parameter deviates significantly from the collective discrete parameter of the distribution area, it can effectively identify an abnormal state of the inverter independent of the grid background. Furthermore, this scheme calculates and weights the valley-related discrete parameters and peak-related discrete parameters separately, fully considering the fluctuation information from both undervoltage and overvoltage risks, making fault warnings more comprehensive and accurate. The organic combination of individual and collective discrete analysis provides a reliable quantitative basis for subsequently distinguishing between grid fault warning parameters and component fault warning parameters.

[0071] S400: Based on the individual approximation degree discrete parameters and the transformer area approximation degree discrete parameters, perform fault warning classification to obtain power grid fault warning parameters and component fault warning parameters.

[0072] In existing fault early warning technologies for residential photovoltaic inverters, the individual approximation degree discrete parameters and the transformer area approximation degree discrete parameters obtained after discreteness analysis are often used only as independent statistical indicators. There is a lack of a complete decision-making mechanism to transform these parameters into fault category early warning results that can directly guide operation and maintenance. While existing methods can identify inverter anomalies, they cannot further determine whether the anomaly originates from the grid side or the inverter's internal components. This results in maintenance personnel still needing to conduct extensive on-site investigations after receiving early warnings, hindering accurate work order dispatch and targeted repairs.

[0073] Meanwhile, existing methods often rely solely on individual inverter data or simple threshold comparisons when determining fault categories. They fail to build classification models based on historical fault data or incorporate the collective dispersion of the transformer area as a reference benchmark to correct the classification results, resulting in insufficient accuracy and robustness in fault category determination. Therefore, how to construct a complete transformation mechanism from discrete parameters to fault warning parameters, using historical fault data to train a classification model for initial category determination, then introducing the similarity between individual discrete parameters and transformer area discrete parameters to adaptively correct the classification results, and finally outputting grid fault warning parameters and component fault warning parameters to achieve accurate fault source location and quantitative early warning, has become a pressing technical problem to be solved.

[0074] Step S400 in the method provided in this embodiment of the invention includes: Based on the first approximation sequence and the second approximation sequence, the mean value is calculated to obtain the first average approximation and the second average approximation, and inverter fault approximation calculation is performed to obtain the basic fault risk rate. Based on the individual approximation degree discrete parameters, fault categories are classified to obtain individual power grid category rate and individual component category rate; Calculate the similarity between the individual approximation degree discrete parameters and the transformer area approximation degree discrete parameters, configure the grid category rate correction coefficient, and perform correction calculations on the individual grid category rate and the individual component category rate to obtain the corrected individual grid category rate and the corrected individual component category rate; Based on the basic fault risk rate, and combined with the modified individual grid category rate and the modified individual component category rate, the grid fault early warning parameters and component fault early warning parameters are calculated.

[0075] In this embodiment of the invention, the mean values ​​of the first and second approximation sequences are first calculated to obtain the first average approximation and the second average approximation. Then, inverter fault approximation calculations are performed to obtain the basic fault risk rate. The basic fault risk rate is used to quantitatively assess the degree to which the inverter as a whole approaches the fault threshold. Since a higher mean of the first approximation indicates a higher risk of undervoltage, and a higher mean of the second approximation indicates a higher risk of overvoltage, the basic fault risk rate should comprehensively reflect both aspects of risk.

[0076] In this embodiment of the invention, the basic fault risk rate is calculated as the average of the first average approximation and the second average approximation. The basic fault risk rate reflects the overall level of the inverter's comprehensive approach to the fault threshold. The closer the value is to one, the closer the overall operating state of the inverter is to the fault boundary, and the higher the risk.

[0077] Furthermore, based on the individual approximation degree discrete parameters, fault category classification is performed to obtain the individual grid category rate and the individual component category rate, including: Based on the fault monitoring data of residential photovoltaic inverters, a set of discrete parameters for the approximation degree of individual samples is collected, and the inverter fault category corresponding to the discrete parameter for the approximation degree of each individual sample is obtained, resulting in a set of fault categories for sample inverters. Each fault category for a sample inverter includes either a grid category or a component category. The set of discrete parameters for the approximation degree of individual samples and the set of fault categories of sample inverters are split into K-fold partitions to obtain K sets of sample category classification data, and K fault category classification agents are constructed. The discrete parameters of the individual approximation degree are input into K fault category classification agents, and K inverter fault categories are output. The proportions of grid category and component category are calculated respectively to obtain the individual grid category rate and the individual component category rate.

[0078] In this embodiment of the invention, firstly, based on the historical fault monitoring data of the residential photovoltaic inverter, a set of discrete parameters for individual sample approximation is collected, and the inverter fault category corresponding to each sample is obtained.

[0079] Specifically, a large amount of sample data of known fault categories is collected from the historical operation records of the transformer substation or inverters of the same model. For each historical fault sample, its individual approximation discrete parameter is calculated. At the same time, the actual fault category of the sample is recorded. The fault categories are divided into two categories: grid category, that is, the fault is caused by grid-side anomalies, such as grid voltage drop, grid frequency fluctuation, etc.; and component category, that is, the fault is caused by internal component anomalies of the inverter, such as capacitor aging, power switch performance degradation, etc.

[0080] Assume that a total of one thousand historical fault samples are collected. Each sample contains two pieces of information: the individual approximation degree discrete parameter value and the corresponding fault category label. These one thousand sample data constitute the sample individual approximation degree discrete parameter set and the sample inverter fault category set.

[0081] Furthermore, the set of discrete parameters for the approximation degree of individual samples and the set of fault categories of sample inverters are split into K-fold partitions to obtain K sets of sample category classification data, and K fault category classification agents are constructed.

[0082] K-fold partitioning is a commonly used cross-validation method. Specifically, all sample data is randomly divided into K equal parts, where K is a positive integer greater than or equal to two, and each part contains approximately the same number of samples. In this embodiment of the invention, K is preferably set to five, that is, one thousand samples are randomly divided into five equal parts, each containing two hundred samples.

[0083] Then, using these K datasets, K fault category classification agents are constructed. Each classification agent is constructed as follows: K minus one dataset is selected as the training set, and the remaining dataset is used as the validation set to train a classification model. Specifically: The first classification agent: trained using the second to fifth sets of data as the training set and the first set of data as the validation set; The second classification agent is trained using the first, third, fourth, and fifth sets of data as the training set and the second set of data as the validation set. In this way, a total of five classification agents are constructed.

[0084] The classification agent can be implemented using various machine learning classification algorithms, such as the K-nearest neighbors algorithm, support vector machines, decision trees, random forests, or neural networks. In this embodiment of the invention, the K-nearest neighbors algorithm is used because it is computationally simple, highly interpretable, and suitable for deployment on embedded or edge computing devices. The input to the classification agent is the discrete parameter of the individual approximation degree, and the output is the fault category.

[0085] By constructing the K-fold split and K classification agents described above, the overfitting problem caused by training data bias in a single classifier can be effectively avoided, thereby improving the robustness and reliability of the classification results.

[0086] Furthermore, the discrete parameter of the individual approximation degree of the target inverter is input into K fault category classification agents to obtain K inverter fault category output results. Then, the proportions of grid category and component category are calculated respectively. For example, the discrete parameter of the individual approximation degree of the target inverter is approximately 0.00183. This value is input into five classification agents, assuming the fault category output results of the five agents are as follows: First classification agent output: component category; Second classification agent output: grid category; Third classification agent output: component category; Fourth classification agent output: component category; Fifth classification agent output: component category.

[0087] Statistically analyzing the five output results above, the grid category appears once, and the component category appears four times. Therefore, the individual grid category rate is the number of times the grid category appears divided by the total number of occurrences, i.e., 1 divided by 5, which equals 0.20; the individual component category rate is the number of times the component category appears divided by the total number of occurrences, i.e., four divided by five, which equals 0.80. This result indicates that, based on the classification judgment of the individual approximation degree discrete parameter, this inverter has a 20% probability of belonging to the grid category fault and an 80% probability of belonging to the component category fault.

[0088] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0089] Calculate the similarity between the individual approximation degree discrete parameters and the transformer area approximation degree discrete parameters, configure the grid category rate correction coefficient, and perform correction calculations on the individual grid category rate and the individual component category rate to obtain the corrected individual grid category rate and the corrected individual component category rate, including: The similarity between the discrete parameters of the individual approximation degree and the discrete parameters of the station area approximation degree is calculated to obtain the discrete consistency coefficient; The ratio of the discrete consistency coefficient to the average discrete consistency coefficient over a historical period is calculated to obtain the power grid category rate correction coefficient. The individual power grid category rate is corrected by using the power grid category rate correction coefficient to obtain the corrected individual power grid category rate, and the corrected individual component category rate is also calculated.

[0090] In this embodiment of the invention, the similarity between the individual approximation degree discrete parameters and the distribution area approximation degree discrete parameters is calculated to obtain the discrete consistency coefficient. The discrete consistency coefficient is used to measure the consistency between the individual discreteness of the target inverter and the overall discreteness of the distribution area. When the individual approximation degree discrete parameters are closer to the distribution area approximation degree discrete parameters, it indicates that the fluctuation characteristics of the inverter are more consistent with the overall trend of the distribution area, and its anomalies are more likely to be caused by systemic factors on the grid side; when the difference between the two is greater, it indicates that the fluctuation characteristics of the inverter deviate significantly from the overall trend of the distribution area, and its anomalies are more likely to be caused by component failures of the inverter itself. In this embodiment of the invention, the discrete consistency coefficient is equal to the smaller value of the individual approximation degree discrete parameters and the distribution area approximation degree discrete parameters divided by the larger value.

[0091] For example, if the individual approximation degree discrete parameter is approximately 0.000183 and the distribution area approximation degree discrete parameter is approximately 0.000181, then the discrete consistency coefficient = 0.000181 / 0.000183 ≈ 0.989. A discrete consistency coefficient close to one indicates that the individual discreteness of the target inverter is highly consistent with the overall discreteness of the distribution area, and the fluctuation characteristics of the inverter basically match the overall trend of the distribution area.

[0092] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0093] Furthermore, the ratio of the discrete consistency coefficient to the average discrete consistency coefficient over historical time is calculated to obtain the power grid category rate correction coefficient.

[0094] The historical average discrete consistency coefficient refers to the average discrete consistency coefficient of a transformer substation under normal operating conditions over a past period. This historical average reflects the individual-collective consistency level of the substation under normal conditions.

[0095] For example, assuming that historical data shows the average discrete consistency coefficient for this distribution area under normal operating conditions was approximately 0.950, and the currently calculated discrete consistency coefficient is approximately 0.989, the grid category rate correction factor equals the current discrete consistency coefficient divided by the historical average discrete consistency coefficient, i.e., grid category rate correction factor = 0.989 / 0.950 = 1.041. This correction factor being greater than one indicates that the current individual-collective consistency is higher than the historical normal level, meaning the inverter's fluctuation characteristics are more consistent with the overall trend of the distribution area, and the probability of grid category faults is relatively increased.

[0096] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0097] Finally, the grid category rate correction coefficient is used to correct the individual grid category rate, resulting in the corrected individual grid category rate, and then the corrected individual component category rate is calculated. The individual grid category rate refers to the probability or proportion of an inverter being classified as a grid category fault after classification based on the individual approximation discrete parameters of the target inverter; the individual component category rate refers to the probability or proportion of an inverter being classified as a component category fault after classification based on the individual approximation discrete parameters of the target inverter. The corrected individual grid category rate equals the individual grid category rate multiplied by the grid category rate correction coefficient. Since the sum of the individual grid category rate and the individual component category rate equals one, the sum of the corrected individual grid category rate and the corrected individual component category rate should also equal one. Therefore, the corrected individual component category rate equals one minus the corrected individual grid category rate.

[0098] Furthermore, based on the basic fault risk rate, combined with the modified individual grid category rate and the modified individual component category rate, the grid fault early warning parameters and component fault early warning parameters are calculated.

[0099] Grid fault early warning parameters are used to quantitatively assess the probability or risk level of a grid-side fault occurring in the inverter. Component fault early warning parameters are used to quantitatively assess the probability or risk level of a component-side fault occurring in the inverter.

[0100] For example, if the individual grid category rate is set to 0.20 and the grid category rate correction factor is approximately 1.041, then the corrected individual grid category rate = 0.20 × 1.041 = 0.208, and the corrected individual component category rate = 1 - 0.208 = 0.792.

[0101] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0102] In this embodiment of the invention, the power grid fault early warning parameter is calculated as follows: the power grid fault early warning parameter equals the basic fault risk rate multiplied by the modified individual power grid category rate. The component fault early warning parameter is calculated as follows: the component fault early warning parameter equals the basic fault risk rate multiplied by the modified individual component category rate.

[0103] This invention first calculates the basic fault risk rate based on the first and second average approximation degrees, thus achieving a quantitative assessment of the overall inverter's proximity to the fault threshold. Based on this, K classification agents are constructed using historical fault data to classify individual approximation degree discrete parameters, obtaining individual grid category rates and individual component category rates, achieving data-driven judgment of fault categories. Simultaneously, the classification results are corrected by calculating the discrete consistency coefficient and its ratio to the historical average, incorporating the collective dispersion of the distribution area as a reference benchmark into the judgment process, significantly improving the accuracy of fault category differentiation. Finally, grid fault warning parameters and component fault warning parameters are output, providing maintenance personnel with clear and quantitative criteria for judging fault sources, facilitating accurate dispatching and targeted maintenance.

[0104] In summary, the embodiments of the present invention have at least the following technical effects: This invention extracts valley and peak values ​​within a unit time window, compressing massive amounts of high-frequency sampling data into ordered feature sequences, significantly reducing the data processing burden. Simultaneously, using a natural day as the statistical period fully aligns with the daily cycle of photovoltaic power generation, naturally separating normal daily variations from abnormal fluctuations, thus reducing false alarms. By acquiring fault electrical thresholds and calculating the approximation degree between the valley and peak value sequences and their respective thresholds, the original voltage data is uniformly transformed into an approximation degree sequence. This quantifies the safety margin of the inverter from faults, providing a clear numerical basis for early warning decisions. It also eliminates dimensional differences between different devices, making the monitoring data of multiple inverters in a distribution area comparable horizontally. By calculating the variance and weighted summing of the valley and peak value approximation degree sequences of a single inverter, a quantitative assessment of the inverter's own operational stability is achieved. Furthermore, by collecting the same family of approximation degree sequences from other inverters in the same distribution area and performing collective discreteness analysis, a reference benchmark for the overall operational status of the distribution area is constructed. The organic combination of individual and collective discreteness analysis effectively distinguishes between grid faults and component faults. Based on this, a classification agent is constructed using historical fault data to classify the discrete parameters of individual approximation. Then, a discrete consistency coefficient is introduced to correct the classification results. Finally, power grid fault warning parameters and component fault warning parameters are output, providing maintenance personnel with clear and quantitative basis for judging the source of faults, which facilitates accurate dispatching and targeted maintenance, and effectively reduces the cost and time of fault investigation.

[0105] In summary, this invention solves the technical problem that existing technologies cannot effectively distinguish whether a fault in a residential photovoltaic inverter originates from the grid side or from the inverter's own components, leading to low operation and maintenance efficiency.

[0106] Example 2, as Figure 3As shown, based on the same inventive concept as the fault early warning method for residential photovoltaic inverters provided in Embodiment 1, this embodiment of the invention also provides a fault early warning system for residential photovoltaic inverters, including: The data acquisition and feature extraction module 11 is used to collect the historical electrical parameter set sequence of household photovoltaic in the past preset time range, extract the valley value and the peak value respectively, and obtain the historical valley value electrical parameter sequence and the historical peak value electrical parameter sequence; The approximation calculation module 12 is used to calculate the approximation degree between the historical valley electrical parameter sequence and the first fault electrical threshold, and between the historical peak electrical parameter sequence and the second fault electrical threshold, respectively, to obtain the first approximation degree sequence and the second approximation degree sequence; The discreteness analysis module 13 is used to perform individual discreteness analysis on the first approximation sequence and the second approximation sequence, and to perform collective discreteness analysis on the monitoring data of the same station area, so as to obtain individual approximation discreteness parameters and station area approximation discreteness parameters. The fault warning classification module 14 is used to classify fault warnings based on the individual proximity degree discrete parameters and the distribution area proximity degree discrete parameters, and obtain power grid fault warning parameters and component fault warning parameters.

[0107] In one embodiment, the data acquisition and feature extraction module 11 is specifically used for: The electrical parameters of residential photovoltaic systems are collected over multiple unit time periods within a preset time range to obtain a sequence of historical electrical parameters, where each historical electrical parameter includes voltage. Extract the valley and peak values ​​from each historical electrical parameter set to obtain the historical valley electrical parameter sequence and the historical peak electrical parameter sequence.

[0108] In one embodiment, the approximation calculation module 12 is specifically used for: Obtain the first fault electrical threshold and the second fault electrical threshold for a residential photovoltaic inverter. Calculate the similarity between each valley electrical parameter in the historical valley electrical parameter sequence and the first fault electrical threshold to obtain the first approximation sequence; The similarity between each peak electrical parameter in the historical peak electrical parameter sequence and the second fault electrical threshold is calculated to obtain the second approximation sequence.

[0109] In one embodiment, the discreteness analysis module 13 is specifically used for: Individual discreteness analysis was performed on the first approximation sequence and the second approximation sequence respectively to obtain the individual approximation discreteness parameters; Collect multiple sets of first approximation sequences and multiple sets of second approximation sequences of other household photovoltaic inverters within the same distribution area where the household photovoltaic system is located; Based on the multiple sets of first approximation degree sequences and multiple sets of second approximation degree sequences of the same family, a collective discreteness analysis is performed to obtain the discretization parameters of the transformer area approximation degree.

[0110] Individual discreteness analysis was performed on the first and second approximation sequences respectively to obtain individual approximation discreteness parameters, including: The first volume approximation degree discrete parameters are obtained by performing discrete parameter calculation on the first approximation degree sequence. The second volume approximation degree discrete parameters are obtained by performing discrete parameter calculation on the second approximation degree sequence. Calculate the mean of the first approximation sequence and the second approximation sequence to obtain the first average approximation and the second average approximation, and assign the first approximation weight and the second approximation weight. The first approximation weight and the second approximation weight are used to calculate the first individual approximation degree discrete parameter and the second individual approximation degree discrete parameter by weighting.

[0111] Based on the multiple sets of first approximation degree sequences and multiple sets of second approximation degree sequences of the same family, a collective discreteness analysis is performed to obtain the discrete parameters of the transformer area approximation degree, including: Discrete parameters are calculated for multiple sets of first-degree-of-approximation sequences and multiple sets of second-degree-of-approximation sequences of the same family, respectively, to obtain multiple first-degree-of-approximation discrete parameters and multiple second-degree-of-approximation discrete parameters; Calculate the mean of multiple sets of first approximation degree sequences and multiple sets of second approximation degree sequences of the same family respectively, to obtain multiple first average approximation degree sequences and multiple second average approximation degree sequences of the same family, and assign multiple first approximation degree sequences and multiple second approximation degree sequences of the same family. Multiple first-family approximation weights and multiple second-family approximation weights are used to perform weighted calculations on multiple first-family approximation degree discrete parameters and multiple second-family approximation degree discrete parameters to obtain multiple weighted collective approximation degree discrete parameters, and the station area approximation degree discrete parameters are calculated.

[0112] In one embodiment, the fault warning classification module 14 is specifically used for: Based on the first approximation sequence and the second approximation sequence, the mean value is calculated to obtain the first average approximation and the second average approximation, and inverter fault approximation calculation is performed to obtain the basic fault risk rate. Based on the individual approximation degree discrete parameters, fault categories are classified to obtain individual power grid category rate and individual component category rate; Calculate the similarity between the individual approximation degree discrete parameters and the transformer area approximation degree discrete parameters, configure the grid category rate correction coefficient, and perform correction calculations on the individual grid category rate and the individual component category rate to obtain the corrected individual grid category rate and the corrected individual component category rate; Based on the basic fault risk rate, and combined with the modified individual grid category rate and the modified individual component category rate, the grid fault early warning parameters and component fault early warning parameters are calculated.

[0113] Based on the individual approximation discrete parameters, fault categories are classified to obtain individual grid category rates and individual component category rates, including: Based on the fault monitoring data of residential photovoltaic inverters, a set of discrete parameters for the approximation degree of individual samples is collected, and the inverter fault category corresponding to the discrete parameter for the approximation degree of each individual sample is obtained, resulting in a set of fault categories for sample inverters. Each fault category for a sample inverter includes either a grid category or a component category. The set of discrete parameters for the approximation degree of individual samples and the set of fault categories of sample inverters are split into K-fold partitions to obtain K sets of sample category classification data, and K fault category classification agents are constructed. The discrete parameters of the individual approximation degree are input into K fault category classification agents, and K inverter fault categories are output. The proportions of grid category and component category are calculated respectively to obtain the individual grid category rate and the individual component category rate.

[0114] Calculate the similarity between the individual approximation degree discrete parameters and the transformer area approximation degree discrete parameters, configure the grid category rate correction coefficient, and perform correction calculations on the individual grid category rate and the individual component category rate to obtain the corrected individual grid category rate and the corrected individual component category rate, including: The similarity between the discrete parameters of the individual approximation degree and the discrete parameters of the station area approximation degree is calculated to obtain the discrete consistency coefficient; The ratio of the discrete consistency coefficient to the average discrete consistency coefficient over a historical period is calculated to obtain the power grid category rate correction coefficient. The individual power grid category rate is corrected by using the power grid category rate correction coefficient to obtain the corrected individual power grid category rate, and the corrected individual component category rate is also calculated.

[0115] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0116] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0117] This specification and accompanying drawings are merely illustrative examples of the invention and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its scope. Therefore, if such modifications and modifications fall within the scope of the invention and its equivalents, the invention is intended to include these modifications and modifications.

Claims

1. A fault early warning method for residential photovoltaic inverters, characterized in that, The method includes: Collect the historical electrical parameter set sequence of household photovoltaic in the past preset time range, and extract the valley value and peak value respectively to obtain the historical valley value electrical parameter sequence and the historical peak value electrical parameter sequence; The approximation degree between the historical valley electrical parameter sequence and the first fault electrical threshold, and between the historical peak electrical parameter sequence and the second fault electrical threshold are calculated respectively to obtain the first approximation degree sequence and the second approximation degree sequence; Individual discreteness analysis was performed on the first and second approximation sequences, and collective discreteness analysis was performed on the monitoring data of the same station area to obtain individual approximation discreteness parameters and station area approximation discreteness parameters. Based on the individual approximation degree discrete parameters and the transformer area approximation degree discrete parameters, fault warning classification is performed to obtain power grid fault warning parameters and component fault warning parameters.

2. The fault early warning method for residential photovoltaic inverters according to claim 1, characterized in that, Collect historical electrical parameter sets of residential photovoltaic systems within a preset time range, extract valley and peak values ​​respectively, and obtain historical valley electrical parameter sequences and historical peak electrical parameter sequences, including: The electrical parameters of residential photovoltaic systems are collected over multiple unit time periods within a preset time range to obtain a sequence of historical electrical parameters, where each historical electrical parameter includes voltage. Extract the valley and peak values ​​from each historical electrical parameter set to obtain the historical valley electrical parameter sequence and the historical peak electrical parameter sequence.

3. The fault early warning method for residential photovoltaic inverters according to claim 1, characterized in that, The approximation degree between the historical valley electrical parameter sequence and the first fault electrical threshold, and between the historical peak electrical parameter sequence and the second fault electrical threshold are calculated respectively, to obtain the first approximation degree sequence and the second approximation degree sequence, including: Obtain the first fault electrical threshold and the second fault electrical threshold for a residential photovoltaic inverter. Calculate the similarity between each valley electrical parameter in the historical valley electrical parameter sequence and the first fault electrical threshold to obtain the first approximation sequence; The similarity between each peak electrical parameter in the historical peak electrical parameter sequence and the second fault electrical threshold is calculated to obtain the second approximation sequence.

4. The fault early warning method for residential photovoltaic inverters according to claim 1, characterized in that, Individual dispersion analysis was performed on the first and second approximation sequences, respectively, and collective dispersion analysis was performed on the monitoring data of the same monitoring area, to obtain individual approximation dispersion parameters and monitoring area approximation dispersion parameters, including: Individual discreteness analysis was performed on the first approximation sequence and the second approximation sequence respectively to obtain the individual approximation discreteness parameters; Collect multiple sets of first approximation sequences and multiple sets of second approximation sequences of other household photovoltaic inverters within the same distribution area where the household photovoltaic system is located; Based on the multiple sets of first approximation degree sequences and multiple sets of second approximation degree sequences of the same family, a collective discreteness analysis is performed to obtain the discretization parameters of the transformer area approximation degree.

5. The fault early warning method for residential photovoltaic inverters according to claim 4, characterized in that, Individual discreteness analysis was performed on the first and second approximation sequences respectively to obtain individual approximation discreteness parameters, including: The first volume approximation degree discrete parameters are obtained by performing discrete parameter calculation on the first approximation degree sequence. The second volume approximation degree discrete parameters are obtained by performing discrete parameter calculation on the second approximation degree sequence. Calculate the mean of the first approximation sequence and the second approximation sequence to obtain the first average approximation and the second average approximation, and assign the first approximation weight and the second approximation weight. The first approximation weight and the second approximation weight are used to calculate the first individual approximation degree discrete parameter and the second individual approximation degree discrete parameter by weighting.

6. The fault early warning method for residential photovoltaic inverters according to claim 4, characterized in that, Based on the multiple sets of first approximation degree sequences and multiple sets of second approximation degree sequences of the same family, a collective discreteness analysis is performed to obtain the discrete parameters of the transformer area approximation degree, including: Discrete parameters are calculated for multiple sets of first-degree-of-approximation sequences and multiple sets of second-degree-of-approximation sequences of the same family, respectively, to obtain multiple first-degree-of-approximation discrete parameters and multiple second-degree-of-approximation discrete parameters; Calculate the mean of multiple sets of first approximation degree sequences and multiple sets of second approximation degree sequences of the same family respectively, to obtain multiple first average approximation degree sequences and multiple second average approximation degree sequences of the same family, and assign multiple first approximation degree sequences and multiple second approximation degree sequences of the same family. Multiple first-family approximation weights and multiple second-family approximation weights are used to perform weighted calculations on multiple first-family approximation degree discrete parameters and multiple second-family approximation degree discrete parameters to obtain multiple weighted collective approximation degree discrete parameters, and the station area approximation degree discrete parameters are calculated.

7. The fault early warning method for residential photovoltaic inverters according to claim 1, characterized in that, Based on the individual approximation degree discrete parameters and the transformer area approximation degree discrete parameters, fault early warning classification is performed to obtain power grid fault early warning parameters and component fault early warning parameters, including: Based on the first approximation sequence and the second approximation sequence, the mean value is calculated to obtain the first average approximation and the second average approximation, and inverter fault approximation calculation is performed to obtain the basic fault risk rate. Based on the individual approximation degree discrete parameters, fault categories are classified to obtain individual power grid category rate and individual component category rate; Calculate the similarity between the individual approximation degree discrete parameters and the transformer area approximation degree discrete parameters, configure the grid category rate correction coefficient, and perform correction calculations on the individual grid category rate and the individual component category rate to obtain the corrected individual grid category rate and the corrected individual component category rate; Based on the basic fault risk rate, and combined with the modified individual grid category rate and the modified individual component category rate, the grid fault early warning parameters and component fault early warning parameters are calculated.

8. The fault early warning method for residential photovoltaic inverters according to claim 7, characterized in that, Based on the individual approximation discrete parameters, fault categories are classified to obtain individual grid category rates and individual component category rates, including: Based on the fault monitoring data of residential photovoltaic inverters, a set of discrete parameters for the approximation degree of individual samples is collected, and the inverter fault category corresponding to the discrete parameter for the approximation degree of each individual sample is obtained, resulting in a set of fault categories for sample inverters. Each fault category for a sample inverter includes either a grid category or a component category. The set of discrete parameters for the approximation degree of individual samples and the set of fault categories of sample inverters are split into K-fold partitions to obtain K sets of sample category classification data, and K fault category classification agents are constructed. The discrete parameters of the individual approximation degree are input into K fault category classification agents, and K inverter fault categories are output. The proportions of grid category and component category are calculated respectively to obtain the individual grid category rate and the individual component category rate.

9. The fault early warning method for residential photovoltaic inverters according to claim 7, characterized in that, Calculate the similarity between the individual approximation degree discrete parameters and the transformer area approximation degree discrete parameters, configure the grid category rate correction coefficient, and perform correction calculations on the individual grid category rate and the individual component category rate to obtain the corrected individual grid category rate and the corrected individual component category rate, including: The similarity between the discrete parameters of the individual approximation degree and the discrete parameters of the station area approximation degree is calculated to obtain the discrete consistency coefficient; The ratio of the discrete consistency coefficient to the average discrete consistency coefficient over a historical period is calculated to obtain the power grid category rate correction coefficient. The individual power grid category rate is corrected by using the power grid category rate correction coefficient to obtain the corrected individual power grid category rate, and the corrected individual component category rate is also calculated.

10. A fault early warning system for a residential photovoltaic inverter, characterized in that, The method for performing fault early warning of a residential photovoltaic inverter according to any one of claims 1 to 9 includes: The data acquisition and feature extraction module is used to collect the historical electrical parameter set sequence of household photovoltaic systems within a preset time range, extract the valley value and peak value respectively, and obtain the historical valley value electrical parameter sequence and the historical peak value electrical parameter sequence. The approximation calculation module is used to calculate the approximation degree between the historical valley electrical parameter sequence and the first fault electrical threshold, and between the historical peak electrical parameter sequence and the second fault electrical threshold, respectively, to obtain the first approximation degree sequence and the second approximation degree sequence; The discreteness analysis module is used to perform individual discreteness analysis on the first approximation sequence and the second approximation sequence, and to perform collective discreteness analysis on the monitoring data of the same station area, so as to obtain individual approximation discreteness parameters and station area approximation discreteness parameters. The fault warning classification module is used to classify fault warnings based on the individual proximity degree discrete parameters and the distribution area proximity degree discrete parameters, and to obtain power grid fault warning parameters and component fault warning parameters.

Images