A household photovoltaic fault diagnosis method and system

Abstract

The present application belongs to the technical field of household photovoltaic, and particularly relates to a household photovoltaic fault diagnosis method and system, which comprises: collecting the volt-ampere characteristic curve data of a photovoltaic system and synchronous environmental irradiance; extracting at least one volt-ampere characteristic curve parameter, including short-circuit current, open-circuit voltage, maximum power or fill factor; according to the current irradiance, based on a preset dynamic health baseline, judging whether the system has a fault through a first preset rule. The first preset rule realizes fault diagnosis by obtaining a parameter expected health value and comparing it with an actual extracted value. The present application adopts a dynamic health baseline to adapt to environmental changes, combines an efficient parameter extraction algorithm with an intelligent fault judgment mechanism, significantly improves the diagnosis accuracy, reduces false positives and false negatives, and is suitable for low-cost, automated operation and maintenance of household photovoltaic systems.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of household photovoltaic, and particularly relates to a household photovoltaic fault diagnosis method and system. BACKGROUND

[0002] With the accelerated transformation of global energy structure, household photovoltaic power generation system, as an important part of distributed energy, is popularizing at an unprecedented speed worldwide. However, unlike large-scale ground photovoltaic power stations, household systems are usually installed on the roofs of residents, and lack professional operation and maintenance teams for daily inspection and maintenance.

[0003] At present, in large-scale photovoltaic power stations, professional I-V curve testers are widely used. These devices can accurately locate various complex faults such as component hot spots, bypass diode failure, junction box failure, and string mismatch by accurately measuring and analyzing collected parameters.

[0004] However, directly transplanting this system to the household market faces insurmountable obstacles: first, high cost and complex operation threshold. Professional I-V curve testers cost tens of thousands of yuan, which is an additional expense that ordinary household users cannot afford. More importantly, the operation and data analysis of these devices require technical personnel with certain photovoltaic professional knowledge to complete. Ordinary users neither have relevant theoretical knowledge nor have practical operation experience, and cannot independently complete testing and interpret reports.

[0005] Second, the fundamental defect of traditional diagnosis logic, the limitation of static baseline comparison method. Whether it is a simple diagnosis algorithm embedded in the inverter or an analysis software matched with a professional tester, the core logic of most of them relies on comparing the current measurement data with a fixed, ideal state baseline data set at the factory. This method seems intuitive, but it has a fatal flaw: it completely ignores the strong dependence of photovoltaic system output performance on environmental conditions, especially irradiance.

[0006] The power generation capacity of a photovoltaic system is not constant, but dynamically fluctuates with real-time changes in meteorological factors such as solar irradiance, ambient temperature, and wind speed. For example, a photovoltaic system with completely normal performance will have a full and standard I-V curve at noon on a sunny day. However, in the same system, the I-V curve will shrink as a whole in cloudy weather or weak light in the morning / evening, showing a significant decrease in short-circuit current Isc and maximum power Pmax. If it is still compared with the ideal baseline at noon on a sunny day, the data collected in cloudy weather will almost certainly be misjudged by the system as a serious degradation of the component or a string fault, resulting in a large number of invalid alarms.

[0007] Conversely, when the light condition is poor, some real but less severe faults (such as early performance degradation of components, slight local shading, etc.) may be masked by the environmental factor (low irradiance). In this case, the system output, although lower than it should be at this irradiance, may still be within the preset normal threshold compared to the ideal baseline, resulting in false negatives. This false alarm or no alarm situation allows small faults to continue to develop and eventually evolve into big problems, causing greater power generation losses. SUMMARY

[0008] To this end, the present application provides a household photovoltaic fault diagnosis method and system to solve the above technical problems.

[0009] The present application provides a household photovoltaic fault diagnosis method, comprising the following method steps: collecting a set of volt-ampere characteristic curve data points of a household photovoltaic system, and synchronously collecting the environmental irradiance corresponding to the time of the data point set; extracting at least one volt-ampere characteristic curve parameter value from the set of volt-ampere characteristic curve data points, the volt-ampere characteristic curve parameters including short-circuit current Isc, open-circuit voltage Voc, maximum power Pmax, or fill factor FF; determining whether the household photovoltaic system has a fault based on a preset dynamic health baseline using a first preset rule according to the currently collected environmental irradiance; wherein the first preset rule includes obtaining an expected health value of the volt-ampere characteristic curve parameter, comparing the extracted volt-ampere characteristic curve parameter value with the parameter expected health value, and determining whether the household photovoltaic system has a fault based on the comparison result.

[0010] Another aspect of the present application also provides a household photovoltaic fault diagnosis system, comprising: a collection module for collecting a set of volt-ampere characteristic curve data points of a household photovoltaic system, and synchronously collecting the environmental irradiance corresponding to the time of the data point set; a parameter extraction module for extracting at least one volt-ampere characteristic curve parameter value from the set of volt-ampere characteristic curve data points, the volt-ampere characteristic curve parameters including short-circuit current Isc, open-circuit voltage Voc, maximum power Pmax, or fill factor FF; a fault judgment module for determining whether the household photovoltaic system has a fault based on a preset dynamic health baseline using a first preset rule according to the currently collected environmental irradiance; wherein the first preset rule includes obtaining an expected health value of the volt-ampere characteristic curve parameter, comparing the extracted volt-ampere characteristic curve parameter value with the parameter expected health value, and determining whether the household photovoltaic system has a fault based on the comparison result.

[0011] The application solves the problems of high false positive rate and false negative rate caused by traditional static baseline due to environmental changes (such as cloudy weather) by constructing a dynamic health baseline, comparing the current performance of the system with the health state that should be under the same irradiance, and combining the efficient Pmax extraction algorithm of golden section and parabolic fitting, the shadow shielding identification of second derivative mutation analysis, and the isolated forest machine learning model enabled when the data is sparse, to realize accurate and comprehensive diagnosis of component failure. The method does not require expensive professional equipment, the diagnosis process is fully automatic and intelligent, significantly reduces the user threshold, and effectively guarantees the system power generation efficiency and long-term stable operation. BRIEF DESCRIPTION OF DRAWINGS

[0012] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor.

[0013] Other features, objects and advantages of the present application will become more apparent from the following detailed description of non-limiting embodiments with reference to the following drawings.

[0014] Figure 1 A household photovoltaic fault diagnosis method flowchart provided for the embodiments of the present application.

[0015] Figure 2 A maximum power extraction method schematic diagram provided for the embodiments of the present application.

[0016] Figure 3 A first preset rule-based parameter expected health value acquisition schematic diagram provided for the embodiments of the present application.

[0017] Figure 4 A household photovoltaic fault diagnosis system schematic diagram provided for the embodiments of the present application. DETAILED DESCRIPTION

[0018] In order to make the purposes, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions of the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0019] As Figure 1As shown, the embodiment of the present application discloses a household photovoltaic fault diagnosis method 100, comprising the following method steps: S101, collecting a set of voltage-current characteristic curve data points of a household photovoltaic system, and synchronously collecting environmental irradiance corresponding to the time of the data point set.

[0020] S102, extracting at least one voltage-current characteristic curve parameter value from the set of voltage-current characteristic curve data points, the voltage-current characteristic curve parameters including short-circuit current Isc, open-circuit voltage Voc, maximum power Pmax or fill factor FF.

[0021] S103, according to the current collected environmental irradiance, determining whether the household photovoltaic system has a fault based on a preset dynamic health baseline and using a first preset rule.

[0022] Among them, the first preset rule includes obtaining an expected health value of the voltage-current characteristic curve parameter, comparing the extracted voltage-current characteristic curve parameter value with the parameter expected health value, and determining whether the household photovoltaic system has a fault based on the comparison result.

[0023] In one embodiment, for S101, the inverter or optimizer usually has the function of scanning the I-V curve, and the household photovoltaic system usually integrates the inverter or optimizer, and the present application can be integrated with the firmware or supporting APP of these devices to directly read the I-V curve data generated by them. Or a dedicated, low-cost I-V curve scanning module can be installed. The module can be a small electronic load device, which periodically (such as at noon every day) performs a quick load scan on the photovoltaic string, and simultaneously collects environmental data through a low-cost irradiance meter.

[0024] Based on the above-mentioned device, a set of voltage-current characteristic curve data points of the household photovoltaic system is collected, and the data point set is composed of a series of (voltage V, current I) coordinate points, which depicts the complete electrical output characteristic of the system at a certain time. At the same time, the environmental irradiance data corresponding to the data point set collection time is synchronously collected. The irradiance data, as described above, can be obtained through the irradiance sensor installed on the module plane.

[0025] In one embodiment, at least one key voltage-current characteristic curve parameter value is extracted from the collected set of voltage-current characteristic curve data points. These parameters include but are not limited to short-circuit current (Isc), open-circuit voltage (Voc), maximum power (Pmax) and fill factor (FF). These parameters are core indicators of system performance, and abnormal changes directly reflect the existence of system faults.

[0026] Optionally, regarding the extraction of short-circuit current and open-circuit voltage, for example, the short-circuit current and open-circuit voltage are located at two boundaries of the I-V curve, and actual measurement data points are often very sparse, and there may even be no accurate zero voltage or zero current point. The embodiment adopts an adaptive linear interpolation algorithm to solve this problem.

[0027] Specifically, for Isc: in the sparse data point region where the voltage value tends to zero (such as 0-0.5V), several points closest to zero voltage are selected, linear interpolation is performed, and the current value when the voltage is 0V is extrapolated, that is, Isc.

[0028] For Voc: in the sparse data point region where the current value tends to zero (such as 0-0.1A), several points closest to zero current are selected, linear interpolation is performed, and the voltage value when the current is 0A is extrapolated, that is, Voc.

[0029] At the same time, according to the density and distribution of data points, the number and range of points used for interpolation are automatically adjusted to ensure the stability and accuracy of the interpolation result.

[0030] Optionally, for the extraction of maximum power (Pmax), the maximum power point is a single-peak extreme point on the I-V curve. The embodiment proposes a hybrid algorithm combining golden section search and quadratic polynomial fitting, which can quickly and accurately locate Pmax in the case of sparse data points or large noise. For example, as shown in Figure 2 S201, based on the volt-ampere characteristic curve, an initial search first voltage interval is set, a preset number of golden section searches are performed, and a single-peak second voltage interval containing the true power point is obtained, wherein the second voltage interval range is smaller than the first voltage interval.

[0031] S202, three voltage points are selected in the second voltage interval range, and a parabola passing through the three voltage points is constructed by quadratic polynomial fitting, wherein the three voltage points are respectively the interval midpoint Vb, one voltage point Va on the left side of the interval midpoint, and one voltage point Vc on the right side of the interval midpoint.

[0032] S203, taking the vertex of the parabola as the initial voltage prediction value corresponding to the maximum power Pmax, updating Vb with the initial voltage prediction value, and updating Va and Vc based on the updated Vb.

[0033] S204, repeating S202 and S203 until the change between the voltage prediction values of the last two times is less than a preset threshold, or the interval length of Va and Vc is less than a preset accuracy; the voltage prediction value obtained in the last time is the maximum power point voltage value.

[0034] Specifically, an initial voltage search interval is set (e.g. 0V to Voc), and then a preset number of times (e.g. 3-5 times) of golden section search is performed. Each search will reduce the interval, and eventually a "single-peak second voltage interval" with a smaller range is obtained, ensuring that the true maximum power point is located in this interval.

[0035] Three voltage points are selected in the second voltage interval: the midpoint Vb, and Va on the left and Vc on the right. A quadratic polynomial (parabola) is constructed using the three points (Va, Pa), (Vb, Pb), (Vc, Pc).

[0036] The vertex voltage of the parabola is calculated as the initial predicted value of the Pmax corresponding voltage. Update Vb with this predicted value, and accordingly update Va and Vc so that they are symmetrically distributed around the new Vb. Then repeat steps S202 and S203 for iterative fitting.

[0037] When the change between the voltage predicted values obtained by two consecutive iterations is less than a preset small threshold (e.g. 0.1V), or the interval length between Va and Vc is less than a preset accuracy requirement, the iteration stops. The voltage predicted value obtained in the last iteration is the final maximum power point voltage, and the corresponding power is Pmax.

[0038] The algorithm combines the global search capability of the golden section method and the local high-precision characteristics of the parabola fitting, has high computational efficiency and strong noise resistance, and is very suitable for deployment on resource-limited household devices.

[0039] Optionally, for the fill factor FF, after extracting Isc and Voc and Pmax for each I-V curve by the above embodiments, FF is calculated according to the formula FF = Pmax / (Isc * Voc).

[0040] In an embodiment, for S103, the dynamic health baseline in the present application is not a fixed value, but a database or mathematical model that stores or describes the health values of the system's parameters under different environmental irradiance. This baseline is established at the initial installation of the system or when the system is known to be in a healthy state, by collecting I-V curve data and extracting parameters under different irradiance conditions multiple times. Over time, the baseline can also be updated and optimized to adapt to the natural aging process of the components.

[0041] Optionally, the establishment of the dynamic baseline can include the first 1-2 weeks after the installation and commissioning of the system, which is considered to be the golden period when the system is in a "healthy" state. During this period, the system collects I-V curve data and corresponding irradiance at a high frequency (e.g. once an hour).

[0042] The collected raw (V, I) data points are cleaned to remove obviously erroneous data (e.g. negative voltage or current, values beyond physical limits, etc.).

[0043] For each I-V curve, Isc and Voc are extracted as described in the foregoing embodiments, and Pmax is extracted, and FF is calculated according to the formula FF = Pmax / (Isc*Voc).

[0044] The extracted parameters (Isc, Voc, Pmax, FF) are packaged with the irradiance G at the time of collection to form a health data tuple: (G, Isc, Voc, Pmax, FF).

[0045] All health data tuples are stored in a health baseline database. The database can be a simple CSV file, or a relational database or a time series database, and the present application does not limit it.

[0046] After 1-2 weeks of intensive collection, the database will accumulate hundreds to thousands of health data points covering different irradiance (from low irradiance in the morning to high irradiance at noon).

[0047] Optionally, photovoltaic modules will experience a slow natural degradation over time (about 0.5%-1% per year). If the baseline is fixed, then after a year, even if the system is completely healthy, its performance will be lower than the initial baseline due to natural degradation, and it will be misjudged as a failure. To solve this problem, the baseline needs to have the ability to update itself adaptively.

[0048] For example, the system can set a predetermined time period (e.g. every quarter), and automatically perform an I-V curve scan during a period of clear weather and stable irradiance.

[0049] The parameters extracted from this scan are compared with the historical data at the same irradiance in the baseline. If the performance degradation is within a reasonable range of natural degradation (e.g. quarterly degradation <0.3%), the system is considered to be still healthy.

[0050] The new data tuple confirmed as healthy this time is added to the health baseline database, gradually replacing the earliest data, so that the baseline can evolve smoothly following the natural aging process of the system, and always maintain an accurate description of the current health status.

[0051] Optionally, based on the health baseline database built based on the foregoing embodiments, a parameter expected health value of the volt-ampere characteristic curve parameter is obtained based on a first preset rule, for example, as shown in the following table: Figure 3 The method includes the following steps: S301, all health data tuples in the health baseline database are divided into multiple intervals according to the order of environmental irradiance size.

[0052] S302, locate the first interval and adjacent intervals to which the current environmental irradiance belongs.

[0053] S303, obtain the first health data tuple set corresponding to the first interval and adjacent intervals.

[0054] S304, fit the linear relationship between the parameter value and irradiance based on the first health data tuple set.

[0055] S305, calculate the expected health value corresponding to the current environmental irradiance based on the linear relationship and the current environmental irradiance.

[0056] Each health data tuple includes a set of parameter values of the photovoltaic under health state within a predetermined time range and the corresponding irradiance.

[0057] Specifically, all health data tuples in the health baseline database, as in the foregoing embodiments, each tuple contains a set of parameter values under health state and corresponding irradiance. According to the size of the irradiance, they are sequentially divided into multiple continuous intervals (such as one interval per 100 W / m²).

[0058] According to the currently collected environmental irradiance, locate the first interval to which it belongs, and at the same time obtain its adjacent left and right two intervals (if any).

[0059] From the first interval and its adjacent intervals, extract all corresponding health data tuples to form a first health data tuple set. The purpose of this is to obtain enough data points close to the current irradiance to improve the accuracy of fitting.

[0060] Based on the first health data tuple set, linearly fit the relationship between the target parameter (such as Pmax) and the irradiance (G) to obtain a linear equation: parameter = a*G + b.

[0061] Substitute the current environmental irradiance G into the above linear equation to calculate the expected health value of the parameter under this irradiance.

[0062] Then, compare the parameter value of the current volt-ampere characteristic curve extracted based on the foregoing embodiments with the expected health value, and make a judgment based on the comparison result. For example, if the actual Pmax value is lower than 95% of the expected health value, or the actual FF value is lower than 90% of the expected health value, it can be determined that the system has a performance degradation fault.

[0063] In an optional embodiment, at the beginning of system operation, or under some rare irradiance conditions, there might not be enough data in the health baseline database (i.e. the number of first health data tuples in S304 is less than a preset threshold, e.g. less than 5). In this case, the reliability of linear fitting will be greatly compromised. To this end, the second preset rule is introduced as a supplement.

[0064] Before S304, it is first determined whether the number of health data tuples in the adjacent interval is sufficient.

[0065] If the data is sufficient, linear calculation is performed according to S305.

[0066] If the data is insufficient, the second preset rule is enabled: the current collected environmental irradiance and the corresponding I-V characteristic curve parameter value are input into a pre-trained machine learning model. The model can predict whether the parameter value is abnormal under the current irradiance based on more extensive, non-linear historical data. If the model predicts that the result is “abnormal”, it is determined that the system has a fault.

[0067] Exemplarily, considering the requirements of household scenarios for computing resources and real-time performance, as well as the characteristics of the fault diagnosis problem (low input feature dimension, relatively limited sample size, and the need for strong generalization ability), the present embodiment adopts Isolation Forest (iForest) as a specific machine learning model.

[0068] iForest is an unsupervised anomaly detection algorithm that only needs health data for training. In the present embodiment, the model can be trained using the hundreds of health data tuples accumulated in the initialization phase. iForest has very fast training and inference speed, and small memory occupation, and is very suitable for deployment on resource-constrained edge devices (such as home gateways or inverters).

[0069] iForest isolates samples by randomly selecting features and split points. An abnormal point is usually isolated faster (i.e. separated out with fewer split times) because it is different from others, and its “Anomaly Score” will be higher.

[0070] Exemplarily, for the training of the iForest model, the training data is, for example, all health data tuples. The feature vector of each sample can be [G, Pmax] or [G, Isc, Voc] or the like.

[0071] Key parameters such as n_estimators (number of trees, set to 100) and contamination (estimated proportion of anomalies, set to 0.1, i.e. 10%).

[0072] When a new sample [G=250, Pmax=800] is input, the model outputs an "abnormal score". If the score is higher than a preset threshold (e.g. 0.5), it is determined to be abnormal.

[0073] In an optional embodiment, in addition to the overall performance decline, local shadow shielding is one of the common faults of household systems. This embodiment provides a method specifically for identifying shadow shielding, which exemplarily comprises: smoothing the currently collected I-V characteristic curve data; calculating the second derivative of each point on the smoothed curve, traversing all point second derivative values, and counting the number of mutation points whose second derivative values exceed a preset threshold; calculating the ratio of the short-circuit current extracted from the mutation points to the expected short-circuit current calculated based on the first preset rule, and determining that the household photovoltaic system has shadow shielding when the ratio corresponding to any mutation point is lower than a preset ratio.

[0074] Specifically, first, the original I-V curve data currently collected is smoothed (such as Savitzky-Golay filtering) to eliminate measurement noise.

[0075] Then, the second derivative of each data point on the smoothed curve is calculated. The I-V curve is smooth and monotonically decreasing under normal circumstances, and its second derivative is close to zero. However, when there is shadow shielding, the curve will appear "step" distortion, and the second derivative at these distortion points will suddenly increase.

[0076] Traverse all data points and count the number of "mutation points" whose second derivative values exceed a preset threshold. The existence of mutation points is a strong signal of shadow shielding.

[0077] To further confirm, the local short-circuit current (i.e. the current value corresponding to the step) extracted from each mutation point is calculated. This local short-circuit current is compared with the system's overall expected short-circuit current calculated based on the first preset rule, and the ratio is calculated.

[0078] If the ratio corresponding to any mutation point is lower than a preset ratio (e.g. 0.8), it can be very determined that the system has a shadow shielding fault. This is because shadow shielding will cause the component current output of the shaded part to drop significantly, forming a clear current step.

[0079] For better understanding of the inventive concept, as a specific example, after the baseline is established, the system enters the regular fault diagnosis process. The following takes the diagnosis of whether the maximum power Pmax is abnormal as an example to elaborate a complete diagnosis process.

[0080] Step 1: Trigger data collection: Diagnostics can be triggered manually by the user or automatically by the system. The strategy for automatic triggering can be to perform once a day when the irradiance reaches a peak value (such as 1000 W / m² or above) and the weather is stable, to ensure that the data is comparable.

[0081] Step 2: Collection and synchronization: Assume that at 12:00 on a certain day, the system collects a set of I-V curve data points, a total of 50 (V, I) pairs. At the same time, the irradiance sensor reading is G = 950 W / m².

[0082] Step 3: Data preprocessing and Pmax extraction: Smooth filtering is performed on the 50 data points to eliminate high-frequency noise.

[0083] Apply the preceding embodiment to extract Pmax, for example: Set the initial voltage interval as [0V, 40V] (assuming Voc is about 40V). After performing 3 times of golden section search, the interval is reduced to [30V, 35V].

[0084] Take Va = 31V, Vb = 32.5V, Vc = 34V within [30V, 35V]. Query or interpolate to get the corresponding power Pa, Pb, Pc.

[0085] Use three-point fitting parabola to calculate the vertex voltage as 32.8V. Update Vb = 32.8V, Va = 32.3V, Vc = 33.3V.

[0086] After repeating twice, the voltage prediction value stabilizes at 32.75V, with a change of less than 0.05V, meeting the convergence condition. Look up or interpolate to get the current at this voltage, and calculate Pmax = 3200W.

[0087] Step 4: Fault diagnosis based on dynamic baseline (first preset rule): S301: The irradiance in the healthy baseline database is divided into an interval of every 50 W / m². 950 W / m² belongs to the interval [950, 1000).

[0088] S302: Locate the first interval [950, 1000) and get its adjacent intervals [900, 950) and [1000, 1050).

[0089] S303: Extract a total of 15 healthy data tuples from the three intervals.

[0090] S304: Determine that 15 > preset threshold 5, the data is sufficient, and enter linear fitting. Perform linear regression with Pmax as Y axis and irradiance G as X axis, and fit the equation: Pmax_expected = 3.4*G-200.

[0091] S305: Substituting G=950 into the equation, the expected health Pmax is calculated to be 3.4*950-200=3030W.

[0092] Comparison and judgment: Actual Pmax = 3200W > Expected 3030W. The system is judged as "no fault, good performance".

[0093] Step 5: Special case handling, enabling machine learning (second preset rule): Assume that on another cloudy day, the collected irradiance G = 250 W / m². Repeat Step 4: S301-S303: Locate the interval [250, 300) and its adjacent intervals, but only extract 3 health data tuples.

[0094] S304: If the value of 3 is less than the preset threshold of 5, the data is insufficient. Linear fitting will not be performed, and the second preset rule will be used instead.

[0095] Machine learning model inference: Input the current data (G=250, Pmax_actual=800W) into the pre-trained machine learning model. The model outputs an anomaly score of 0.2, which is below the threshold and is judged as "normal".

[0096] Step 6: Shadow Occlusion Detection: Smooth the acquired IV curve.

[0097] Calculate the second derivative for each data point. Suppose we find that the absolute value of the second derivative at 3 points is much larger than that at the other points, and these points are marked as "breakdown points".

[0098] Calculate the local current values ​​corresponding to these three abrupt change points, which are 8A, 7.5A, and 8.2A, respectively.

[0099] Meanwhile, based on the first preset rule, the expected Isc of the system under G=950W / m² is calculated to be 9A.

[0100] Calculate the ratios: 8 / 9≈0.89, 7.5 / 9≈0.83, 8.2 / 9≈0.91.

[0101] Since 7.5 / 9 = 0.83 < the preset ratio of 0.85, the system determines that there is a shadow occlusion fault. Optionally, it can push an alarm to the user: "Component occlusion detected. Please check for bird droppings or leaves."

[0102] Figure 4 A residential photovoltaic (PV) fault diagnosis system 400 is shown. This device embodiment is similar to... Figure 1 Corresponding to the method embodiment shown, it specifically includes: a data acquisition module 401, used to acquire a set of data points of the current-voltage characteristic curve of the household photovoltaic system, and to simultaneously acquire the ambient irradiance corresponding to the time of the data point set.

[0103] The parameter extraction module 402 is configured to extract at least one value of a parameter of the I-V characteristic curve from the set of I-V characteristic curve data points, the parameter of the I-V characteristic curve comprising a short-circuit current Isc, an open-circuit voltage Voc, a maximum power Pmax, or a fill factor FF.

[0104] The fault determination module 403 is configured to determine, according to a current collected ambient irradiance, whether the household photovoltaic system has a fault based on a preset dynamic health baseline and using a first preset rule.

[0105] The first preset rule comprises obtaining an expected health value of the parameter of the I-V characteristic curve, comparing the extracted value of the parameter of the I-V characteristic curve with the expected health value of the parameter of the I-V characteristic curve, and determining, based on a comparison result, whether the household photovoltaic system has a fault.

[0106] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts of each embodiment can be referred to each other. For the device disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method.

[0107] The above description of the disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to the embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for diagnosing a fault of a household photovoltaic system, characterized by, The method comprises the following steps: Collecting a set of data points of a voltage-current characteristic curve of a household photovoltaic system, and synchronously collecting environmental irradiance corresponding to the time of the set of data points; extracting at least one voltage-current characteristic curve parameter value from the set of data points of the voltage-current characteristic curve, the voltage-current characteristic curve parameters including short-circuit current Isc, open-circuit voltage Voc, maximum power Pmax or fill factor FF; determining whether the household photovoltaic system has a fault based on a preset dynamic health baseline and a first preset rule according to the currently collected environmental irradiance; wherein the first preset rule comprises obtaining an expected health value of the voltage-current characteristic curve parameters, comparing the extracted voltage-current characteristic curve parameter value with the parameter expected health value, and determining whether the household photovoltaic system has a fault based on the comparison result; the extraction method of the maximum power Pmax comprises the following steps: S201, based on the voltage-current characteristic curve, setting an initial search first voltage interval, performing a preset number of golden section searches, and obtaining a single-peak second voltage interval containing the real power point, wherein the range of the second voltage interval is smaller than that of the first voltage interval; S202, selecting three voltage points in the range of the second voltage interval, and constructing a parabola passing through the three voltage points by quadratic polynomial fitting, wherein the three voltage points are the interval midpoint Vb, the voltage point Va on the left side of the interval midpoint, and the voltage point Vc on the right side of the interval midpoint; S203, taking the vertex of the parabola as the initial voltage prediction value corresponding to the maximum power Pmax, updating Vb with the initial voltage prediction value, and updating Va and Vc based on the updated Vb; S204, repeating S202 and S203 until the change between the voltage prediction values of the last two times is less than a preset threshold, or the interval length of Va and Vc is less than a preset accuracy; the voltage prediction value obtained at the last time is the maximum power point voltage value.

2. The method of claim 1, wherein the method further comprises: The method comprises the following steps: S301, dividing all health data tuples in a health baseline database into multiple intervals in order based on the size of environmental irradiance; S302, positioning to the first interval and adjacent intervals in which the current environmental irradiance is located; S303, obtaining a first set of health data tuples corresponding to the first interval and adjacent intervals; S304, fitting a linear relationship between the parameter value and irradiance based on the first set of health data tuples; S305, calculating the expected health value corresponding to the current environmental irradiance based on the linear relationship and the current environmental irradiance; wherein each health data tuple comprises a set of voltage-current characteristic curve parameter values and corresponding irradiance of the household photovoltaic system in a predetermined time range under a healthy state.

3. The method of claim 2, wherein the method further comprises: Before step S304, further comprising: judging whether the number of the first health data tuples is greater than a preset threshold; if yes, entering step S305; otherwise, determining whether the household photovoltaic system has a fault by using a second preset rule, wherein the second preset rule comprises: inputting the current collected environmental irradiance and the corresponding I-V characteristic curve parameter value into a pre-trained machine learning model, and predicting whether the I-V characteristic curve parameter value is abnormal based on the machine learning model; if the I-V characteristic curve parameter value is abnormal, it is determined that the household photovoltaic system has a fault.

4. The method of claim 1, wherein the method further comprises: Further comprising: performing smoothing processing on the current collected I-V characteristic curve data; calculating the second derivative of each point on the curve after the smoothing processing, traversing all point second derivative values, and counting the number of mutation points whose second derivative values exceed a preset threshold; calculating the ratio of the short-circuit current extracted from the mutation points to the expected short-circuit current calculated based on the first preset rule; and when the ratio corresponding to any mutation point is lower than a preset ratio, it is determined that the household photovoltaic system has a shadow obstruction.

5. The method of claim 1, wherein the method further comprises: The extraction of the short-circuit current Isc and the open-circuit voltage Voc specifically comprises: using an adaptive linear interpolation algorithm to calculate the short-circuit current Isc and the open-circuit voltage Voc values in the sparse data point area where the voltage or current approaches zero.

6. A home-use photovoltaic fault diagnostic system characterized by comprising: Comprise: a collection module configured to collect a set of I-V characteristic curve data points of a household photovoltaic system and synchronously collect environmental irradiance corresponding to the time of the set of data points; a parameter extraction module configured to extract at least one I-V characteristic curve parameter value from the set of I-V characteristic curve data points, the I-V characteristic curve parameters comprising a short-circuit current Isc, an open-circuit voltage Voc, a maximum power Pmax, or a fill factor FF; a fault determination module configured to determine whether the household photovoltaic system has a fault based on a current collected environmental irradiance and a preset dynamic health baseline by using a first preset rule; wherein the first preset rule comprises: obtaining an expected health value of the I-V characteristic curve parameters, comparing the extracted I-V characteristic curve parameter value with the parameter expected health value, and determining whether the household photovoltaic system has a fault based on the comparison result; further comprising a maximum power Pmax extraction module configured to set an initial search first voltage interval based on the I-V characteristic curve, perform a preset number of golden section searches, obtain a single-peak second voltage interval containing a real power point, wherein the range of the second voltage interval is smaller than that of the first voltage interval; select three voltage points in the range of the second voltage interval, construct a parabola passing through the three voltage points by quadratic polynomial fitting, wherein the three voltage points are a middle point Vb of the interval, a voltage point Va on the left side of the middle point, and a voltage point Vc on the right side of the middle point; take the vertex of the parabola as an initial voltage prediction value corresponding to the maximum power Pmax, update Vb based on the initial voltage prediction value, and update Va and Vc based on the updated Vb; until the change between the voltage prediction values of two consecutive times is less than a preset threshold, or the interval length of Va and Vc is less than a preset accuracy; and the voltage prediction value obtained in the last time is the maximum power point voltage value.

7. A home-use photovoltaic fault diagnostic system according to claim 6, wherein The application also comprises an expected health value obtaining module, which is used for dividing all health data tuples in the health baseline database into multiple intervals in sequence based on the size of environmental irradiance; positioning to the first interval and adjacent interval where the current environmental irradiance is located based on the current environmental irradiance; and obtaining the corresponding first health data tuple set in the first interval and adjacent interval. Based on the first health data tuple set, a linear relationship between the parameter value and irradiance is fitted; and the expected health value corresponding to the current environmental irradiance is calculated based on the linear relationship and the current environmental irradiance; wherein each health data tuple comprises a set of parameter values of the photovoltaic in the health state within a predetermined time range and the corresponding irradiance.

8. The home photovoltaic fault diagnostic system of claim 6, wherein, The application also comprises a shielding judgment module, which is used for smoothing the current collected volt-ampere characteristic curve data; calculating the second derivative of each point on the curve after smoothing; traversing all point second derivative values, and counting the number of mutation points whose second derivative values exceed a preset threshold; calculating the ratio of the short-circuit current extracted from the mutation point to the expected short-circuit current calculated based on the first preset rule; and determining that the household photovoltaic system is shadowed when the ratio corresponding to any mutation point is lower than a preset ratio.

Images