Photovoltaic string state detection method and device, controller and readable storage medium

Abstract

The invention relates to a photovoltaic string state detection method and device, a controller and a readable storage medium. The method comprises the following steps: acquiring respective actual current values and respective actual irradiance values of a plurality of effective photovoltaic strings; current features are extracted based on the actual current values of the multiple effective photovoltaic group strings; determining an initial state detection result of each effective photovoltaic string through a trained initial state detection model according to the respective actual current values and current characteristics of the plurality of effective photovoltaic strings; when a target photovoltaic string with the initial state detection result representing an abnormal state exists in the multiple effective photovoltaic strings, determining a predicted current value of the target photovoltaic string through a trained current value prediction model based on an actual irradiance value of the target photovoltaic string; and determining a target state detection result of the target photovoltaic string according to the actual current value and the predicted current value of the target photovoltaic string. By adopting the method, the photovoltaic string state detection accuracy can be improved.

Description

Technical Field

[0001] This application relates to the field of photovoltaic technology, and in particular to a method, apparatus, controller and readable storage medium for detecting the status of photovoltaic strings. Background Technology

[0002] During the operation of a photovoltaic (PV) power plant, the PV strings generate direct current (DC) which is then supplied to the inverter, where it is converted into alternating current (AC). Therefore, the performance of the PV strings directly affects the power generation efficiency of the PV system. Any abnormality in the PV strings can not only result in power generation loss but may also lead to equipment failure. Thus, monitoring the status of the PV strings is crucial. Currently, most PV power plants rely on experience-based fixed threshold judgments for fault alarms related to PV strings.

[0003] However, in actual operating environments, photovoltaic systems often face various fluctuations, such as complex weather conditions and operational fluctuations, which makes traditional methods prone to false alarms and results in low accuracy for the status detection of photovoltaic strings. Summary of the Invention

[0004] Therefore, it is necessary to provide a photovoltaic string state detection method, device, controller, and readable storage medium that can improve the accuracy of state detection, addressing the aforementioned technical problems.

[0005] Firstly, this application provides a method for detecting the state of a photovoltaic string, including:

[0006] Collect the actual current value and actual irradiance value of each of the multiple effective photovoltaic strings;

[0007] Based on the actual current values ​​of each of the multiple effective photovoltaic strings, current characteristics are extracted;

[0008] Based on the actual current values ​​and current characteristics of each of the multiple effective photovoltaic strings, the preliminary state detection result of each effective photovoltaic string is determined through the trained preliminary state detection model.

[0009] If there is a target photovoltaic string among the multiple effective photovoltaic strings whose preliminary state detection results indicate an abnormal state, then based on the actual irradiance value of the target photovoltaic string, the predicted current value of the target photovoltaic string is determined through a trained current value prediction model.

[0010] The target state detection result of the target photovoltaic string is determined based on the actual current value and the predicted current value.

[0011] Secondly, this application also provides a photovoltaic string status detection device, comprising:

[0012] The acquisition module is used to acquire the actual current value and the actual irradiance value of each of the multiple valid photovoltaic strings.

[0013] The first detection module is used to extract current features based on the actual current values ​​of each of the multiple effective photovoltaic strings; and to determine the preliminary state detection result of each effective photovoltaic string based on the actual current values ​​of each of the multiple effective photovoltaic strings and the current features, through a trained preliminary state detection model.

[0014] The second detection module is used to determine the predicted current value of the target photovoltaic string based on its actual irradiance value and a trained current value prediction model when there is a target photovoltaic string among the multiple valid photovoltaic strings whose preliminary state detection results indicate an abnormal state; and to determine the target state detection result of the target photovoltaic string based on the actual current value and the predicted current value.

[0015] Thirdly, this application also provides a controller, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0016] Collect the actual current value and actual irradiance value of each of the multiple effective photovoltaic strings;

[0017] Based on the actual current values ​​of each of the multiple effective photovoltaic strings, current characteristics are extracted;

[0018] Based on the actual current values ​​and current characteristics of each of the multiple effective photovoltaic strings, the preliminary state detection result of each effective photovoltaic string is determined through the trained preliminary state detection model.

[0019] If there is a target photovoltaic string among the multiple effective photovoltaic strings whose preliminary state detection results indicate an abnormal state, then based on the actual irradiance value of the target photovoltaic string, the predicted current value of the target photovoltaic string is determined through a trained current value prediction model.

[0020] The target state detection result of the target photovoltaic string is determined based on the actual current value and the predicted current value.

[0021] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0022] Collect the actual current value and actual irradiance value of each of the multiple effective photovoltaic strings;

[0023] Based on the actual current values ​​of each of the multiple effective photovoltaic strings, current characteristics are extracted;

[0024] Based on the actual current values ​​and current characteristics of each of the multiple effective photovoltaic strings, the preliminary state detection result of each effective photovoltaic string is determined through the trained preliminary state detection model.

[0025] If there is a target photovoltaic string among the multiple effective photovoltaic strings whose preliminary state detection results indicate an abnormal state, then based on the actual irradiance value of the target photovoltaic string, the predicted current value of the target photovoltaic string is determined through a trained current value prediction model.

[0026] The target state detection result of the target photovoltaic string is determined based on the actual current value and the predicted current value.

[0027] The aforementioned photovoltaic string state detection method, device, controller, and readable storage medium first determine the preliminary state detection result of each effective photovoltaic string based on its actual current value and current characteristics using a trained preliminary state detection model. This allows for the initial screening of target photovoltaic strings that may exhibit abnormal states. For each target photovoltaic string, a predicted current value is further determined based on its actual irradiance value using a trained current prediction model. Given the strong physical relationship between irradiance and current in photovoltaic strings, the predicted current value is relatively accurate. The target state detection result of the target photovoltaic string can then be determined by combining its actual and predicted current values. Thus, by integrating multiple physical quantities such as current and irradiance values ​​through these multiple detection steps, the accuracy of photovoltaic string state detection can be improved compared to traditional methods that rely on fixed thresholds. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a schematic diagram of the photovoltaic string status detection process in one embodiment;

[0030] Figure 2 This is a schematic diagram of the effective string identification steps in one embodiment;

[0031] Figure 3 This is a structural block diagram of a photovoltaic string status detection device in one embodiment;

[0032] Figure 4 This is a diagram of the internal structure of the controller in one embodiment. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0034] It should be noted that the terms "first," "second," etc., used in this application may be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more.

[0035] In one embodiment, such as Figure 1 As shown, a photovoltaic string status detection method is provided. This embodiment illustrates the application of this method to a controller. In this embodiment, the method includes the following steps:

[0036] Step 110: Collect the actual current value and actual irradiance value of each of the multiple valid photovoltaic strings.

[0037] In this context, an effective photovoltaic (PV) string refers to a PV string with configured PV modules. The actual current value is the output current actually collected from the PV string during the PV system's operation. This output current is fed into the inverter, where it is converted into alternating current (AC). The actual irradiance value is the irradiance actually collected from the PV string during the PV system's operation. This irradiance value can be measured using a reference radiometer corresponding to the PV string, which is installed on the same plane and facing the same direction as the PV panels. The unit of irradiance value can be W / m². 2 (Watts per square meter)

[0038] For example, the controller can periodically collect the actual current values ​​of each of the multiple valid photovoltaic strings through current sensors and measure their actual irradiance values ​​through a reference irradiance meter. The controller can be a standalone dedicated monitoring device, the main controller inside the inverter, or a monitoring unit in the combiner box. The controller can execute steps 110 to 150 in each cycle, output the target state detection result, and also push alarms based on the target state detection result. This cycle can be each preset time interval, such as once every 5 minutes, once every 10 minutes, or others.

[0039] Step 120: Extract current features based on the actual current values ​​of each of the multiple valid photovoltaic strings.

[0040] The current characteristics can be the overall characteristics of the actual current values ​​of multiple effective photovoltaic strings, such as at least one of current peak characteristics, current fluctuation characteristics, current average characteristics, and current dispersion.

[0041] For example, the current characteristics include current peak characteristics, current fluctuation characteristics, current mean characteristics, and current dispersion. The controller can determine the maximum and minimum values ​​from the actual current values ​​of each of the multiple effective photovoltaic strings to obtain the current peak characteristics. The standard deviation, average value, and coefficient of variation calculated based on the actual current values ​​of each of the multiple effective photovoltaic strings are respectively determined as the current fluctuation characteristics, current mean characteristics, and current dispersion.

[0042] The formula for calculating the standard deviation can be: Where σ represents the standard deviation, N represents the total number of data points in the population, and x i The coefficient of variation can be represented by each data point in the overall data, and μ can represent the mean of the overall data. The formula for calculating the coefficient of variation can be... Where σ is the standard deviation, I avg This represents the average of the actual current values ​​of each of the multiple effective photovoltaic strings.

[0043] In one embodiment, before extracting current features, the actual current values ​​and actual irradiance values ​​of each of the multiple valid photovoltaic strings can be preprocessed. Specifically, preprocessing may include removing data from periods with zero actual irradiance, filtering outlier data, and imputing missing data. Outlier data may include at least one of negative actual current values ​​and negative actual irradiance values. Imputation can be achieved through interpolation, for example, using linear interpolation with the two preceding data points of the missing data.

[0044] Step 130: Based on the actual current values ​​and current characteristics of each of the multiple effective photovoltaic strings, determine the preliminary state detection result of each effective photovoltaic string using the trained preliminary state detection model.

[0045] The initial state detection model is used to initially screen out potentially abnormal photovoltaic strings. This model can be obtained by training a classification model, which can employ Gradient Boosting Decision Tree (GBDT), XGBoost (eXtreme Gradient Boosting), Random Forest, or others. After training and generating the initial state detection model, it can be loaded into the controller. The preliminary state detection results can include detection results representing normal or abnormal states.

[0046] For example, the controller can input the actual current values ​​and current characteristics of each of the multiple effective photovoltaic strings into the trained preliminary state detection model to obtain the preliminary state detection results of each effective photovoltaic string output by the preliminary state detection model.

[0047] Step 140: If there is a target photovoltaic string among multiple effective photovoltaic strings whose preliminary state detection results indicate an abnormal state, then based on the actual irradiance value of the target photovoltaic string, the predicted current value of the target photovoltaic string is determined through the trained current value prediction model.

[0048] The target photovoltaic string is the photovoltaic string whose abnormal state is characterized by preliminary state detection results. Based on the photovoltaic effect, it is known that there is a certain physical relationship between the current of the photovoltaic string and the irradiance, so the current of the photovoltaic string can be predicted by the irradiance.

[0049] The current prediction model can be obtained by training a regression model, specifically a linear regression model. After training and generating the current prediction model, it can be loaded into the controller. The current prediction model can specifically use irradiance G as the independent variable and the photovoltaic string current I as the variable. pead It can be expressed as a mathematical relationship between the dependent and dependent variables, for example, it can be denoted as I. pred =f(G) = a×G + b×G 2 +c, where the coefficients a, b, and c can be obtained by least squares fitting. The predicted current value is the output current value predicted by the current prediction model for the target photovoltaic string.

[0050] For example, when there is a target photovoltaic string among multiple valid photovoltaic strings whose preliminary state detection results indicate an abnormal state, the controller can input the actual irradiance value of the target photovoltaic string into the trained current value prediction model to obtain the predicted current value of the target photovoltaic string output by the current value prediction model.

[0051] In one embodiment, the current prediction model can be trained by grouping according to the irradiance value range. Each irradiance range corresponds to a current prediction model. In this embodiment, when there is a target photovoltaic string among multiple effective photovoltaic strings whose preliminary state detection results indicate an abnormal state, the controller can determine the irradiance range where the actual irradiance value of the target photovoltaic string is located, input the actual irradiance value of the target photovoltaic string into the current prediction model corresponding to the irradiance group, and obtain the predicted current value of the target photovoltaic string output by the current prediction model.

[0052] Different irradiance groups have different irradiance value ranges. For example, they can be divided into: extremely low irradiance range: less than 200 W / m². 2 Low irradiance range: 200W / m 2 Up to 400W / m 2 Medium irradiance range: 400W / m 2 Up to 600W / m 2 High irradiance range: 600W / m 2 Up to 800W / m 2 Extremely high irradiance range: greater than 800 W / m 2 .

[0053] Step 150: Determine the target state detection result of the target photovoltaic string based on the actual current value and the predicted current value.

[0054] Among them, the target state detection result is the detection result after secondary confirmation of the target photovoltaic string, which may include the detection result representing the normal state or the abnormal state.

[0055] For example, the controller can determine the deviation between the actual current value of the target photovoltaic string and the predicted current value of the target photovoltaic string, and determine the target state detection result of the target photovoltaic string based on the deviation.

[0056] In one embodiment, the actual current value and the actual irradiance value are collected periodically. Step 150 further includes: determining the relative deviation ratio of the actual current value of the target photovoltaic string to the predicted current value of the target photovoltaic string; when the relative deviation ratio is within a preset abnormal ratio range for a consecutive preset number of periods, the target state detection result of the target photovoltaic string is determined to represent an abnormal state.

[0057] The relative deviation ratio is the ratio of the deviation between the actual and predicted current values ​​to the predicted current value. The deviation must be no less than zero and can be obtained by subtracting the predicted current value from the actual current value and taking the absolute value. In one scenario, both the preset number of cycles and the preset anomaly ratio range can be set to one, corresponding to only one type of anomaly state. In another scenario, multiple preset number of cycles and preset anomaly ratio ranges can be set, corresponding to multiple anomaly states that can represent different degrees of anomaly.

[0058] In one embodiment, when the relative deviation ratio is within a preset abnormal ratio range for a consecutive preset number of periods, the step of determining the target state detection result of the target photovoltaic string as representing an abnormal state may include: when the relative deviation ratio is greater than a first preset abnormal ratio and not greater than a second preset abnormal ratio for a consecutive first preset number of periods, the target state detection result of the target photovoltaic string is determined to represent a first-level abnormal state; when the relative deviation ratio is greater than a second preset abnormal ratio and not greater than a third preset abnormal ratio for a consecutive second preset number of periods, the target state detection result of the target photovoltaic string is determined to represent a second-level abnormal state; when the relative deviation ratio is greater than a third preset abnormal ratio for a consecutive third preset number of periods, the target state detection result of the target photovoltaic string is determined to represent a third-level abnormal state; wherein, the first preset number of periods, the second preset number of periods, and the third preset number of periods decrease sequentially, the first preset abnormal ratio, the second preset abnormal ratio, and the third preset abnormal ratio increase sequentially, and the degree of abnormality represented by the first-level abnormal state, the second-level abnormal state, and the third-level abnormal state increases sequentially.

[0059] In this system, Level 1 anomalies represent minor anomalies, Level 2 anomalies represent general anomalies, and Level 3 anomalies represent severe anomalies. For example, the actual current value of the target photovoltaic string can be denoted as I. actual The predicted current value of the target photovoltaic string can be denoted as I. pred The relative deviation ratio can be expressed as The first preset number of cycles can be 10, the second preset number of cycles can be 5, and the third preset number of cycles can be 3. The first preset abnormality ratio can be 10%, the second preset abnormality ratio can be 15%, and the third preset abnormality ratio can be 25%. Then, when 10 consecutive cycles meet the requirements... It can be determined that the target photovoltaic string is in a slightly abnormal state, when the conditions are met for 5 consecutive cycles. It can be determined that the target photovoltaic string is in a general abnormal state, when the conditions are met for three consecutive cycles. It can be determined that the target photovoltaic string is in a severely abnormal state.

[0060] In another possible implementation, the controller can also determine whether the target photovoltaic string is in an abnormal state based on the slope change of the current-irradiance curve. Specifically, when there is a target photovoltaic string among multiple valid photovoltaic strings whose preliminary state detection results indicate an abnormal state, the controller can determine the first average rate of change of the actual current value and the actual irradiance value of the target photovoltaic string within a historical consecutive preset number of periods, and determine the second average rate of change of the actual current value and the actual irradiance value of the target photovoltaic string within a historical statistical period. When the first average rate of change deviates from the second average rate of change by more than a preset deviation amount, it can be determined that the target state detection result of the target photovoltaic string indicates an abnormal state.

[0061] The duration of the historical statistical period is longer than the duration of the preset quantity period. The period can be every 5 minutes, the preset quantity can be 10, then the duration of the preset quantity period can be 50 minutes, and the historical statistical period can be the past day, i.e., the past 24 hours. The first average rate of change and the second average rate of change can be the average slope of the current-irradiance curve within the corresponding period. The preset deviation can be, for example, 10%.

[0062] In the aforementioned photovoltaic string state detection method, firstly, based on the actual current values ​​and current characteristics of multiple effective photovoltaic strings, a pre-trained preliminary state detection model is used to determine the preliminary state detection result of each effective photovoltaic string. This allows for the initial screening of target photovoltaic strings that may have abnormal states. For each target photovoltaic string, a predicted current value is further determined based on its actual irradiance value using a pre-trained current prediction model. Given the working principle of photovoltaic strings, there is a strong physical relationship between irradiance and current, resulting in a relatively accurate predicted current value. Based on the actual and predicted current values ​​of the target photovoltaic string, the target state detection result can be determined. Thus, by combining multiple physical quantities such as current and irradiance values ​​through these multiple detection steps, the accuracy of photovoltaic string state detection can be improved compared to traditional methods that rely on fixed thresholds.

[0063] In an exemplary embodiment, prior to step 110, the photovoltaic string status detection method further includes the following valid string identification step: collecting the output current values ​​of each of the multiple photovoltaic strings connected to the inverter within a continuous time period; for each photovoltaic string, calculating the target proportion of data points whose output current values ​​do not reach a preset current value among all data points within the continuous time period; when the target proportion does not exceed the preset proportion upper limit, the photovoltaic string is determined to be a valid photovoltaic string.

[0064] The flowchart of the steps in this embodiment can be found as follows: Figure 2The diagram illustrates the valid string identification process. Valid string identification can be performed before starting to detect the status of photovoltaic strings. A continuous time period is a continuous period of time. A data point refers to an output current value of a photovoltaic string and its corresponding acquisition time. The preset current value can be, for example, 5%, 10%, or other values ​​of the rated current. The preset upper limit of the ratio can be, for example, 85%, 90%, or other values. A valid photovoltaic string refers to a photovoltaic string with configured modules, while an invalid photovoltaic string refers to a photovoltaic string without configured modules. If the target ratio exceeds the preset upper limit of the ratio, the photovoltaic string is determined to be an invalid photovoltaic string.

[0065] When calculating the target proportion, the weather conditions at the time of collecting the output current value can also be considered for statistical analysis. Data points during characteristic weather periods, such as rainy weather periods and heavy snow weather periods, can be excluded.

[0066] In an exemplary embodiment, after step 150, the photovoltaic string status detection method further includes: for photovoltaic strings exhibiting abnormalities, pushing abnormal alarm information to the operation and maintenance system. The abnormal alarm information may include the abnormal photovoltaic string number, the time of the abnormality, and the relative deviation ratio of the abnormal photovoltaic string. When pushing to the operation and maintenance system, the alarm can be selectively pushed according to the severity of the abnormality. For example, for severe abnormalities, an immediate alarm can be issued; for general abnormalities, a warning label can be added locally; and for minor abnormalities, an observation label can be added locally. Alternatively, all abnormalities can be pushed to the operation and maintenance system, and different labels can be added to the abnormal alarm information according to the severity of the abnormality. For example, a warning label can be added for general abnormalities.

[0067] In an exemplary embodiment, the preliminary state detection model is trained through a first training step, which includes: acquiring first sample data for each of multiple photovoltaic strings, the first sample data including first sample current values ​​and corresponding sample state information; identifying multiple valid photovoltaic strings based on the first sample current values ​​of each of the multiple photovoltaic strings, and preprocessing the first sample current values ​​of each of the multiple valid photovoltaic strings; extracting sample current features based on the preprocessed first sample current values; training the classification model to be trained using the sample current features and the preprocessed first sample current values ​​as input data, and using the corresponding sample state information as labels; and obtaining the trained preliminary state detection model after training is completed.

[0068] The first sample data is the sample data used in the first training step, the first sample current value is the output current value of the photovoltaic string used in the first training step, and the sample state information corresponds to the first sample current value and can be used to represent the state of the photovoltaic string. For example, the sample state information can be 1 or 0, where 1 can represent an abnormal state and 0 can represent a normal state.

[0069] The first sample data can be obtained from historical operational data. This operational data can include the photovoltaic (PV) string number where the anomaly occurred, the output current value of that PV string when the anomaly occurred, and the time of the anomaly. The historical time period can be, for example, 1 to 2 years. Anomalies in PV strings can manifest as: persistently low current, drastic current fluctuations, or zero current, etc.

[0070] Based on the first sample current values ​​of each of the multiple photovoltaic (PV) strings, multiple valid PV strings are identified. The implementation principle is the same as the valid string identification step described above, but the data used differs. The first sample current values ​​can replace the output current values ​​in the valid string identification step, and historical time periods can replace continuous time periods. By identifying valid PV strings and preprocessing the first sample current values ​​of each valid PV string, interference from invalid PV string sample data can be avoided during model training.

[0071] Preprocessing of the first sample data may include removing data from periods with zero actual irradiance, filtering outlier data, imputing missing data, and normalization. Sample current characteristics may include peak current characteristics, current fluctuation characteristics, mean current characteristics, and current dispersion, the calculation principles of which are the same as those for the aforementioned peak current characteristics, current fluctuation characteristics, mean current characteristics, and current dispersion.

[0072] The classification model can employ gradient boosting trees, XGBoost, random forests, or others. During model training, the preprocessed first sample data can be divided into training and test sets according to a certain ratio (e.g., 7:3). Model training can be performed on the H2O platform (an open-source deep learning platform). When using gradient boosting trees, the maximum tree depth can be 10, the learning rate can be 0.1, and the number of trees can be 100. Cross-validation is used to optimize model parameters and ensure the model's generalization ability. The test set can be used to evaluate model performance, requiring an accuracy of at least 90% and a recall of at least 85%.

[0073] In an exemplary embodiment, the current prediction model is trained through a second training step, which includes: acquiring second sample data for each of multiple photovoltaic strings, the second sample data including sample irradiance values ​​and corresponding second sample current values; identifying multiple valid photovoltaic strings based on the second sample current values ​​of each of the multiple photovoltaic strings, and preprocessing the second sample data of each of the multiple valid photovoltaic strings; grouping the second sample data of each of the multiple valid photovoltaic strings according to the irradiance value range, and within each group, using the preprocessed sample irradiance value as input data and the corresponding preprocessed second sample current value as label, training the regression model to be trained; and obtaining the trained current prediction model after training is completed.

[0074] The second sample data refers to the sample data used during the second training step. The sample irradiance value is the irradiance value of the photovoltaic string used during the second training step, and the second sample current value is the output current value of the photovoltaic string used during the second training step. The second sample data can be obtained from historical operational data. The operational data may also include sample irradiance values ​​for multiple photovoltaic strings, the corresponding second sample current values, and the corresponding acquisition time. The historical time period can be, for example, 1 to 2 years.

[0075] Based on the second sample current values ​​of each of the multiple photovoltaic strings, multiple valid photovoltaic strings are identified. The implementation principle can be the same as that of the above valid string identification step, but the data used is different. The second sample current value can be used to replace the output current value in the above valid string identification step, and the historical time period can be used to replace the continuous time period.

[0076] Preprocessing of the second sample data can include data cleaning according to predefined rules. Specifically, data points that do not conform to physical laws can be removed by analyzing the correlation between the sample irradiance and the second sample current value. For example, the second sample current value remains unchanged throughout the day, or the sample irradiance value is very low, but the corresponding second sample current value is very large.

[0077] The range of irradiance values ​​can be divided, for example, into: extremely low irradiance range: less than 200 W / m². 2 Low irradiance range: 200W / m 2 Up to 400W / m 2 Medium irradiance range: 400W / m 2 Up to 600W / m 2 High irradiance range: 600W / m 2 Up to 800W / m 2 Extremely high irradiance range: greater than 800 W / m 2 .

[0078] The regression model can specifically employ a linear regression model. During training, the model can be trained on the H2O platform, with training conducted separately for different groups based on the range of irradiance values. In the training of each group, the model hyperparameters are optimized through grid search to minimize the prediction error. When evaluating and validating the model, the coefficient of determination can be calculated, ideally not lower than 0.85, and the predictive stability of the model under different seasons and weather conditions can be verified.

[0079] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0080] Based on the same inventive concept, this application also provides a photovoltaic string state detection device for implementing the photovoltaic string state detection method described above. The solution provided by this device is similar to the implementation described in the above method; therefore, the specific limitations in one or more photovoltaic string state detection device embodiments provided below can be found in the limitations of the photovoltaic string state detection method described above, and will not be repeated here.

[0081] In one exemplary embodiment, such as Figure 3 As shown, a photovoltaic string status detection device 300 is provided, including: a data acquisition module 310, a first detection module 320, and a second detection module 330, wherein:

[0082] The acquisition module 310 is used to acquire the actual current value and the actual irradiance value of each of the multiple valid photovoltaic strings.

[0083] The first detection module 320 is used to extract current features based on the actual current values ​​of each of the multiple effective photovoltaic strings; and to determine the preliminary state detection result of each effective photovoltaic string based on the actual current values ​​and current features of each of the multiple effective photovoltaic strings through a trained preliminary state detection model.

[0084] The second detection module 330 is used to determine the predicted current value of the target photovoltaic string based on the actual irradiance value of the target photovoltaic string and through a trained current value prediction model when there is a target photovoltaic string among multiple effective photovoltaic strings whose preliminary state detection results indicate an abnormal state; and to determine the target state detection result of the target photovoltaic string based on the actual current value and the predicted current value of the target photovoltaic string.

[0085] In an exemplary embodiment, the actual current value and the actual irradiance value are collected periodically. The second detection module 330 is also used to determine the relative deviation ratio of the actual current value of the target photovoltaic string to the predicted current value of the target photovoltaic string. When the relative deviation ratio is within the preset abnormal ratio range for a consecutive preset number of cycles, the target state detection result of the target photovoltaic string is determined to represent an abnormal state.

[0086] In an exemplary embodiment, the second detection module 330 is further configured to determine that the target state detection result of the target photovoltaic string represents a first-level abnormal state when the relative deviation ratio is greater than a first preset abnormal ratio and not greater than a second preset abnormal ratio within a consecutive first preset number of cycles; to determine that the target state detection result of the target photovoltaic string represents a second-level abnormal state when the relative deviation ratio is greater than a second preset abnormal ratio and not greater than a third preset abnormal ratio within a consecutive second preset number of cycles; and to determine that the target state detection result of the target photovoltaic string represents a third-level abnormal state when the relative deviation ratio is greater than a third preset abnormal ratio within a consecutive third preset number of cycles. The first, second, and third preset number of cycles decrease sequentially, the first, second, and third preset abnormal ratios increase sequentially, and the degree of abnormality represented by each of the first, second, and third-level abnormal states increases sequentially.

[0087] In an exemplary embodiment, the current characteristics include current peak characteristics, current fluctuation characteristics, current mean characteristics, and current dispersion. The first detection module 320 is further configured to determine the maximum and minimum values ​​from the actual current values ​​of each of the multiple effective photovoltaic strings to obtain the current peak characteristics; and to determine the standard deviation, average value, and coefficient of variation calculated based on the actual current values ​​of each of the multiple effective photovoltaic strings as the current fluctuation characteristics, current mean characteristics, and current dispersion, respectively.

[0088] In an exemplary embodiment, the photovoltaic string status detection device 300 further includes a valid photovoltaic string identification module, which is used to collect the output current values ​​of each of the multiple photovoltaic strings connected to the inverter within a continuous time period; for each photovoltaic string, it counts the proportion of data points whose output current values ​​do not reach a preset current value to the target proportion of all data points within the continuous time period; when the target proportion does not exceed the preset proportion upper limit, the photovoltaic string is determined to be a valid photovoltaic string.

[0089] In an exemplary embodiment, the preliminary state detection model is trained by a first training module, which can be integrated into the photovoltaic string state detection device 300 or located in a server. The first training module is used to acquire first sample data for each of multiple photovoltaic strings, including first sample current values ​​and corresponding sample state information. Based on the first sample current values ​​of each of the multiple photovoltaic strings, multiple valid photovoltaic strings are identified, and the first sample current values ​​of each of the multiple valid photovoltaic strings are preprocessed. Sample current features are extracted based on the preprocessed first sample current values. The classification model to be trained is trained using the sample current features and the preprocessed first sample current values ​​as input data and the corresponding sample state information as labels. After training, the trained preliminary state detection model is obtained.

[0090] In an exemplary embodiment, the current prediction model is trained by a second training module, which can be integrated into the photovoltaic string state detection device 300 or located in a server. The second training module can acquire second sample data for each of multiple photovoltaic strings, including sample irradiance values ​​and corresponding second sample current values. Based on the second sample current values ​​of each of the multiple photovoltaic strings, multiple valid photovoltaic strings are identified, and the second sample data of each of the multiple valid photovoltaic strings is preprocessed. The second sample data of each of the multiple valid photovoltaic strings are grouped according to the irradiance value range, and within each group, the preprocessed sample irradiance value is used as input data, and the corresponding preprocessed second sample current value is used as label to train the regression model to be trained. After training, the trained current prediction model is obtained.

[0091] Each module in the aforementioned photovoltaic string status monitoring device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of the controller in hardware form or independent of it, or stored in the memory of the controller in software form, so that the processor can call and execute the corresponding operations of each module.

[0092] In one exemplary embodiment, a controller is provided, the internal structure of which can be shown in the following diagram. Figure 4As shown, the controller includes a processor, memory, input / output interfaces, and a communication interface. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the input / output interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a photovoltaic string state detection method.

[0093] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the controller to which the present application is applied. A specific controller may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0094] In one embodiment, a controller is also provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps in the above method embodiments.

[0095] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0096] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0097] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0098] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0099] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for detecting the status of photovoltaic strings, characterized in that, The method includes: Collect the actual current value and actual irradiance value of each of the multiple effective photovoltaic strings; Based on the actual current values ​​of each of the multiple effective photovoltaic strings, current characteristics are extracted; Based on the actual current values ​​and current characteristics of each of the multiple effective photovoltaic strings, the preliminary state detection result of each effective photovoltaic string is determined through the trained preliminary state detection model. If there is a target photovoltaic string among the multiple effective photovoltaic strings whose preliminary state detection results indicate an abnormal state, then based on the actual irradiance value of the target photovoltaic string, the predicted current value of the target photovoltaic string is determined through a trained current value prediction model. The target state detection result of the target photovoltaic string is determined based on the actual current value and the predicted current value.

2. The method according to claim 1, characterized in that, The actual current value and the actual irradiance value are collected periodically. Determining the target state detection result of the target photovoltaic string based on the actual current value and the predicted current value includes: Determine the relative deviation ratio between the actual current value of the target photovoltaic string and the predicted current value of the target photovoltaic string; If the relative deviation ratio is within the preset abnormal ratio range for a consecutive preset number of cycles, then the target state detection result of the target photovoltaic string is determined to represent an abnormal state.

3. The method according to claim 2, characterized in that, When the relative deviation ratio is within a preset abnormal ratio range for a consecutive preset number of periods, the target state detection result of the target photovoltaic string is determined to represent an abnormal state, including: If the relative deviation ratio is greater than the first preset abnormal ratio and not greater than the second preset abnormal ratio within a continuous first preset number of cycles, then the target state detection result of the target photovoltaic string is determined to represent a first-level abnormal state. If the relative deviation ratio is greater than the second preset abnormal ratio and not greater than the third preset abnormal ratio within a consecutive second preset number of cycles, then the target state detection result of the target photovoltaic string is determined to represent a second-level abnormal state. If the relative deviation ratio is greater than the third preset abnormal ratio within a consecutive third preset number of cycles, then the target state detection result of the target photovoltaic string is determined to represent a level three abnormal state. Among them, the first preset number of cycles, the second preset number of cycles, and the third preset number of cycles decrease in sequence, the first preset abnormality ratio, the second preset abnormality ratio, and the third preset abnormality ratio increase in sequence, and the degree of abnormality represented by the first-level abnormality state, the second-level abnormality state, and the third-level abnormality state increases in sequence.

4. The method according to claim 1, characterized in that, The current characteristics include peak current characteristics, current fluctuation characteristics, average current characteristics, and current dispersion. The extraction of current characteristics based on the actual current values ​​of each of the multiple effective photovoltaic strings includes: The maximum and minimum values ​​are determined from the actual current values ​​of each of the multiple effective photovoltaic strings to obtain the current peak characteristics; The standard deviation, average value, and coefficient of variation calculated based on the actual current values ​​of each of the multiple effective photovoltaic strings are respectively determined as current fluctuation characteristics, current mean characteristics, and current dispersion.

5. The method according to claim 1, characterized in that, The method further includes: Collect the output current values ​​of each of the multiple photovoltaic strings connected to the inverter within a continuous time period; For each photovoltaic string, the target proportion of data points whose output current value did not reach the preset current value is calculated out of all data points in the continuous time period. If the target ratio does not exceed the preset upper limit, the photovoltaic string is determined to be a valid photovoltaic string.

6. The method according to any one of claims 1-5, characterized in that, The preliminary state detection model is obtained through a first training step, which includes: Acquire first sample data for each of multiple photovoltaic strings, wherein the first sample data includes a first sample current value and corresponding sample status information; Based on the first sample current value of each of the multiple photovoltaic strings, multiple valid photovoltaic strings are identified, and the first sample current value of each of the multiple valid photovoltaic strings is preprocessed. Extract sample current features based on the preprocessed first sample current value; The classification model to be trained is trained using the sample current characteristics and the preprocessed first sample current value as input data, and the corresponding sample state information as labels. After training, a preliminary state detection model is obtained.

7. The method according to any one of claims 1-5, characterized in that, The current prediction model is trained through a second training step, which includes: Acquire the second sample data for each of the multiple photovoltaic strings. The second sample data includes the sample irradiance value and the corresponding second sample current value. Based on the second sample current value of each of the multiple photovoltaic strings, multiple valid photovoltaic strings are identified, and the second sample data of each of the multiple valid photovoltaic strings are preprocessed. According to the range of irradiance values, the second sample data of each of the multiple effective photovoltaic strings are grouped. Within each group, the preprocessed sample irradiance value is used as input data and the corresponding preprocessed second sample current value is used as label to train the regression model to be trained. After training, the trained current value prediction model is obtained.

8. A photovoltaic string status detection device, characterized in that, The device includes: The acquisition module is used to acquire the actual current value and the actual irradiance value of each of the multiple valid photovoltaic strings. The first detection module is used to extract current features based on the actual current values ​​of each of the multiple effective photovoltaic strings; and to determine the preliminary state detection result of each effective photovoltaic string based on the actual current values ​​of each of the multiple effective photovoltaic strings and the current features, through a trained preliminary state detection model. The second detection module is used to determine the predicted current value of the target photovoltaic string based on its actual irradiance value and a trained current value prediction model when there is a target photovoltaic string among the multiple valid photovoltaic strings whose preliminary state detection results indicate an abnormal state; and to determine the target state detection result of the target photovoltaic string based on the actual current value and the predicted current value.

9. A controller comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images