A photovoltaic power generation fault prediction and diagnosis system and method based on data drive

By collecting photovoltaic panel feature data in real time and adjusting the temperature difference threshold dynamically, and building fault prediction and diagnosis models, the problems of inflexible prediction lag and inflexible threshold settings in the existing technology are solved, and early identification and maintenance of photovoltaic panel faults are realized Improved efficiency.

CN118232832BActive Publication Date: 2024-11-05SHENZHEN LYTRAN TECH
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202410332923.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-03-22
Publication Date
2024-11-05
Estimated Expiration
2044-03-22

AI Technical Summary

Technical Problem

The existing photovoltaic power generation fault prediction and diagnosis system has problems such as lag and inflexible threshold settings, which makes it difficult to detect faults in a timely manner and balance maintenance workload, affecting the normal operation and power generation efficiency of photovoltaic panels.

Method used

By dividing the photovoltaic panel into multiple areas, collect feature data in real time and input the pre-built fault prediction model, combine real-time picture data of temperature abnormalities, and use dynamically adjusted local temperature difference thresholds to build fault prediction and diagnostic models to identify fault types to identify fault types using dynamically adjusted local temperature difference thresholds. and reasons.

Benefits of technology

It realizes early prediction and accurate diagnosis of photovoltaic panel faults, reduces the probability of failure, improves maintenance efficiency, and ensures the normal operation and power generation efficiency of photovoltaic panels.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118232832B_ABST
    Figure CN118232832B_ABST
Patent Text Reader

Abstract

The present invention discloses a photovoltaic power generation fault prediction and diagnosis system and method based on data-driven, the method comprising: pre-dividing a photovoltaic panel into i divided areas, collecting real-time characteristic data of the photovoltaic panels in the i divided areas within a unit time; inputting the real-time characteristic data into a pre-constructed fault prediction model to obtain the operating state of the photovoltaic panel after a unit time; the operating state includes the absence of a fault and the presence of a fault, and the presence of a fault includes local blocking damage and temperature anomaly; collecting real-time image data of the corresponding photovoltaic panel according to the temperature anomaly; inputting the collected image data into a pre-constructed fault diagnosis model to obtain the fault type, and the fault type includes internal anomaly and external anomaly; facilitating and timely repairing or replacing the faulty part, reducing the probability of fault occurrence, timely discovering heat spots and improving the efficiency of treating heat spots, so as to ensure the normal operation and power generation efficiency of photovoltaic panel power generation components.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of power grid monitoring, and in particular to a photovoltaic power generation fault prediction and diagnosis system and method based on data drive. Background Art

[0002] Photovoltaic power generation systems play an important role in the field of renewable energy. However, due to environmental and equipment factors, photovoltaic power generation systems may fail, resulting in reduced power generation efficiency and equipment damage. Therefore, it is of great significance to develop a system and method that can accurately predict and diagnose photovoltaic power generation system failures.

[0003] Most of the existing photovoltaic power generation fault prediction and diagnosis systems are based on the analysis and diagnosis of data after the fault occurs. The prediction of specific faults has a certain lag. At this time, the fault has already occurred and the probability of fault occurrence cannot be reduced, thus affecting the normal operation, power generation efficiency and maintenance efficiency of photovoltaic panel power generation components in photovoltaic power generation.

[0004] Secondly, when it comes to threshold comparison in analysis, diagnosis and prediction, the thresholds are mostly fixed. For example, if the local temperature difference threshold is fixed, it cannot be adjusted in real time according to the actual situation. If the fixed value is low, false alarms and maintenance workload will increase; otherwise, it will be the opposite, making it difficult to balance timely detection and maintenance workload.

[0005] If this continues, it will be difficult to ensure the normal operation, power generation efficiency, maintenance efficiency and lifespan of photovoltaic panel power generation components. Summary of the invention

[0006] The purpose of the present invention is to provide a photovoltaic power generation fault prediction and diagnosis system and method based on data drive to solve the deficiencies in the background technology.

[0007] In order to achieve the above-mentioned object, the present invention provides the following technical solution: a data-driven photovoltaic power generation fault prediction and diagnosis method, comprising: pre-dividing the photovoltaic panel into i divided areas, collecting the real-time characteristic data of the photovoltaic panel in the i divided areas per unit time;

[0008] Input the real-time characteristic data into the pre-built fault prediction model to obtain the operating status of the photovoltaic panel after a unit time; the operating status includes the absence of faults and the presence of faults, and the presence of faults includes local blocking damage and temperature abnormality;

[0009] Collect real-time image data of corresponding photovoltaic panels according to temperature anomalies;

[0010] The collected image data is input into the pre-built fault diagnosis model to obtain the fault type, which includes internal abnormality and external abnormality.

[0011] Preferably, the real-time characteristic data includes n current change rates, n voltage change rates, temperature differences, and local temperature difference thresholds corresponding to the temperature differences within a unit time.

[0012] Preferably, the n current change rate acquisition methods include:

[0013] The unit time is equally divided into n+1 continuous time intervals, and the current value of the photovoltaic panel is collected once in each time interval;

[0014] The current value in the next time interval is subtracted from the current value in the previous time interval to obtain the current difference between adjacent regions, and the current difference is divided by the current value in the previous time interval to obtain a current change rate, and so on to obtain n current change rates.

[0015] Preferably, the n voltage change rate acquisition methods include:

[0016] The voltage value of the photovoltaic panel is collected once in each time interval;

[0017] The voltage value in the next time interval is subtracted from the voltage value in the previous time interval to obtain the voltage difference between adjacent regions, and the voltage difference is divided by the voltage value in the previous time interval to obtain a voltage change rate, and so on to obtain n voltage change rates.

[0018] Preferably, the temperature difference acquisition method comprises:

[0019] Mark the temperature value collected in the last time interval of each divided area as the regional temperature value;

[0020] Subtract the regional temperature value of the adjacent divided area from the regional temperature value to obtain k temperature difference values.

[0021] Preferably, the local temperature difference threshold acquisition method includes:

[0022] Pre-collecting the threshold setting data of the unit time before the photovoltaic panel has heat spots, the threshold setting data including the historical temperature difference characteristic data and the historical local temperature difference threshold value corresponding to the historical temperature difference characteristic data; the historical temperature difference characteristic data includes the ambient temperature, light intensity, wind speed, solar radiation and panel material;

[0023] Each group of historical temperature difference feature data is used as an input of a difference threshold setting model in the form of a vector. The difference threshold setting model uses a historical local temperature difference threshold corresponding to a group of historical temperature difference feature data as an output, uses a historical local temperature difference threshold actually corresponding to a group of historical temperature difference feature data as a prediction target, and uses minimizing the difference threshold setting model loss function value as a training target; when the difference threshold setting model loss function value is less than or equal to a preset target loss value, the training is stopped to obtain a difference threshold setting model, and the difference threshold setting model is one of a SVM regression model, a random forest regression model, a neural network regression model, and the like;

[0024] The threshold setting data collected in real time is input into the trained difference threshold setting model to obtain the local temperature difference threshold.

[0025] Preferably, the method for pre-constructing the fault prediction model includes:

[0026] Pre-collecting operation data per unit time during the operation stage of the photovoltaic panel, the operation data including historical characteristic data and operation status corresponding to the historical characteristic data; wherein, when the operation status is that there is a fault, the corresponding historical characteristic data is the historical characteristic data per unit time before the photovoltaic panel has heat spots during the operation stage;

[0027] Convert historical feature data into feature vectors as follows:

[0028] , The corresponding operating status is ; is the first set of eigenvectors; is the current change rate of the nth time interval in the first set of eigenvectors; is the voltage change rate of the nth time interval in the first set of eigenvectors; is the temperature difference between a temperature value of a region in the first set of eigenvectors and the temperature value of the region divided in the kth direction;

[0029] is the local temperature difference threshold corresponding to the temperature difference in the first set of eigenvectors; include , and , To avoid any faults, It is a local obstruction damage. Abnormal temperature;

[0030] The operating data is divided into a training set and a test set, and a first classifier is constructed. The historical feature data in the training set is used as input data, and the operating status in the training set is used as output data. The first classifier is trained to obtain an initial first classifier, and the initial first classifier is tested using the test set. The first classifier that meets the preset accuracy is output as a fault prediction model, and the fault prediction model is a naive Bayes model or a support vector machine model.

[0031] Preferably, the external anomalies include cracks in the photovoltaic panel, contaminants in the photovoltaic panel and shadows in the photovoltaic panel; the internal anomalies include uneven battery cells and aging of battery cells.

[0032] Preferably, the method for constructing the fault diagnosis model includes:

[0033] Obtain r groups of data, where r is a positive integer greater than 1, and the data includes historical image data and historical fault types. The historical image data and historical fault types are used as sample sets, and the sample sets are divided into training sets and test sets. A second classifier is constructed, and the historical image data in the training set is used as input data, and the historical fault types in the training set are used as output data. The second classifier is trained to obtain an initial second classifier, and the initial second classifier is tested using the test set. The second classifier that meets the preset accuracy is output as a fault diagnosis model, and the fault diagnosis model is one of a convolutional neural network and a recurrent neural network.

[0034] A photovoltaic power generation fault prediction and diagnosis system based on data drive, used to implement the photovoltaic power generation fault prediction and diagnosis method based on data drive, comprising:

[0035] The first data acquisition module divides the photovoltaic panel into i divided areas in advance, and collects the real-time characteristic data of the photovoltaic panels in the i divided areas per unit time;

[0036] The data analysis module is used to input the real-time characteristic data into the pre-built fault prediction model to obtain the operating status of the photovoltaic panel after a unit time; the operating status includes the absence of faults and the presence of faults, and the presence of faults includes local blocking damage and temperature abnormality;

[0037] The second data acquisition module is used to collect real-time image data of the corresponding photovoltaic panel according to the temperature anomaly;

[0038] The fault diagnosis module is used to input the collected image data into a pre-built fault diagnosis model to obtain the fault type, which includes internal abnormality and external abnormality.

[0039] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0040] By pre-dividing the photovoltaic panel into i divided areas, real-time feature data of each divided area is collected in real time, and the collected real-time feature data is input into a pre-trained fault prediction model, it is possible to obtain in real time whether each divided area has a fault after a unit time. If a fault exists, the fault can be predicted in advance, and the specific fault can be identified and the cause of the fault can be located, so as to facilitate timely repair or replacement of the faulty part, reduce the probability of fault occurrence, timely discover heat spots and improve the efficiency of treating heat spots, so as to ensure the normal operation and power generation efficiency of photovoltaic panel power generation components.

[0041] Secondly, in the process of setting the local temperature difference threshold, comprehensive consideration of multiple factors can more comprehensively consider the impact of multiple factors on the temperature of the photovoltaic panel, and improve the accuracy of judging local temperature anomalies; real-time setting of the corresponding local temperature difference threshold can achieve a balance between timely detection and maintenance workload; determining the appropriate local temperature difference threshold can reduce the maintenance workload while also timely detecting heat spots and improving the efficiency of treating heat spots, effectively ensuring the normal operation, power generation efficiency and life of photovoltaic panel power generation components. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0043] Figure 1 A schematic diagram of a photovoltaic power generation fault prediction and diagnosis system based on data drive of the present invention;

[0044] Figure 2 This is a schematic diagram of a photovoltaic panel of the present invention;

[0045] Figure 3 It is a schematic diagram of a data-driven photovoltaic power generation fault prediction and diagnosis method of the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be an intermediate element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be an intermediate element at the same time. Example 1

[0048] like Figure 1 As shown, this embodiment is a data-driven photovoltaic power generation fault prediction and diagnosis system, including a first data acquisition module, a data analysis module, a second data acquisition module and a fault diagnosis module, and the first data acquisition module, the data analysis module, the second data acquisition module and the fault diagnosis module are connected by wire and / or wirelessly.

[0049] The first data acquisition module divides the photovoltaic panel into i divided areas in advance, and collects real-time characteristic data of the photovoltaic panels in the i divided areas per unit time. The real-time characteristic data includes n current change rates, n voltage change rates, temperature differences, and local temperature difference thresholds corresponding to the temperature differences per unit time.

[0050] The photovoltaic panel is divided into i division areas, and the technicians make corresponding divisions according to the design circuit of the photovoltaic cell in the photovoltaic panel. One photovoltaic cell can be divided into one area, or multiple photovoltaic cells can be divided into one area. The reason for dividing the photovoltaic panel into i division areas is that it is convenient to timely discover the differences in current, voltage and temperature changes in each area of ​​the photovoltaic panel, so as to timely discover the generation of heat spots on the photovoltaic panel;

[0051] The method for obtaining n current change rates is as follows:

[0052] The unit time is equally divided into n+1 continuous time intervals, and the current value of the photovoltaic panel is collected once in each time interval. It should be noted that the current value can be the start node, the end node, or any node from the start node to the end node of each time interval;

[0053] The current value in the next time interval is subtracted from the current value in the previous time interval to obtain the current difference in adjacent regions. The current difference is divided by the current value in the previous time interval to obtain a current change rate. This is repeated to obtain n current change rates.

[0054] The method for obtaining n voltage change rates is similar to the method for obtaining n current change rates, as follows:

[0055] The voltage value of the photovoltaic panel is collected once in each time interval;

[0056] Subtract the voltage value in the previous time interval from the voltage value in the next time interval to obtain the voltage difference between adjacent regions. Divide the voltage difference by the voltage value in the previous time interval to obtain a voltage change rate. This process is repeated to obtain n voltage change rates.

[0057] The temperature difference is obtained as follows:

[0058] Mark the temperature value collected in the last time interval of each divided area as the regional temperature value;

[0059] Subtract the regional temperature value from the regional temperature value of the adjacent divided area to obtain k temperature difference values. If the divided area corresponding to the regional temperature value has four adjacent divided areas, k is 4; if there are two adjacent divided areas, k is 2. The divided area corresponding to the regional temperature value is located at the corner of the photovoltaic panel, such as Figure 2 The filled portion of the rectangle shown.

[0060] The method for obtaining the local temperature difference threshold is as follows:

[0061] Pre-collecting threshold setting data per unit time before heat spots appear on photovoltaic panels, the threshold setting data including historical temperature difference characteristic data and local temperature difference thresholds corresponding to the historical temperature difference characteristic data; the historical temperature difference characteristic data including ambient temperature, light intensity, wind speed, solar radiation and panel material;

[0062] Each group of historical temperature difference feature data is used in the form of a vector as the input of a difference threshold setting model. The difference threshold setting model uses the historical local temperature difference threshold corresponding to a group of historical temperature difference feature data as the output, and the historical local temperature difference threshold actually corresponding to a group of historical temperature difference feature data as the prediction target. Minimizing the loss function value of the difference threshold setting model is used as the training target; when the loss function value of the difference threshold setting model is less than or equal to the preset target loss value, the training is stopped to obtain the difference threshold setting model, and the difference threshold setting model is one of the models such as SVM regression, random forest regression or neural network regression; the threshold setting data collected in real time is input into the trained difference threshold setting model to obtain the local temperature difference threshold.

[0063] The loss function value is mean square error;

[0064] Loss function formula Minimize as the goal to train the model, the loss function The model loss function value is set for the difference threshold, x is the historical temperature difference feature data group number; m is the number of historical temperature difference feature data groups; is the local temperature difference threshold predicted by the x-th group of historical temperature difference feature data, is the historical local temperature difference threshold actually corresponding to the x-th group of historical temperature difference feature data.

[0065] In order to timely detect and treat heat spots, the existing local temperature difference threshold is usually set at a lower local temperature difference threshold, which will increase false alarms and maintenance workload; in order to reduce the maintenance workload, the local temperature difference threshold is correspondingly increased, making it difficult to balance timely detection and maintenance workload; therefore, determining an appropriate local temperature difference threshold can reduce the maintenance workload while also timely detecting heat spots and improving the efficiency of treating heat spots.

[0066] In the process of setting the local temperature difference threshold, the inventor found that the ambient temperature, light intensity, wind speed, solar radiation and panel material are closely related to the local temperature difference threshold. By setting the appropriate local temperature difference threshold in real time according to the actual situation, the warning sensitivity can be improved while also improving the warning accuracy.

[0067] Among them, the ambient temperature refers to the temperature around the photovoltaic panel, which will directly affect the operating temperature of the photovoltaic panel. Generally speaking, when the ambient temperature is high, the temperature of the photovoltaic panel will also increase accordingly. Therefore, when the ambient temperature is high, the local temperature difference threshold needs to be adjusted to a smaller value accordingly, so as to more sensitively detect local temperature anomalies.

[0068] Light intensity refers to the intensity of solar radiation on the photovoltaic panel, which can reflect seasonal changes. It will directly affect the power generation efficiency and temperature of the photovoltaic panel. When the light intensity is high, the temperature of the photovoltaic panel may increase. Therefore, under high light intensity, the local temperature difference threshold needs to be adjusted to a smaller value accordingly to more accurately determine whether there is a local temperature anomaly.

[0069] Wind speed refers to the velocity of wind, which affects the heat dissipation effect of photovoltaic panels. When the wind speed is high, the heat dissipation effect of photovoltaic panels will be enhanced, thereby reducing the panel temperature. Therefore, under high wind speed, the local temperature difference threshold needs to be adjusted to a larger value accordingly to avoid misjudging local temperature anomalies.

[0070] Solar radiation refers to the radiation intensity of sunlight, which directly affects the power generation efficiency and temperature of photovoltaic panels. When the solar radiation is high, the temperature of the photovoltaic panel may increase. Therefore, under high solar radiation, the local temperature difference threshold needs to be adjusted to a smaller value accordingly to more accurately determine whether there is a local temperature anomaly.

[0071] Different panel materials have different thermal conductivities and heat capacities, which will affect the thermal conduction and heat storage capacity of the photovoltaic panel. Therefore, the characteristics of the panel material also need to be taken into account to determine the appropriate local temperature difference threshold.

[0072] It should be noted that the above historical temperature difference characteristic data, in actual application, needs to be different according to factors such as the design and manufacturing process of the photovoltaic panel. Therefore, the specific historical temperature difference characteristic data needs to be adjusted and optimized according to the specific situation to ensure that the local temperature anomaly of the photovoltaic panel can be accurately judged.

[0073] The data analysis module is used to input real-time feature data into a pre-built fault prediction model to obtain the operating status of the photovoltaic panel after a unit time; the operating status includes the absence of a fault and the presence of a fault, and the presence of a fault includes local blocking damage and temperature anomaly.

[0074] When a certain divided area of ​​the photovoltaic panel is partially blocked or damaged, the current flow in the divided area will be blocked or interrupted, which will cause the battery cells in the divided area to generate excessive heat during operation. When a certain divided area is partially blocked or damaged, the current output of the divided area will decrease or be interrupted, forming heat spots; the current value is monitored in real time by installing a current sensor or shunt in the photovoltaic cell circuit.

[0075] When a certain area of ​​the photovoltaic panel is partially blocked or damaged, the voltage output of the area will drop or be interrupted. The voltage value is obtained by installing a voltage sensor or a voltage divider in the photovoltaic cell circuit.

[0076] The method for pre-building the fault prediction model is as follows:

[0077] The unit time may be any time between 3 minutes and 10 minutes;

[0078] The operation data per unit time during the operation stage of the photovoltaic panel is collected in advance, and the operation data includes historical characteristic data and the operation status corresponding to the historical characteristic data; among them, when the operation status is a fault, the corresponding historical characteristic data is the historical characteristic data per unit time before the photovoltaic panel has heat spots during the operation stage; the historical characteristic data is converted into a characteristic vector as follows: , The corresponding operating status is ; is the first set of eigenvectors; is the current change rate of the nth time interval in the first set of eigenvectors; is the voltage change rate of the nth time interval in the first set of eigenvectors; It is the temperature difference between a regional temperature value in the first set of eigenvectors and the regional temperature value of the divided region in the kth direction.

[0079] Exemplarily, if k is 1, the adjacent divided area is located on the left side of the divided area, if k is 2, it is located on the right side, if k is 3, it is located above, and if k is 4, it is located below.

[0080] is the local temperature difference threshold corresponding to the temperature difference in the first set of eigenvectors; include , and , To avoid any faults, It is a local obstruction damage. The temperature is abnormal.

[0081] The operating data is divided into a training set and a test set, and a first classifier is constructed. The historical feature data in the training set is used as input data, and the operating status in the training set is used as output data. The first classifier is trained to obtain an initial first classifier, and the initial first classifier is tested using the test set. The first classifier that meets the preset accuracy is output as a fault prediction model. The fault prediction model is preferably a naive Bayes model or a support vector machine model.

[0082] The second data acquisition module is used to collect real-time image data of the corresponding photovoltaic panel according to the temperature anomaly; specifically, the data is collected through an optical camera set next to the photovoltaic panel.

[0083] It should be noted that, in this embodiment, the optical camera operates only when the fault prediction model outputs an abnormal temperature, thereby reducing the energy consumption of photovoltaic panel monitoring and the service life of the optical camera.

[0084] The fault diagnosis module is used to input the collected image data into a pre-built fault diagnosis model to obtain the fault type, which includes internal abnormality and external abnormality.

[0085] External anomalies include cracks in the photovoltaic panel, contaminants in the photovoltaic panel, and shadows in the photovoltaic panel; internal anomalies include uneven battery cells and battery cell aging;

[0086] In the practice of identifying the types of photovoltaic panel failures, there may be some non-uniformity in the production of solar cells when photovoltaic panels are manufactured. For example, there may be differences in the thickness, material quality or manufacturing process of the solar cells. These non-uniformities will cause some solar cells to generate excessive heat during operation and form heat spots. As the use time of photovoltaic panels increases, the solar cells may have problems due to aging. Aging causes the electrical conductivity of the solar cells to decrease, thereby generating excessive heat and leading to the formation of heat spots.

[0087] The construction method of the fault diagnosis model is as follows:

[0088] Obtain r groups of data, where r is a positive integer greater than 1, and the data includes historical image data and historical fault types. The historical image data and historical fault types are used as sample sets, and the sample sets are divided into training sets and test sets. A second classifier is constructed, and the historical image data in the training set is used as input data, and the historical fault types in the training set are used as output data. The second classifier is trained to obtain an initial second classifier, and the initial second classifier is tested using the test set, and a second classifier that meets a preset accuracy is output as a fault diagnosis model, and the fault diagnosis model is a convolutional neural network, a recurrent neural network, or other suitable models;

[0089] Before training the fault diagnosis model, each historical image data is manually labeled in advance for the corresponding historical fault type. The historical fault types include cracks in the photovoltaic panel, contaminants in the photovoltaic panel, shadows in the photovoltaic panel, and internal abnormalities. The historical fault types are first converted into digital labels. For example, cracks in the photovoltaic panel are marked as 1, contaminants in the photovoltaic panel are marked as 2, and shadows in the photovoltaic panel are marked as 3. The absence of cracks in the photovoltaic panel, the absence of contaminants in the photovoltaic panel, and the absence of shadows in the photovoltaic panel are all marked as 0, i.e., internal abnormalities.

[0090] This embodiment divides the photovoltaic panel into i divided areas in advance, collects real-time feature data of each divided area in real time, inputs the collected real-time feature data into a pre-trained fault prediction model, and obtains in real time whether each divided area has a fault after a unit time. If a fault exists, the specific fault is identified at the same time, and the cause of the fault is located, the faulty part is repaired or replaced in time, heat spots are discovered in time and the efficiency of treating heat spots is improved, so as to ensure the normal operation and power generation efficiency of the photovoltaic panel power generation components.

[0091] Secondly, in the process of setting the local temperature difference threshold, comprehensive consideration of multiple factors can more comprehensively consider the impact of multiple factors on the temperature of the photovoltaic panel, and improve the accuracy of judging local temperature anomalies; real-time setting of the corresponding local temperature difference threshold can achieve a balance between timely detection and maintenance workload; determining the appropriate local temperature difference threshold can reduce the maintenance workload while also timely detecting heat spots and improving the efficiency of treating heat spots, effectively ensuring the normal operation, power generation efficiency and life of photovoltaic panel power generation components. Example 2

[0092] like Figure 3 As shown, this embodiment provides a photovoltaic power generation fault prediction and diagnosis method based on data drive, including:

[0093] The photovoltaic panels are divided into i divided areas in advance, and real-time characteristic data of the photovoltaic panels in the i divided areas per unit time are collected;

[0094] Input the real-time characteristic data into the pre-built fault prediction model to obtain the operating status of the photovoltaic panel after a unit time; the operating status includes the absence of faults and the presence of faults, and the presence of faults includes local blocking damage and temperature abnormality;

[0095] Collect real-time image data of corresponding photovoltaic panels according to temperature anomalies;

[0096] The collected image data is input into the pre-built fault diagnosis model to obtain the fault type, which includes internal abnormality and external abnormality.

[0097] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A data-driven photovoltaic power generation fault prediction and diagnosis method, characterized in that: include: The photovoltaic panels are divided into i divided areas in advance, and real-time characteristic data of the photovoltaic panels in the i divided areas per unit time are collected; Input the real-time characteristic data into the pre-built fault prediction model to obtain the operating status of the photovoltaic panel after a unit time; the operating status includes the absence of faults and the presence of faults, and the presence of faults includes local blocking damage and temperature abnormality; The method for pre-constructing the fault prediction model includes: Pre-collecting operation data per unit time during the operation stage of the photovoltaic panel, the operation data including historical characteristic data and operation status corresponding to the historical characteristic data; wherein, when the operation status is that there is a fault, the corresponding historical characteristic data is the historical characteristic data per unit time before the photovoltaic panel has heat spots during the operation stage; Convert historical feature data into feature vectors as follows: , The corresponding operating status is ; is the first set of eigenvectors; is the current change rate of the nth time interval in the first set of eigenvectors; is the voltage change rate of the nth time interval in the first set of eigenvectors; is the temperature difference between a temperature value of a region in the first set of eigenvectors and the temperature value of the region divided in the kth direction; is the local temperature difference threshold corresponding to the temperature difference in the first set of eigenvectors; include , and , To avoid any faults, It is a local obstruction damage. Abnormal temperature; Divide the operation data into a training set and a test set, construct a first classifier, use the historical feature data in the training set as input data, use the operation status in the training set as output data, train the first classifier to obtain an initial first classifier, use the test set to test the initial first classifier, and output a first classifier that meets a preset accuracy as a fault prediction model, wherein the fault prediction model is a naive Bayes model or a support vector machine model; Collect real-time image data of corresponding photovoltaic panels according to temperature anomalies; Input the collected image data into the pre-built fault diagnosis model to obtain the fault type, which includes internal abnormality and external abnormality; The method for constructing the fault diagnosis model comprises: Obtain r groups of data, where r is a positive integer greater than 1, and the data includes historical image data and historical fault types. The historical image data and historical fault types are used as sample sets, and the sample sets are divided into training sets and test sets. A second classifier is constructed, and the historical image data in the training set is used as input data, and the historical fault types in the training set are used as output data. The second classifier is trained to obtain an initial second classifier, and the initial second classifier is tested using the test set. The second classifier that meets the preset accuracy is output as a fault diagnosis model, and the fault diagnosis model is one of a convolutional neural network and a recurrent neural network.

2. The photovoltaic power generation fault prediction and diagnosis method based on data drive according to claim 1 is characterized by: The real-time characteristic data includes n current change rates, n voltage change rates, temperature differences, and local temperature difference thresholds corresponding to the temperature differences within a unit time.

3. The data-driven photovoltaic power generation fault prediction and diagnosis method according to claim 2, characterized in that: The n current change rate acquisition methods include: The unit time is equally divided into n+1 continuous time intervals, and the current value of the photovoltaic panel is collected once in each time interval; The current value in the next time interval is subtracted from the current value in the previous time interval to obtain the current difference between adjacent regions, and the current difference is divided by the current value in the previous time interval to obtain a current change rate, and so on to obtain n current change rates.

4. The photovoltaic power generation fault prediction and diagnosis method based on data drive according to claim 2 is characterized by: The n voltage change rate acquisition methods include: The voltage value of the photovoltaic panel is collected once in each time interval; The voltage value in the next time interval is subtracted from the voltage value in the previous time interval to obtain the voltage difference between adjacent regions, and the voltage difference is divided by the voltage value in the previous time interval to obtain a voltage change rate, and so on to obtain n voltage change rates.

5. The photovoltaic power generation fault prediction and diagnosis method based on data drive according to claim 2 is characterized by: The temperature difference acquisition method comprises: Mark the temperature value collected in the last time interval of each divided area as the regional temperature value; Subtract the regional temperature value of the adjacent divided area from the regional temperature value to obtain k temperature difference values.

6. The data-driven photovoltaic power generation fault prediction and diagnosis method according to claim 2, characterized in that: The local temperature difference threshold acquisition method comprises: Pre-collecting the threshold setting data of the unit time before the photovoltaic panel has heat spots, the threshold setting data including the historical temperature difference characteristic data and the historical local temperature difference threshold value corresponding to the historical temperature difference characteristic data; the historical temperature difference characteristic data includes the ambient temperature, light intensity, wind speed, solar radiation and panel material; Each group of historical temperature difference feature data is used as an input of a difference threshold setting model in the form of a vector. The difference threshold setting model uses a historical local temperature difference threshold corresponding to a group of historical temperature difference feature data as an output, uses a historical local temperature difference threshold actually corresponding to a group of historical temperature difference feature data as a prediction target, and uses minimizing the difference threshold setting model loss function value as a training target; when the difference threshold setting model loss function value is less than or equal to a preset target loss value, the training is stopped to obtain a difference threshold setting model, and the difference threshold setting model is one of an SVM regression model, a random forest regression model, or a neural network regression model; The threshold setting data collected in real time is input into the trained difference threshold setting model to obtain the local temperature difference threshold.

7. The data-driven photovoltaic power generation fault prediction and diagnosis method according to claim 1, characterized in that: The external anomalies include cracks in the photovoltaic panel, contaminants in the photovoltaic panel and shadows in the photovoltaic panel; the internal anomalies include uneven battery cells and aging of battery cells.

8. A data-driven photovoltaic power generation fault prediction and diagnosis system, characterized in that: A data-driven photovoltaic power generation fault prediction and diagnosis method for implementing any one of claims 1 to 7, comprising: The first data acquisition module divides the photovoltaic panel into i divided areas in advance, and collects the real-time characteristic data of the photovoltaic panels in the i divided areas per unit time; The data analysis module is used to input the real-time characteristic data into the pre-built fault prediction model to obtain the operating status of the photovoltaic panel after a unit time; the operating status includes the absence of faults and the presence of faults, and the presence of faults includes local blocking damage and temperature abnormality; The second data acquisition module is used to collect real-time image data of the corresponding photovoltaic panel according to the temperature anomaly; The fault diagnosis module is used to input the collected image data into a pre-built fault diagnosis model to obtain the fault type, which includes internal abnormality and external abnormality.

Citation Information

Patent Citations

  • Photovoltaic array monitoring system with fault diagnosis and power prediction functions

    CN111460644A

  • Fault recording system for photovoltaic system and configuration self-adaption method

    CN114726088A

  • Photovoltaic array fault diagnosis method based on ML-RFKNN in sand and dust weather

    CN117235617A

  • System for monitoring failure forecasts of solar module

    KR102080164B1