Photovoltaic hot spot detection method, storage medium and equipment

Abstract

The invention discloses a photovoltaic hot spot detection method, a storage medium and equipment, and the method comprises the steps: pasting a fiber Bragg grating on each solar cell in a photovoltaic module, and collecting a plurality of full-spectrum signals; determining the number of harmonic peak extreme values of the fiber Bragg grating array according to each group of full-spectrum signals through a genetic algorithm, dividing harmonic peaks, and calculating the temperature change value of the solar cell according to the harmonic peak wavelength value of the fiber Bragg grating in combination with the calibrated sensitivity parameter; establishing a hot spot detection model taking the temperature change value of the solar cell as input and the working condition type of the photovoltaic module as a label, and optimizing an activation function of the hot spot detection model through a particle swarm optimization algorithm to obtain an optimal hot spot detection model; full-spectrum signals are collected in real time through the optical fiber Bragg array, temperature change values of the corresponding solar cells are calculated and input into the optimal hot spot detection model, and the working condition type of the photovoltaic module is predicted. According to the invention, the detection accuracy of the photovoltaic hot spots is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic thermal management, and specifically, to a method for detecting photovoltaic hot spots, a storage medium, and a device. Background Art

[0002] Due to the long-term service of solar cells in harsh environments, performance degradation and safety hazards often occur in actual working conditions. Currently, photovoltaic modules generate hot spot effects due to shadow occlusion, cracking and defects, animal traces, uneven light concentration, and aging. It is worth noting that photovoltaic hot spots will inhibit the output power and cause economic losses to photovoltaic power stations. In addition, most seriously, photovoltaic hot spots will extend from the solar cells with hot spots to the surrounding solar cells, thereby causing irreversible damage to photovoltaic power stations.

[0003] The existing methods for detecting the temperature and hot spots of photovoltaic modules are roughly divided into two categories: contact measurement based on digital temperature sensors such as thermocouples and thermal resistance sensors, and non-contact measurement based on infrared thermal imaging. Digital temperature sensors are usually installed on the back of solar panels to measure the temperature after calibration. Although electrical sensors have the characteristics of high-precision measurement and low cost, due to the size of the sensors, the surface temperature of solar panels cannot be measured, and a large number of sensors need to be arranged in photovoltaic power stations; while the accuracy of thermal imaging technology for measuring photovoltaic temperature is greatly affected by environmental and climatic factors. In addition, thermal imaging measurement has high costs, complex processing procedures, and low detection efficiency, and cannot monitor the distributed hot spots of photovoltaic arrays in real time. At the same time, the current photovoltaic hot spot evaluation system and feedback mechanism are relatively single, the evaluation standard is manual judgment, and the means for dealing with hot spots is to use diode disconnection or directly replace the photovoltaic panel, with high costs and complex processing. Summary of the Invention

[0004] Aiming at the problems existing in the prior art, the present invention provides a method for detecting photovoltaic hot spots, a storage medium, and a device, which perform distributed real-time accurate measurement on photovoltaic modules based on a fiber Bragg grating array, and use a hot spot detection model to detect hot spots on the surface of photovoltaic modules, improving the accuracy and speed of hot spot recognition; at the same time, the present invention cools the photovoltaic hot spot area based on hygroscopic hydrogel, improving the power generation efficiency of photovoltaic modules and reducing costs.

[0005] To achieve the above technical objectives, the present invention adopts the following technical solutions: A method for detecting photovoltaic hot spots specifically includes the following steps:

[0006] Step S1: Paste fiber Bragg gratings on each solar cell in a photovoltaic module to form a fiber Bragg grating array, and use the fiber Bragg array to collect a plurality of full-spectrum signals;

[0007] Step S2: Determine the number of resonant peak extrema of the fiber Bragg grating array for each group of full-spectrum signals, divide the resonant peaks, perform one-dimensional Gaussian filtering on the divided resonant peaks of the fiber Bragg gratings, calculate the full-width at half-maximum of the resonant peaks of each filtered fiber Bragg grating, fit the wavelength values of the resonant peaks of each fiber Bragg grating, and calculate the temperature change value of the solar cell in combination with the calibrated sensitivity parameter;

[0008] Step S3: Establish a hot spot detection model with the temperature change value of the solar cell as the input and the photovoltaic module working condition type as the label, and optimize the activation function of the hot spot detection model through the particle swarm optimization algorithm to obtain the optimal hot spot detection model;

[0009] Step S4: Collect full-spectrum signals in real time through the fiber Bragg array, calculate the corresponding temperature change value of the solar cell through Step S2, input it into the optimal hot spot detection model, and predict the working condition type of the photovoltaic module.

[0010] Further, the specific process of determining the number of resonant peak extrema of the fiber Bragg grating array for each group of full-spectrum signals in Step S2 is as follows:

[0011] Step S2.1: Randomly binary encode each point in each group of full-spectrum signals to obtain an encoded sequence, where 0 represents a non-extreme point and 1 represents an extreme point;

[0012] Step S2.2: Determine the number of resonant peak extrema for each group according to the judgment function where n represents the number of points in the encoded sequence, y a represents the binary encoding of the a-th point, and g(a) represents the judgment function. If the spectral signal value of the a-th point is larger than the spectral signal values of the two adjacent points, the a-th point is a valid maximum value, g(a) = 1; otherwise, g(a) = 0;

[0013] Step S2.3: Perform selection, crossover, and mutation operations on each group of encoded sequences, update the encoded sequences, and repeat Step S2.2;

[0014] Step S2.4: Repeat Step S2.3 until the maximum number of iterations is reached to obtain the optimal number of resonant peak extrema.

[0015] Further, the selection operation in Step S2.3 is performed by roulette, and the selection probability of roulette is determined by the ratio of the number of resonant peak extrema of each group to the number of resonant peak extrema of all groups;

[0016] The probability of the crossover operation is:

[0017] The probability of the mutation operation is:

[0018] Among them, P Cmin and P Cmax respectively represent the minimum and maximum values of the crossover rate, and P mmin and P mmax respectively represent the minimum and maximum values of the mutation rate, and t is the ratio of the current iteration number to the maximum iteration number.

[0019] Furthermore, the calculation process of the temperature change value ΔT of the solar cell in step S2 is as follows:

[0020]

[0021] Among them, λ B represents the resonant peak wavelength value of the fiber Bragg grating, Δλ B represents the change value of the resonant peak wavelength of the fiber Bragg grating, α 0 represents the thermal expansion coefficient of the calibrated fiber Bragg grating, and β 0 represents the thermo-optic coefficient of the calibrated fiber Bragg grating.

[0022] Furthermore, step S3 includes the following sub-steps:

[0023] Step S3.1: Connect the hidden layer with 8 neurons and the hidden layer with 4 neurons through a fully connected layer to form a hot spot detection model, and use the hard-sigmoid function as the activation function;

[0024] Step S3.2: Input the infrared imaging pictures of the photovoltaic modules into the EfficientNet introduced with the self-attention mechanism to extract feature vectors, and use the optimized K-Means clustering algorithm to classify the working conditions of the photovoltaic modules as the labels of the hot spot detection model;

[0025] Step S3.3: Use the temperature change value of the solar cell as the input of the hot spot detection model, and use the particle swarm optimization algorithm to optimize the slope and bias of the activation function of the hot spot detection model until the mean square error between the predicted working condition type of the photovoltaic module by the hot spot detection model and the label is minimized, and obtain the optimal hot spot detection model.

[0026] Furthermore, step S3.2 includes the following sub-steps:

[0027] Step S3.2.1: Input the infrared imaging pictures of several photovoltaic modules into the EfficientNet, calculate the attention scores through the self-attention mechanism, then perform softmax normalization to obtain the attention weights, and perform weighted summation of the attention scores and the attention weights to obtain the feature vectors of the infrared imaging pictures;

[0028] Step S3.2.2: Map all the extracted feature vectors to a high-dimensional feature space and randomly select four clustering centroids;

[0029] Step S3.2.3: Calculate the Euclidean distance from the feature vectors to each clustering centroid, and assign each feature vector to the cluster where the clustering centroid with the minimum Euclidean distance is located;

[0030] Step S3.2.4: Update the clustering centroid of each cluster through the Gaussian kernel function where, x i and x j respectively represent any two feature vectors in each cluster C j , K(x i , x j ) represents the Gaussian kernel function, ||x i - x j || represents the Euclidean distance between x i and x j , and σ represents the kernel bandwidth parameter;

[0031] Step S3.2.5: Repeat Step S3.2.3 - Step S3.2.4 until the clustering centroids of each cluster no longer change, and obtain the clustering result of the photovoltaic module operating conditions types, where the photovoltaic module operating conditions types are divided into: normal, defective, reversible hot spot, and irreversible hot spot.

[0032] Furthermore, Step S3.3 includes the following sub-steps:

[0033] Step S3.3.1: In the particle swarm optimization algorithm, use the slope and bias of the activation function of the hot spot detection model as the particle individuals, randomly initialize the particle individuals, set the particle population and the maximum number of iterations, and use the mean square error between the predicted photovoltaic module operating conditions type of the hot spot detection model and the label as the fitness function;

[0034] Step S3.3.2: Use the particle individuals to determine the activation function of the hot spot detection model, input the temperature change value of the solar cell into the hot spot detection model, predict the photovoltaic module operating conditions type, and calculate the fitness function;

[0035] Step S3.3.3: In the next iteration, retain the particle individual with the minimum fitness function value, and update the remaining particle individuals by updating the positions and velocities of the particle individuals;

[0036] Step S3.3.4: Repeat Step S3.3.2 - Step S3.3.3 until the maximum number of iterations is reached, and use the particle individual with the minimum fitness function value as the optimal slope and bias of the activation function to obtain the optimal hot spot detection model.

[0037] Furthermore, if the predicted working condition of the photovoltaic component is a reversible hot spot, the obstructions on the solar cell corresponding to the fiber Bragg grating detection will be cleaned; if the predicted working condition of the photovoltaic component is an irreversible hot spot, a hygroscopic hydrogel will be pasted on the back of the solar cell corresponding to the fiber Bragg grating detection for cooling.

[0038] Furthermore, the present invention also provides a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the photovoltaic hot spot detection method.

[0039] Furthermore, the present invention also provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the photovoltaic hot spot detection method is implemented.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] (1) The photovoltaic hot spot detection method of the present invention is based on the distributed real-time precise measurement of photovoltaic modules by fiber Bragg grating arrays, and the full spectrum signal is used to determine the number of resonance peak extreme values ​​of the fiber Bragg grating through a genetic algorithm. By searching in the entire domain space, it can effectively cope with the complexity of the full spectrum signal. This global search capability is reflected in the processing of multiple resonance peak extreme values ​​through selection, crossover and mutation operations. For multiple resonance peak extreme values, the genetic algorithm can maintain a diverse population so that different full spectrum signal sequences can converge to different peak values, thereby accurately determining the number of function extreme values ​​and having a certain tolerance for signal noise and inaccuracy, achieving accurate division of resonance peaks, and providing a reliable basis for subsequent prediction of photovoltaic module working condition types;

[0042] (2) The photovoltaic hot spot detection method of the present invention uses a hot spot detection model to detect hot spots on the surface of a photovoltaic module. Optimizing the activation function of the hot spot detection model through a particle swarm optimization algorithm can avoid missing activation function parameter settings that are more suitable for the hot spot detection model, and more fully explore the effective information in the hot spot data, such as the shape, size, temperature distribution and other differences of the hot spots, to meet the requirements of efficient global search capabilities, fast convergence and high integration. At the same time, the present invention divides the working conditions of the photovoltaic modules according to the infrared imaging pictures of the photovoltaic modules, which are used as labels of the hot spot detection model. Such labels can intuitively reflect the temperature distribution characteristics of the photovoltaic panels and the temperature gradients between the hot spots and the surrounding normal areas through non-contact detection, so that the model can improve the detection accuracy of different types of hot spots by learning the thermal feature patterns in the infrared imaging pictures.

[0043] In summary, the accuracy and speed of hot spot identification are improved by using an optimized hot spot detection model.

[0044] (3) The present invention cools the photovoltaic hot spot area based on the hygroscopic hydrogel, improves the power generation efficiency of the photovoltaic module, reduces costs, and avoids the risk of fire. Description of the Drawings

[0045] Figure 1 is a schematic diagram of the photovoltaic hot spot detection device in the present invention;

[0046] Figure 2 is a flowchart of the detection method for the photovoltaic hot spot of the present invention. Detailed Embodiments

[0047] The technical solutions of the present invention will be further explained below with reference to the drawings.

[0048] As Figure 1 is a schematic diagram of the photovoltaic hot spot detection device in the present invention. The photovoltaic hot spot detection device includes: a photovoltaic module 1, an optical fiber temperature sensing array 2, a hygroscopic hydrogel 3, a grating demodulator 4, and a host computer 5. The photovoltaic module 1 is an integrated polycrystalline silicon solar panel, and an external resistor forms a circuit. The optical fiber temperature sensing array 2 is a femtosecond laser direct-written fiber Bragg grating array, and a calibrated fiber Bragg grating is adhesively attached to the area to be measured of each solar panel through thermal conductive silicone grease. The hygroscopic hydrogel 3 is a polyacrylamide / calcium chloride hygroscopic hydrogel prepared by a self-crosslinking process, which is pasted on the back of the hot spot area and can effectively cool the hot spot area. The grating demodulator 4 is an integrated electronic device of a tunable laser, a multi-channel photodetector, and a customized high-speed gate array for driving the laser and processing the photoelectric detection signal. It is connected to the host computer 5 for collecting and processing spectral signals, demodulating temperature sensing signals, and establishing a photovoltaic hot spot detection model, and finally realizing the identification and feedback of the photovoltaic hot spot.

[0049] As Figure 2 is a flowchart of the detection method for the photovoltaic hot spot of the present invention. The detection method for the photovoltaic hot spot specifically includes the following steps:

[0050] Step S1: Paste a temperature-calibrated fiber Bragg grating on each solar cell in the photovoltaic module to form a fiber Bragg grating array. Specifically, the temperature-calibrated fiber Bragg grating is fixed on the surface grid line area of the solar cell using thermal conductive silicone grease to avoid affecting the photovoltaic power generation efficiency. The fiber Bragg array is used to collect a number of full-spectrum signals to achieve distributed measurement of the photovoltaic module.

[0051] Step S2: Determine the number of resonant peak extrema of the fiber Bragg grating array for each set of full-spectrum signals. By searching in the entire domain space, it can effectively handle the complexity of the full-spectrum signals. This global search ability is reflected in the processing of multiple resonant peak extrema through selection, crossover, and mutation operations. For signals with multiple resonant peak extrema, the genetic algorithm can maintain a diverse population, enabling different full-spectrum signal sequences to converge to different peaks, thereby accurately determining the number of function extrema and having a certain tolerance for signal noise and inaccuracies, achieving precise division of the resonant peaks and providing a reliable basis for subsequent prediction of the operating conditions of photovoltaic modules. Divide the full spectrum according to the number of resonant peaks, perform one-dimensional Gaussian filtering on the divided resonant peaks of the fiber Bragg grating, smooth and reduce the noise of the resonant peaks, calculate the full-width at half-maximum of the resonant peaks of each filtered fiber Bragg grating, fit the wavelength values of the resonant peaks of each fiber Bragg grating, and calculate the temperature change value of the solar cell in combination with the calibrated sensitivity parameters:

[0052]

[0053] where, λ B represents the resonant peak wavelength value of the fiber Bragg grating, Δλ B represents the change value of the resonant peak wavelength of the fiber Bragg grating, α 0 represents the thermal expansion coefficient of the calibrated fiber Bragg grating, β 0 represents the thermo-optic coefficient of the calibrated fiber Bragg grating, and α 0 and β 0 constitute the sensitivity parameter;

[0054] The specific process of determining the number of resonant peak extrema of the fiber Bragg grating array for each set of full-spectrum signals in the present invention is as follows:

[0055] Step S2.1: Randomly perform binary encoding on each point in each set of full-spectrum signals to obtain an encoding sequence, where 0 represents a non-extreme point and 1 represents an extreme point;

[0056] Step S2.2: Determine the number of resonant peak extrema for each set according to the judgment function where, n represents the number of points in the encoding sequence, y a represents the binary encoding of the a-th point, and g(a) represents the judgment function. If the spectral signal value of the a-th point is larger than the spectral signal values of the two adjacent points, the a-th point is a valid maximum value, g(a)=1; otherwise, g(a)=0;

[0057] Step S2.3: Select, crossover, and mutate each group of coding sequences to update the coding sequences, and repeat Step S2.2; specifically, the selection operation is performed using the roulette wheel method, and the selection probability of the roulette wheel is determined by the ratio of the number of resonance peak extrema in each group to the number of resonance peak extrema in all groups; the probability of the crossover operation is: The probability of the mutation operation is:

[0058] where P Cmin and P Cmax represent the minimum and maximum values of the crossover rate respectively, P mmin and P mmax represent the minimum and maximum values of the mutation rate respectively, t is the ratio of the current iteration number to the maximum iteration number. At the initial stage of genetic iteration, a higher crossover rate and a lower mutation rate can be set, which helps the genetic algorithm quickly explore various regions of the solution space and increase the chance of finding all extrema; as the number of genetic iterations increases, the sine function causes the crossover rate to gradually decrease and the mutation rate to gradually increase, which helps to reduce the destruction of the excellent individual structure caused by excessive crossover and mutation, and helps to jump out of the local optimal trap to determine the possible number of resonance peaks under the full spectrum.

[0059] Step S2.4: Repeat Step S2.3 until the maximum iteration number is reached to obtain the optimal number of resonance peak extrema.

[0060] Step S3: Establish a hot spot detection model with the temperature change value of the solar cell as the input and the photovoltaic module working condition type as the label, and optimize the activation function of the hot spot detection model through the particle swarm optimization algorithm to obtain the optimal hot spot detection model. The particle swarm optimization algorithm can avoid missing the activation function parameter settings more suitable for the hot spot detection model, and more fully excavate the effective information in the hot spot data such as the differences in the shape, size, and temperature distribution of the hot spots, meeting the requirements of high efficient global search ability, fast convergence, and high integration; it includes the following sub-steps:

[0061] Step S3.1: Connect the hidden layer with 8 neurons and the hidden layer with 4 neurons through a fully connected layer to form a hot spot detection model, which can quickly and accurately perform working condition analysis and hot spot detection on the photovoltaic module using the fiber optic distributed temperature data, with the hard-sigmoid function as the activation function;

[0062] Step S3.2: Input the infrared imaging pictures of photovoltaic modules into EfficientNet with self-attention mechanism to extract feature vectors, and use the optimized K-Means clustering algorithm to classify the working conditions of photovoltaic modules as the labels of the hot spot detection model. In the present invention, the infrared thermal image feature vectors are extracted through a pre-trained model for clustering, rather than directly converting the image into a one-dimensional vector, which avoids mainly focusing on the pixel-level information of the image due to the lack of a complex feature learning process, reduces the serious influence of factors such as image size, rotation, and illumination, and improves the accuracy of clustering; it includes the following sub-steps:

[0063] Step S3.2.1: Input the infrared imaging pictures of several photovoltaic modules into EfficientNet, and calculate the attention scores through the self-attention mechanism Then perform softmax normalization to obtain the attention weight attention = softmax(score(q i , k j ))v j , and perform weighted summation of the attention scores and the attention weights to obtain the feature vectors of the infrared imaging pictures, where The transpose of the i-th query vector, k j is the j-th key vector, and v j is the j-th value vector;

[0064] Step S3.2.2: Map all the extracted feature vectors to a high-dimensional feature space to increase the probability of distinguishing four working conditions of photovoltaic panels in the high-dimensional space, and randomly select four clustering centroids;

[0065] Step S3.2.3: Calculate the Euclidean distance from the feature vectors to each clustering centroid, and assign each feature vector to the cluster where the clustering centroid with the smallest Euclidean distance is located;

[0066] Step S3.2.4: Update each cluster's clustering centroid through the Gaussian kernel function where x i and x j respectively represent any two feature vectors in each cluster C j , and K(x i , x j ) represents the Gaussian kernel function, ||x i - x j || represents x i and x jThe Euclidean distance between them, and σ represents the kernel bandwidth parameter; compared with ordinary K-Means clustering, the K-Means clustering algorithm using the Gaussian kernel function has the ability to process non-linear data, flexible similarity measurement, and better generalization ability, and is more accurate and comprehensive in extracting the characteristics of the temperature spatial distribution of infrared images, and avoids the influence of various noises by virtue of the smoothness of the kernel function.

[0067] Step S3.2.5: Repeat Step S3.2.3 - Step S3.2.4 until the clustering centroids of each cluster no longer change, and obtain the clustering results of the photovoltaic module operating conditions types, where the photovoltaic module operating conditions types are divided into: normal, defective, reversible hot spot, and irreversible hot spot.

[0068] Step S3.3: Use the temperature change value of the solar cell as the input of the hot spot detection model, and use the particle swarm optimization algorithm to optimize the slope and bias of the activation function of the hot spot detection model until the mean square error between the predicted photovoltaic module operating conditions type of the hot spot detection model and the label is minimized, and obtain the optimal hot spot detection model; it includes the following sub-steps:

[0069] Step S3.3.1: In the particle swarm optimization algorithm, use the slope and bias of the activation function of the hot spot detection model as the particle individuals, randomly initialize the particle individuals, set the particle population and the maximum number of iterations, and use the mean square error between the predicted photovoltaic module operating conditions type of the hot spot detection model and the label as the fitness function;

[0070] Step S3.3.2: Use the particle individuals to determine the activation function of the hot spot detection model, input the temperature change value of the solar cell into the hot spot detection model, predict the photovoltaic module operating conditions type, and calculate the fitness function;

[0071] Step S3.3.3: In the next iteration, retain the particle individual with the minimum fitness function value, and update the remaining particle individuals by updating the position and velocity of the particle individuals:

[0072] v i (t + 1) = w·v i (t) + c 1 ·r 1 ·(p i - x i (t)) + c 2 ·r 2 ·(p g - x i (t))

[0073] x i (t + 1) = x i (t) + v i (t + 1)

[0074] Among them, w is the inertia weight, and c 1 and c 2 are both learning factors, r 1 and r 2 are both random numbers, p i is the optimal position of the particle individual, and p g is the position of the global optimal particle individual.

[0075] Step S3.3.4: Repeat steps S3.3.2 - S3.3.3 until the maximum number of iterations is reached. Take the particle individual with the minimum fitness function value as the optimal slope and bias of the activation function to obtain the optimal hot spot detection model.

[0076] Step S4: Collect the full-spectrum signal in real time through the fiber Bragg array, calculate the temperature change value of the corresponding solar cell through step S2, and input it into the optimal hot spot detection model to predict the working condition type of the photovoltaic module.

[0077] By using the hot spot detection model of the present invention for hot spot identification, the accuracy and speed of hot spot identification are improved. Compared with classification models such as decision trees, support vector machines, and K-nearest neighbors, the hot spot detection model of the present invention is superior to other models with an identification accuracy of 97.2%.

[0078] In a technical solution of the present invention, if the predicted working condition of the photovoltaic module is a reversible hot spot, clean the obstruction on the solar cell detected by the corresponding fiber Bragg grating; if the predicted working condition of the photovoltaic module is an irreversible hot spot, paste a hygroscopic hydrogel on the back of the solar cell detected by the corresponding fiber Bragg grating for cooling.

[0079] In a technical solution of the present invention, a computer-readable storage medium is further provided, storing a computer program, and the computer program causes a computer to execute the detection method of the photovoltaic hot spot.

[0080] In a technical solution of the present invention, an electronic device is further provided, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the detection method of the photovoltaic hot spot is implemented.

[0081] In the embodiments disclosed in the present application, the computer storage medium may be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. The computer storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of the computer storage medium would include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0082] Those of ordinary skill in the art will appreciate that the units and algorithm steps of the examples described in connection with the embodiments disclosed in the present application can be implemented in electronic hardware or in a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled artisans may use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the present application.

[0083] The above are only the preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the concept of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art of this technology, several improvements and refinements made without departing from the principle of the present invention should be regarded as within the protection scope of the present invention.

Claims

1. A method for detecting photovoltaic hot spots, characterized in that: The specific steps include: Step S1, pasting a fiber Bragg grating on each solar cell in a photovoltaic module to form a fiber Bragg grating array, and using the fiber Bragg array to collect a number of full-spectrum signals; Step S2, using a genetic algorithm to determine the number of resonance peak extreme values ​​of the fiber Bragg grating array for each group of full-spectrum signals, dividing the resonance peaks, performing one-dimensional Gaussian filtering on the divided resonance peaks of the fiber Bragg gratings, calculating the half-peak width of the resonance peak of each filtered fiber Bragg grating, fitting the resonance peak wavelength value of each fiber Bragg grating, and calculating the temperature change value of the solar cell in combination with the calibrated sensitivity parameters; Step S3, establishing a hot spot detection model with the temperature change value of the solar cell as input and the photovoltaic module working condition type as a label, and optimizing the activation function of the hot spot detection model by using a particle swarm optimization algorithm to obtain an optimal hot spot detection model; Step S4, collect full spectrum signals in real time through the fiber Bragg array, calculate the temperature change value of the corresponding solar cell through step S2, input it into the optimal hot spot detection model, and predict the working condition type of the photovoltaic module.

2. A photovoltaic hot spot detection method according to claim 1, characterized in that: The specific process of determining the number of resonance peak extreme values ​​of the fiber Bragg grating array for each group of full spectrum signals in step S2 is as follows: Step S2.1, perform random binary encoding on each point in each group of full spectrum signals to obtain a coding sequence, where 0 represents a non-extreme point and 1 represents an extreme point; Step S2.2: Root judgment function determines the number of resonance peak extremes in each group Where n represents the number of points in the encoding sequence, y a represents the binary code of the a-th point, g(a) represents the judgment function, if the spectral signal value of the a-th point is larger than the two adjacent spectral signal values, the a-th point is a valid maximum value, g(a) = 1, otherwise, g(a) = 0; Step S2.3, perform selection, crossover and mutation operations on each group of coding sequences, update the coding sequences, and repeat step S2.2; Step S2.4, repeat step S2.3 until the maximum number of iterations is reached to obtain the optimal number of resonance peak extreme values.

3. A photovoltaic hot spot detection method according to claim 2, characterized in that: The selection operation in step S2.3 is performed in a roulette wheel manner, and the probability of selection in the roulette wheel is determined by the ratio of the number of resonance peak extreme values ​​of each group to the number of resonance peak extreme values ​​of all groups; The probability of the crossover operation is: The probability of the mutation operation is: Among them, P Cmin and P Cmax Respectively represent the minimum and maximum values ​​of the crossover rate, P mmin and P mmax They represent the minimum and maximum mutation rates respectively, and t is the ratio of the current number of iterations to the maximum number of iterations.

4. A photovoltaic hot spot detection method according to claim 3, characterized in that: The calculation process of the temperature change value ΔT of the solar cell in step S2 is: Among them, λ B Indicates the resonance peak wavelength of the fiber Bragg grating, Δλ B represents the change value of the resonance peak wavelength of the fiber Bragg grating, α0 represents the thermal expansion coefficient of the calibrated fiber Bragg grating, and β0 represents the thermo-optical coefficient of the calibrated fiber Bragg grating.

5. A photovoltaic hot spot detection method according to claim 1, characterized in that: Step S3 includes the following sub-steps: Step S3.1, connecting the hidden layer containing 8 neurons and the hidden layer containing 4 neurons into a hot spot detection model through a fully connected layer, using the hard-sigmoid function as the activation function; Step S3.2, using the infrared imaging picture of the photovoltaic module as input to extract feature vectors from the EfficientNet with the self-attention mechanism, and using the optimized K-Means clustering algorithm to classify the working conditions of the photovoltaic modules as labels for the hot spot detection model; Step S3.3, using the temperature change value of the solar cell as the input of the hot spot detection model, and using the particle swarm optimization algorithm to optimize the slope and bias of the activation function of the hot spot detection model until the mean square error between the photovoltaic module operating condition type predicted by the hot spot detection model and the label is minimized, and the optimal hot spot detection model is obtained.

6. A photovoltaic hot spot detection method according to claim 5, characterized in that: Step S3.2 includes the following sub-steps: Step S3.2.1, input infrared imaging pictures of several photovoltaic modules into EfficientNet, calculate the attention score through the self-attention mechanism, and then perform softmax normalization to obtain the attention weight, perform weighted summation of the attention score and the attention weight to obtain the feature vector of the infrared imaging picture; Step S3.2.2, map all extracted feature vectors to a high-dimensional feature space and randomly select four cluster centroids; Step S3.2.3, calculate the Euclidean distance from the feature vector to each cluster centroid, and assign each feature vector to the cluster where the cluster centroid with the smallest Euclidean distance is located; Step S3.2.4: Update the cluster centroid of each cluster using the Gaussian kernel function Among them, x i and x j Represents each cluster C j Any two eigenvectors in K(x i ,x j ) represents the Gaussian kernel function, ||x i -x j || represents x i and x j The Euclidean distance between them, σ represents the kernel bandwidth parameter; Step S3.2.5, repeat steps S3.2.3-S3.2.4 until the cluster centroid of each cluster no longer changes, and obtain the clustering results of the photovoltaic module operating condition types, which are divided into: normal, defective, reversible hot spots and irreversible hot spots.

7. A photovoltaic hot spot detection method according to claim 5, characterized in that: Step S3.3 includes the following sub-steps: Step S3.3.1, in the particle swarm optimization algorithm, the slope and bias of the activation function of the hot spot detection model are used as individual particles, the individual particles are randomly initialized, the particle population and the maximum number of iterations are set, and the mean square error between the working condition type of the photovoltaic module predicted by the hot spot detection model and the label is used as the fitness function; Step S3.3.2, using individual particles to determine the activation function of the hot spot detection model, inputting the temperature change value of the solar cell into the hot spot detection model, predicting the working condition type of the photovoltaic module, and calculating the fitness function; Step S3.3.3, in the next iteration, retain the particle individual with the smallest fitness function value, and update the remaining particle individuals by updating the position and velocity of the particle individual; Step S3.3.4, repeat steps S3.3.2-S3.3.3 until the maximum number of iterations is reached, and use the particle individual with the smallest fitness function value as the optimal slope and bias of the activation function to obtain the optimal hot spot detection model.

8. A photovoltaic hot spot detection method according to claim 1, characterized in that: If the predicted working condition of the photovoltaic module is a reversible hot spot, the obstructions on the solar cell corresponding to the fiber Bragg grating detection will be cleaned; if the predicted working condition of the photovoltaic module is an irreversible hot spot, a hygroscopic hydrogel will be pasted on the back of the solar cell corresponding to the fiber Bragg grating detection for cooling.

9. A computer-readable storage medium storing a computer program, characterized in that: The computer program enables a computer to execute the photovoltaic hot spot detection method according to any one of claims 1 to 8.

10. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the photovoltaic hot spot detection method according to any one of claims 1 to 8 is implemented.

Images