Dust monitoring system and method for photovoltaic module

Abstract

The embodiment of the invention discloses a dust monitoring system and method for a photovoltaic module. The method comprises the following steps: firstly, carrying out data preprocessing on illumination intensity values at a plurality of preset time points and output current values of a monitored photovoltaic module at the plurality of preset time points to obtain an illumination intensity time sequence input vector and an output current time sequence input vector; performing time sequence analysis on the illumination intensity time sequence input vector to obtain a sequence of illumination intensity local time sequence feature vectors; and integrating output current time sequence information expressed by the output current time sequence input vector into a sequence of the illumination intensity local time sequence feature vector to obtain an illumination intensity time sequence embedded output current time sequence feature vector, and finally, based on the illumination intensity time sequence embedded output current time sequence feature vector, obtaining an illumination intensity time sequence embedded output current time sequence feature vector. And determining whether the dust coverage rate of the monitored photovoltaic module exceeds a predetermined threshold. In this way, whether cleaning is needed or not can be intelligently judged.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of photovoltaic modules, and more specifically, to a dust monitoring system and method for a photovoltaic module. Background Art

[0002] A photovoltaic module is a device that converts solar energy into electrical energy, and it requires sufficient sunlight for effective operation. However, when a photovoltaic module is exposed to the outdoor environment for a long time, dust is likely to accumulate on its surface, blocking the incident light, reducing the output current of the photovoltaic module, and affecting its power generation efficiency. Therefore, regular cleaning of the photovoltaic module is an important measure to ensure its normal operation.

[0003] However, traditional cleaning methods usually require manual inspection and operation, which are time-consuming and laborious. Therefore, a dust monitoring solution for photovoltaic modules is expected to achieve real-time monitoring of the dust coverage of photovoltaic modules. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art, and provides a dust monitoring system and method for a photovoltaic module.

[0005] In a first aspect, embodiments of the present invention provide a dust monitoring system for a photovoltaic module, including:

[0006] A data acquisition module, configured to acquire the light intensity values at multiple predetermined time points within a predetermined time period and the output current values of the monitored photovoltaic module at the multiple predetermined time points;

[0007] A data preprocessing module, configured to perform data preprocessing on the light intensity values at the multiple predetermined time points and the output current values of the monitored photovoltaic module at the multiple predetermined time points to obtain a light intensity time series input vector and an output current time series input vector;

[0008] A time series analysis module, configured to perform time series analysis on the light intensity time series input vector to obtain a sequence of light intensity local time series feature vectors;

[0009] An information integration module, configured to integrate the output current time series information expressed by the output current time series input vector into the sequence of light intensity local time series feature vectors to obtain a light intensity time series embedded output current time series feature vector; and

[0010] A dust coverage analysis module, configured to determine whether the dust coverage of the monitored photovoltaic module exceeds a predetermined threshold based on the light intensity time series embedded output current time series feature vector.

[0011] In some possible embodiments, the data preprocessing module is configured to:

[0012] Arrange the light intensity values at the multiple predetermined time points and the output current values of the monitored photovoltaic modules at the multiple predetermined time points in the time dimension to form the light intensity time series input vector and the output current time series input vector respectively.

[0013] In some possible embodiments, the time series analysis module includes:

[0014] A vector splitting unit, configured to split the light intensity time series input vector to obtain a sequence of light intensity local time series input vectors; and

[0015] A light intensity time series feature extraction unit, configured to pass the sequence of light intensity local time series input vectors through a light intensity time series feature extractor based on a one-dimensional convolutional layer to obtain a sequence of light intensity local time series feature vectors.

[0016] In some possible embodiments, the light intensity time series feature extraction unit is configured to:

[0017] Use each layer of the light intensity time series feature extractor based on a one-dimensional convolutional layer to perform one-dimensional convolutional processing, pooling processing, and non-linear activation processing on the input data respectively during the forward pass of the layer, so as to output the sequence of light intensity local time series feature vectors by the last layer of the light intensity time series feature extractor based on a one-dimensional convolutional layer, where the input of the first layer of the light intensity time series feature extractor based on a one-dimensional convolutional layer is the sequence of light intensity local time series input vectors.

[0018] In some possible embodiments, the information integration module is configured to:

[0019] Pass the sequence of light intensity local time series feature vectors and the output current time series input vector through a feature embedding module to obtain the light intensity time series embedded output current time series feature vector.

[0020] In some possible embodiments, the information integration module includes:

[0021] A fully convolutional encoding unit, configured to pass the output current time series input vector through a feature extractor based on a fully convolutional network model to obtain an output current time series feature vector;

[0022] A first linear processing unit, configured to perform linear processing on the output current time series feature vector to obtain a linearly processed output current time series feature vector;

[0023] A second linear processing unit, configured to perform linear processing on the sequence of light intensity local time series feature vectors to obtain a sequence of linearly processed light intensity local time series feature vectors;

[0024] A vector fusion unit, configured to fuse sequences of the output current timing feature vectors after the linear processing and the local timing feature vectors of the light intensity after the linear processing to obtain a linearly pre-fused vector;

[0025] A one-dimensional convolution unit, configured to perform one-dimensional convolution processing on the sequence of the local timing feature vectors of the light intensity to obtain a sequence of light intensity timing neighborhood correlation feature vectors; and

[0026] A splicing fusion unit, configured to fuse the sequence of the light intensity timing neighborhood correlation feature vectors and the linearly pre-fused vector in a splicing manner to obtain the light intensity timing embedded output current timing feature vector.

[0027] In some possible embodiments, the dust coverage analysis module includes:

[0028] A feature distribution optimization unit, configured to optimize the feature distribution of the light intensity timing embedded output current timing feature vector to obtain an optimized light intensity timing embedded output current timing feature vector; and

[0029] A classification unit, configured to pass the optimized light intensity timing embedded output current timing feature vector through a classifier to obtain a classification result, where the classification result is used to indicate whether the dust coverage of the monitored photovoltaic module exceeds a predetermined threshold.

[0030] In some possible embodiments, the classification unit includes:

[0031] A fully connected encoding subunit, configured to perform fully connected encoding on the optimized light intensity timing embedded output current timing feature vector using the fully connected layer of the classifier to obtain an encoded classification feature vector; and

[0032] A classification subunit, configured to input the encoded classification feature vector into the Softmax classification function of the classifier to obtain the classification result.

[0033] In a second aspect, an embodiment of the present invention provides a method for monitoring dust of a photovoltaic module, including:

[0034] Obtaining light intensity values at a plurality of predetermined time points within a predetermined time period and output current values of the monitored photovoltaic module at the plurality of predetermined time points;

[0035] Performing data preprocessing on the light intensity values at the plurality of predetermined time points and the output current values of the monitored photovoltaic module at the plurality of predetermined time points to obtain a light intensity timing input vector and an output current timing input vector;

[0036] Performing timing analysis on the light intensity timing input vector to obtain a sequence of local timing feature vectors of the light intensity;

[0037] Integrate the output current timing information expressed by the output current timing input vector into the sequence of the local timing feature vectors of the light intensity to obtain a light intensity timing embedded output current timing feature vector;

[0038] Based on the light intensity timing embedded output current timing feature vector, determine whether the dust coverage rate of the monitored photovoltaic module exceeds a predetermined threshold.

[0039] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it can implement the dust monitoring method described above.

[0040] Compared with the prior art, the dust monitoring system and method for a photovoltaic module provided by the embodiment of the present invention first perform data preprocessing on the light intensity values at multiple predetermined time points and the output current values of the monitored photovoltaic module at the multiple predetermined time points to obtain a light intensity timing input vector and an output current timing input vector. Then, perform timing analysis on the light intensity timing input vector to obtain a sequence of local timing feature vectors of the light intensity. Then, integrate the output current timing information expressed by the output current timing input vector into the sequence of the local timing feature vectors of the light intensity to obtain a light intensity timing embedded output current timing feature vector. Finally, based on the light intensity timing embedded output current timing feature vector, determine whether the dust coverage rate of the monitored photovoltaic module exceeds a predetermined threshold. In this way, it can be intelligently determined whether cleaning is required. Description of the Drawings

[0041] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for use in the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0042] Figure 1 It is a block diagram schematic of a dust monitoring system for a photovoltaic module according to an embodiment of the present invention.

[0043] Figure 2 It is a block diagram schematic of the timing analysis module in the dust monitoring system for a photovoltaic module according to an embodiment of the present invention.

[0044] Figure 3 It is a block diagram schematic of the information integration module in the dust monitoring system for a photovoltaic module according to an embodiment of the present invention.

[0045] Figure 4 Schematic block diagram of the dust coverage analysis module in the dust monitoring system for a photovoltaic module according to an embodiment of the present invention.

[0046] Figure 5 Schematic block diagram of the classification unit in the dust monitoring system for a photovoltaic module according to an embodiment of the present invention.

[0047] Figure 6 Flowchart of the dust monitoring method for a photovoltaic module according to an embodiment of the present invention.

[0048] Figure 7 Schematic diagram of the system architecture of the dust monitoring method for a photovoltaic module according to an embodiment of the present invention.

[0049] Figure 8 Application scenario diagram of the dust monitoring system for a photovoltaic module according to an embodiment of the present invention. Detailed implementation manners

[0050] To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the described embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0051] Unless otherwise specifically stated, the technical terms or scientific terms used in the embodiments of the present invention should be of the ordinary meaning understood by those of ordinary skill in the art to which the present invention pertains. The "including" or "comprising" used in the embodiments of the present invention neither limits the mentioned shapes, numbers, steps, actions, operations, components, elements and / or their groups, nor excludes the occurrence or addition of one or more other different shapes, numbers, steps, actions, operations, components, elements and / or their groups, or the addition of these. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity or order of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of the present invention, "a plurality" means two or more, unless otherwise specifically and clearly defined.

[0052] Unless otherwise specifically stated, the relative settings, numerical expressions, and numerical values of the components and steps set forth in these embodiments do not limit the scope of the present invention. At the same time, it should be understood that for the sake of description, the dimensions of the various parts shown in the drawings are not drawn in actual proportional relationship. For technologies, methods, and devices known to those of ordinary skill in the relevant art, they may not be discussed in detail, but where appropriate, the technologies, methods, and devices shown should be regarded as part of the authorization specification. In all the examples shown and discussed here, any specific other example may have different values. It should be noted that similar symbols and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.

[0053] In the description of the embodiments of the present invention, the description with reference to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In the embodiments of the present invention, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in the embodiments of the present invention and the features of different embodiments or examples.

[0054] In the present invention, flowcharts are used to illustrate the operations performed by the system according to the embodiments of the present invention. It should be understood that the operations before or below do not necessarily need to be precisely executed in sequence. On the contrary, as needed, various steps can be executed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or one or several steps of operations can be removed from these processes.

[0055] Next, example embodiments according to the present invention will be described in detail with reference to the drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments of the present invention. It should be understood that the present invention is not limited by the example embodiments described here.

[0056] It is considered that the dust coverage rate can be estimated by comparing the difference between the actual output current and the theoretical output current. Generally, a high difference between the actual output current and the theoretical output current may indicate that the surface of the photovoltaic module is covered with more dust. Based on this characteristic, the technical concept of the present invention is: using a deep learning algorithm to estimate and predict the theoretical output current of the photovoltaic module from real-time light intensity data, and comparing and analyzing it with the actual output current data, so as to estimate and predict the dust coverage rate on the surface of the photovoltaic module and intelligently determine whether cleaning is required.

[0057] Based on this, Figure 1 FIG. is a schematic block diagram of a dust monitoring system for a photovoltaic module according to an embodiment of the present invention. As Figure 1 shown, the dust monitoring system 100 for a photovoltaic module according to an embodiment of the present invention includes: a data acquisition module 110, configured to acquire light intensity values at a plurality of predetermined time points within a predetermined time period and output current values of the monitored photovoltaic module at the plurality of predetermined time points; a data preprocessing module 120, configured to perform data preprocessing on the light intensity values at the plurality of predetermined time points and the output current values of the monitored photovoltaic module at the plurality of predetermined time points to obtain a light intensity time series input vector and an output current time series input vector; a time series analysis module 130, configured to perform time series analysis on the light intensity time series input vector to obtain a sequence of light intensity local time series feature vectors; an information integration module 140, configured to integrate the output current time series information expressed by the output current time series input vector into the sequence of light intensity local time series feature vectors to obtain a light intensity time series embedded output current time series feature vector; and a dust coverage rate analysis module 150, configured to determine whether the dust coverage rate of the monitored photovoltaic module exceeds a predetermined threshold based on the light intensity time series embedded output current time series feature vector.

[0058] Specifically, in the technical solution of the embodiment of the present invention, first, light intensity values at a plurality of predetermined time points within a predetermined time period and output current values of the monitored photovoltaic module at the plurality of predetermined time points are acquired; and the light intensity values at the plurality of predetermined time points and the output current values of the monitored photovoltaic module at the plurality of predetermined time points are respectively arranged as a light intensity time series input vector and an output current time series input vector according to the time dimension.

[0059] Then, time series analysis is performed on the light intensity time series input vector to obtain a sequence of light intensity local time series feature vectors. That is, the change pattern and trend of the light intensity are captured from the time dimension.

[0060] In a specific example of the present invention, the encoding process for performing temporal analysis on the light intensity temporal input vector to obtain a sequence of light intensity local temporal feature vectors includes: first, performing vector segmentation on the light intensity temporal input vector to obtain a sequence of light intensity local temporal input vectors; then, passing the sequence of light intensity local temporal input vectors through a light intensity temporal feature extractor based on a one-dimensional convolutional layer to obtain a sequence of light intensity local temporal feature vectors.

[0061] That is, by means of vector segmentation, the light intensity temporal input vector is split into light intensity local temporal input vectors, and then a one-dimensional convolutional layer is used to capture the light intensity temporal feature distribution contained in each light intensity local temporal input vector, so as to guide the model to pay attention to more local detailed information.

[0062] Correspondingly, the data preprocessing module 120 is configured to: arrange the light intensity values at the plurality of predetermined time points and the output current values of the monitored photovoltaic modules at the plurality of predetermined time points in the time dimension to form the light intensity temporal input vector and the output current temporal input vector respectively.

[0063] It should be understood that in the data preprocessing module 120, the light intensity temporal input vector contains the light intensity values arranged in time sequence, which can be used to analyze and predict the change trend of light, and this is very important for the performance evaluation and optimization of photovoltaic modules, because light intensity is one of the important factors affecting the output current of photovoltaic modules. The output current temporal input vector contains the output current values of the photovoltaic module arranged in time sequence, which can be used to analyze and predict the current output situation of the photovoltaic module. By analyzing the output current temporal input vector, the working state, performance change and possible faults or abnormalities of the photovoltaic module can be understood. Generally speaking, arranging the light intensity value and the output current value in the time dimension as the temporal input vector can provide an important data basis for the performance analysis, fault diagnosis and optimization of the photovoltaic module. These data can be used for tasks such as model building, data mining and machine learning to improve the efficiency and reliability of the photovoltaic system.

[0064] Further, as Figure 2 shown, the temporal analysis module 130 includes: a vector segmentation unit 131 configured to perform vector segmentation on the light intensity temporal input vector to obtain a sequence of light intensity local temporal input vectors; and a light intensity temporal feature extraction unit 132 configured to pass the sequence of light intensity local temporal input vectors through a light intensity temporal feature extractor based on a one-dimensional convolutional layer to obtain the sequence of light intensity local temporal feature vectors.

[0065] It should be understood that the purpose of the vector segmentation unit 131 is to divide the light intensity data according to time windows, so as to independently analyze and process the light intensity in different time periods. By segmenting the vector, local light intensity information in different time periods can be obtained, making subsequent feature extraction and analysis more refined and accurate. The light intensity time series feature extraction unit 132 uses a light intensity time series feature extractor based on a one-dimensional convolutional layer to process each local time series input vector and extract the feature information therein. The one-dimensional convolutional layer can capture local patterns and trends in the light intensity time series data, thereby extracting useful feature representations. The sequence of the extracted local light intensity time series feature vectors can be used for subsequent model training, prediction, or other analysis tasks. Generally speaking, the vector segmentation unit 131 segments the light intensity time series input vector into a sequence of local time series input vectors, enabling the light intensity data to be independently processed according to time windows. The light intensity time series feature extraction unit 132 extracts the features of the local time series input vectors through a one-dimensional convolutional layer to obtain a sequence of local light intensity time series feature vectors. These feature vectors can be used for subsequent analysis, modeling, and decision-making, such as tasks like performance evaluation, fault detection, or light intensity prediction of a photovoltaic system.

[0066] It is worth mentioning that the one-dimensional convolutional layer is one of the layer types of a commonly used convolutional neural network (CNN) in deep learning. It is mainly used to process one-dimensional sequence data, such as time series data or signal data. The one-dimensional convolutional layer extracts features by sliding a convolutional kernel (also known as a filter) of a fixed size over the input sequence. The convolutional kernel is a small learnable parameter matrix that performs element-wise multiplication and summation operations with the input sequence to generate an output feature map. The sliding process of the convolutional kernel can capture local patterns and features in the input sequence. In the one-dimensional convolutional layer, the convolutional kernel operates in only one direction when sliding over the input sequence, usually from left to right. Each time it slides, the convolutional kernel performs a convolution operation with a part of the input sequence to obtain an output value. By using different convolutional kernels and stride lengths, the one-dimensional convolutional layer can extract features of different scales and different levels of abstraction. The one-dimensional convolutional layer is usually combined with other types of layers (such as pooling layers, fully connected layers, etc.) to build more complex neural network models. It performs well in many tasks, such as text classification in natural language processing, acoustic modeling in speech recognition, time series prediction, etc. Generally speaking, the one-dimensional convolutional layer is a neural network layer for processing one-dimensional sequence data, which extracts local features of the input sequence by sliding the convolutional kernel and provides meaningful feature representations for subsequent tasks.

[0067] Specifically, the light intensity temporal feature extraction unit 132 is configured to: use each layer of the light intensity temporal feature extractor based on a one-dimensional convolutional layer to perform one-dimensional convolutional processing, pooling processing, and non-linear activation processing on the input data respectively during the forward pass of the layer, so as to output a sequence of light intensity local temporal feature vectors by the last layer of the light intensity temporal feature extractor based on a one-dimensional convolutional layer, wherein the input of the first layer of the light intensity temporal feature extractor based on a one-dimensional convolutional layer is the sequence of light intensity local temporal input vectors.

[0068] Next, the output current temporal information expressed by the output current temporal input vector is incorporated into the sequence of light intensity local temporal feature vectors to obtain a light intensity temporal embedded output current temporal feature vector. That is, the output current temporal features are compared with the light intensity temporal features and feature interaction is performed, so that the model can learn the association and difference between the two.

[0069] In a specific example of the present invention, the implementation manner of incorporating the output current temporal information expressed by the output current temporal input vector into the sequence of light intensity local temporal feature vectors to obtain a light intensity temporal embedded output current temporal feature vector is: passing the sequence of light intensity local temporal feature vectors and the output current temporal input vector through a feature embedding module to obtain a light intensity temporal embedded output current temporal feature vector.

[0070] Accordingly, the information integration module 140 is configured to: obtain the light intensity time-series embedded output current time-series feature vector by inputting the sequence of the light intensity local time-series feature vectors and the output current time-series input vector into a feature embedding module. It should be understood that the feature embedding module is a component of the information integration module, which is used to perform feature fusion on the sequence of the light intensity local time-series feature vectors and the output current time-series input vector to obtain the light intensity time-series embedded output current time-series feature vector. The main function of the feature embedding module is to fuse two different types of time-series feature vectors to extract the correlation and complementarity between them. By fusing the time-series features of the light intensity and the output current, the behavior and performance of the photovoltaic system can be described more comprehensively. Specifically, the feature embedding module can adopt various methods for feature fusion, such as concatenation, weighted summation, element-wise multiplication, etc. These methods can combine the time-series features of the light intensity and the output current to generate a new feature vector, which contains the correlation information between the two. The fused light intensity time-series embedded output current time-series feature vector can be used for subsequent model training, prediction, or other analysis tasks. For example, it can be input into a Recurrent Neural Network (RNN) for time-series modeling for performance prediction or anomaly detection of the photovoltaic system. Feature fusion can provide a richer feature representation, which helps to improve the accuracy and robustness of the model. In summary, the feature embedding module is used to perform feature fusion on the light intensity local time-series feature vector and the output current time-series input vector to obtain the light intensity time-series embedded output current time-series feature vector. Such fusion can provide a more comprehensive and rich feature representation for the modeling, prediction, and other analysis tasks of the photovoltaic system.

[0071] Specifically, as Figure 3As shown, the information integration module 140 includes: a fully convolutional encoding unit 141 for obtaining an output current timing feature vector by passing the output current timing input vector through a feature extractor based on a fully convolutional network model; a primary linear processing unit 142 for linearly processing the output current timing feature vector to obtain a linearly processed output current timing feature vector; a secondary linear processing unit 143 for linearly processing the sequence of local illumination intensity timing feature vectors to obtain a sequence of linearly processed local illumination intensity timing feature vectors; a vector fusion unit 144 for fusing the linearly processed output current timing feature vector and the sequence of linearly processed local illumination intensity timing feature vectors to obtain a linearly pre-fused vector; a one-dimensional convolutional unit 145 for performing one-dimensional convolutional processing on the sequence of local illumination intensity timing feature vectors to obtain a sequence of illumination intensity timing neighborhood correlation feature vectors; and a splicing fusion unit 146 for fusing the sequence of illumination intensity timing neighborhood correlation feature vectors and the linearly pre-fused vector in a splicing manner to obtain the illumination intensity timing embedded output current timing feature vector.

[0072] Subsequently, the illumination intensity timing embedded output current timing feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the dust coverage rate of the monitored photovoltaic module exceeds a predetermined threshold.

[0073] Correspondingly, as Figure 4 shown, the dust coverage rate analysis module 150 includes: a feature distribution optimization unit 151 for optimizing the feature distribution of the illumination intensity timing embedded output current timing feature vector to obtain an optimized illumination intensity timing embedded output current timing feature vector; and a classification unit 152 for passing the optimized illumination intensity timing embedded output current timing feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the dust coverage rate of the monitored photovoltaic module exceeds a predetermined threshold.

[0074] It should be understood that in the feature distribution optimization unit 151, the purpose of feature distribution optimization is to make the feature vector more representative and distinguishable by adjusting the value range and distribution of each dimension in the feature vector. This can be achieved through some statistical methods, normalization or standardization techniques, feature selection methods, etc. The optimized feature vector can improve the accuracy and robustness of subsequent classification tasks. In the classification unit 152, this unit is used to pass the optimized light intensity time series embedded output current time series feature vector through a classifier to obtain a classification result. The classifier can be various machine learning algorithms or deep learning models, such as Support Vector Machine (SVM), Random Forest, neural network, etc. The classifier maps the input feature vector to different categories or labels, indicating whether the dust coverage rate of the monitored photovoltaic module exceeds a predetermined threshold. Through the classification result, the dust coverage degree of the photovoltaic module can be judged, and then corresponding maintenance and cleaning measures can be taken. The feature distribution optimization unit and the classification unit are two key parts of the dust coverage rate analysis module. The feature distribution optimization unit improves the representational ability of features by optimizing the distribution of the feature vector; the classification unit classifies the optimized feature vector through a classifier to judge the dust coverage degree of the photovoltaic module. Such analysis can help monitor the operating state of the photovoltaic module, timely detect dust coverage problems, and take corresponding measures to maintain and optimize the performance of the photovoltaic system.

[0075] Here, after passing the sequence of the local time series input vectors of the light intensity through the light intensity time series feature extractor based on a one-dimensional convolutional layer, high-order local time series correlation features within the local time domain of the light intensity values can be extracted, and the output current time series input vector expresses the distribution information of the output current values along the time series direction. Thus, when passing the output current time series input vector and the sequence of the light intensity local time series feature vectors through the feature embedding module, time series dynamic encoding of the high-order local time series correlation features within the local time domain of the light intensity values along the time series direction can be performed based on the time series distribution of the output current values, so that the light intensity time series embedded output current time series feature vector expresses time series multi-order correlation features. However, such time series multi-order correlation features will also cause sparsification of the time series feature distribution representation of the light intensity time series embedded output current time series feature vector, resulting in poor convergence of the probability density distribution of the regression probabilities of each eigenvalue of the light intensity time series embedded output current time series feature vector when performing class probability regression mapping through the classifier, affecting the accuracy of the classification result obtained through the classifier. Therefore, preferably, each eigenvalue of the light intensity time series embedded output current time series feature vector is optimized.

[0076] Correspondingly, in one example, the feature distribution optimization unit 151 is configured to: optimize the feature distribution of the light intensity time series embedded output current time series feature vector according to the following optimization formula to obtain the optimized light intensity time series embedded output current time series feature vector; where the optimization formula is:

[0077]

[0078] where V is the light intensity time series embedded output current time series feature vector, v i and v j are the i-th and j-th eigenvalues of the light intensity time series embedded output current time series feature vector, and is the global feature mean of the light intensity time series embedded output current time series feature vector, exp{·} represents the exponential operation of a numerical value, and the exponential operation of the numerical value represents calculating the value of the natural exponential function with the numerical value as the power, and v′ i is the i-th eigenvalue of the optimized light intensity time series embedded output current time series feature vector.

[0079] Specifically, for the local probability density mismatch of the probability density distribution in the probability space caused by the sparse distribution of the light intensity time series embedded output current time series feature vector in the high-dimensional feature space, through regularized global self-consistent coding, to imitate the global self-consistent relationship of the coding behavior of the high-dimensional feature manifold of the light intensity time series embedded output current time series feature vector in the probability space, so as to adjust the error landscape of the feature manifold in the high-dimensional open space domain, and realize the self-consistent matching type coding of the high-dimensional feature manifold of the light intensity time series embedded output current time series feature vector to the explicit probability space embedding, thereby enhancing the convergence of the probability density distribution of the regression probability of the light intensity time series embedded output current time series feature vector and improving the accuracy of the classification result obtained by the classifier.

[0080] Furthermore, as Figure 5 shown, the classification unit 152 includes: a fully connected coding subunit 1521, configured to perform fully connected coding on the optimized light intensity time series embedded output current time series feature vector using the fully connected layer of the classifier to obtain a coded classification feature vector; and a classification subunit 1522, configured to input the coded classification feature vector into the Softmax classification function of the classifier to obtain the classification result.

[0081] That is, in the technical solution of the present invention, the labels of the classifier include that the dust coverage rate of the monitored photovoltaic module exceeds a predetermined threshold (the first label), and that the dust coverage rate of the monitored photovoltaic module does not exceed the predetermined threshold (the second label). Among them, the classifier determines which classification label the optimized light intensity time series embedding output current time series feature vector belongs to through the softmax function. It should be noted that the first label p1 and the second label p2 here do not contain the concept set by humans. In fact, during the training process, the computer model does not have the concept of "whether the dust coverage rate of the monitored photovoltaic module exceeds the predetermined threshold". It only has two classification labels and the probabilities of the output features under these two classification labels, that is, the sum of p1 and p2 is one. Therefore, the classification result of whether the dust coverage rate of the monitored photovoltaic module exceeds the predetermined threshold is actually transformed into a binary classification probability distribution that conforms to the natural law through the classification label. Essentially, the physical meaning of the natural probability distribution of the label is used, rather than the language text meaning of "whether the dust coverage rate of the monitored photovoltaic module exceeds the predetermined threshold".

[0082] It should be understood that the role of the classifier is to use the given categories and known training data to learn classification rules and classifiers, and then classify (or predict) unknown data. Logistic regression, SVM, etc. are commonly used to solve binary classification problems. For multi-class classification problems, logistic regression or SVM can also be used. However, it requires multiple binary classifications to form a multi-class classification, which is prone to errors and has low efficiency. The commonly used multi-class classification method is the Softmax classification function.

[0083] In summary, the dust monitoring system 100 of the photovoltaic module based on the embodiment of the present invention is clarified, which can intelligently judge whether cleaning is required.

[0084] As described above, the dust monitoring system 100 of the photovoltaic module based on the embodiment of the present invention can be implemented in various terminal devices, such as a server with the dust monitoring algorithm of the photovoltaic module based on the embodiment of the present invention. In one example, the dust monitoring system 100 of the photovoltaic module based on the embodiment of the present invention can be integrated into the terminal device as a software module and / or a hardware module. For example, the dust monitoring system 100 of the photovoltaic module based on the embodiment of the present invention can be a software module in the operating system of the terminal device, or can be an application program developed for the terminal device; of course, the dust monitoring system 100 of the photovoltaic module based on the embodiment of the present invention can also be one of the many hardware modules of the terminal device.

[0085] Alternatively, in another example, the dust monitoring system 100 of the photovoltaic module based on the embodiments of the present invention and the terminal device may also be discrete devices, and the dust monitoring system 100 of the photovoltaic module may be connected to the terminal device through a wired and / or wireless network and transmit and interact information in accordance with a predefined data format.

[0086] Figure 6 FIG. is a flowchart of a dust monitoring method for a photovoltaic module according to an embodiment of the present invention. Figure 7 FIG. is a schematic diagram of a system architecture of a dust monitoring method for a photovoltaic module according to an embodiment of the present invention. As Figure 6 and Figure 7 shown, the dust monitoring method for a photovoltaic module according to an embodiment of the present invention includes: S110, obtaining light intensity values at a plurality of predetermined time points within a predetermined time period and output current values of the monitored photovoltaic module at the plurality of predetermined time points; S120, performing data preprocessing on the light intensity values at the plurality of predetermined time points and the output current values of the monitored photovoltaic module at the plurality of predetermined time points to obtain a light intensity time series input vector and an output current time series input vector; S130, performing time series analysis on the light intensity time series input vector to obtain a sequence of light intensity local time series feature vectors; S140, integrating the output current time series information expressed by the output current time series input vector into the sequence of light intensity local time series feature vectors to obtain a light intensity time series embedded output current time series feature vector; and S150, based on the light intensity time series embedded output current time series feature vector, determining whether the dust coverage rate of the monitored photovoltaic module exceeds a predetermined threshold.

[0087] In a specific example, in the above dust monitoring method for a photovoltaic module, performing data preprocessing on the light intensity values at the plurality of predetermined time points and the output current values of the monitored photovoltaic module at the plurality of predetermined time points to obtain a light intensity time series input vector and an output current time series input vector includes: arranging the light intensity values at the plurality of predetermined time points and the output current values of the monitored photovoltaic module at the plurality of predetermined time points in a time dimension respectively to form the light intensity time series input vector and the output current time series input vector.

[0088] In a specific example, in the above dust monitoring method for a photovoltaic module, performing time series analysis on the light intensity time series input vector to obtain a sequence of light intensity local time series feature vectors includes: performing vector segmentation on the light intensity time series input vector to obtain a sequence of light intensity local time series input vectors; and passing the sequence of light intensity local time series input vectors through a light intensity time series feature extractor based on a one-dimensional convolutional layer to obtain the sequence of light intensity local time series feature vectors.

[0089] Here, those skilled in the art can understand that the specific operations of each step in the above dust monitoring method for photovoltaic modules have been described in detail in the description of the dust monitoring system 100 for photovoltaic modules with reference to Figures 1 to 5 above, and therefore, the repeated description thereof will be omitted.

[0090] Figure 8 FIG. is an application scenario diagram of a dust monitoring system for a photovoltaic module according to an embodiment of the present invention. As Figure 8 shown, in this application scenario, first, light intensity values at multiple predetermined time points within a predetermined time period (e.g., Figure 8 D1 shown in Figure 8 ) and output current values of the monitored photovoltaic module at the multiple predetermined time points (e.g., Figure 8 D2 shown in

[0091] ) are obtained. Then, the light intensity values at the multiple predetermined time points and the output current values of the monitored photovoltaic module at the multiple predetermined time points are input into a server (e.g., Figure 8 S shown in

[0091] ) deployed with a dust monitoring algorithm for photovoltaic modules. Among them, the server can use the dust monitoring algorithm for photovoltaic modules to process the light intensity values at the multiple predetermined time points and the output current values of the monitored photovoltaic module at the multiple predetermined time points to obtain a classification result indicating whether the dust coverage rate of the monitored photovoltaic module exceeds a predetermined threshold.

[0092] It is understandable that the above embodiments are merely exemplary embodiments adopted to illustrate the principles of the present invention. However, the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also regarded as the protection scope of the present invention.

Claims

1. A dust monitoring system for a photovoltaic module, characterized in that, Including: A data acquisition module, configured to acquire light intensity values at a plurality of predetermined time points within a predetermined time period and output current values of the monitored photovoltaic module at the plurality of predetermined time points; A data preprocessing module, configured to perform data preprocessing on the light intensity values at the plurality of predetermined time points and the output current values of the monitored photovoltaic module at the plurality of predetermined time points to obtain a light intensity time series input vector and an output current time series input vector; A time series analysis module, configured to perform time series analysis on the light intensity time series input vector to obtain a sequence of light intensity local time series feature vectors; An information integration module, configured to integrate the output current time series information expressed by the output current time series input vector into the sequence of the light intensity local time series feature vectors to obtain a light intensity time series embedded output current time series feature vector; And A dust coverage analysis module, configured to determine whether the dust coverage of the monitored photovoltaic module exceeds a predetermined threshold based on the light intensity time series embedded output current time series feature vector.

2. The dust monitoring system for a photovoltaic module according to claim 1, wherein The data preprocessing module is configured to: Arrange the light intensity values at the plurality of predetermined time points and the output current values of the monitored photovoltaic module at the plurality of predetermined time points in the time dimension respectively to form the light intensity time series input vector and the output current time series input vector.

3. The dust monitoring system for a photovoltaic module according to claim 2, characterized in that The time series analysis module includes: A vector segmentation unit, configured to segment the light intensity time series input vector to obtain a sequence of light intensity local time series input vectors; and A light intensity time series feature extraction unit, configured to pass the sequence of the light intensity local time series input vectors through a light intensity time series feature extractor based on a one-dimensional convolutional layer to obtain the sequence of the light intensity local time series feature vectors.

4. The dust monitoring system for a photovoltaic module according to claim 3, wherein, The light intensity time series feature extraction unit is configured to: Use each layer of the light intensity time series feature extractor based on the one-dimensional convolutional layer to perform one-dimensional convolutional processing, pooling processing, and non-linear activation processing on the input data respectively during the forward pass of the layer, so as to output the sequence of the light intensity local time series feature vectors from the last layer of the light intensity time series feature extractor based on the one-dimensional convolutional layer, wherein the input of the first layer of the light intensity time series feature extractor based on the one-dimensional convolutional layer is the sequence of the light intensity local time series input vectors.

5. The dust monitoring system for a photovoltaic module according to claim 4, wherein, The information integration module is configured to: Pass the sequence of the light intensity local time series feature vectors and the output current time series input vector through a feature embedding module to obtain the light intensity time series embedded output current time series feature vector.

6. The dust monitoring system for a photovoltaic module according to claim 5, wherein, The information integration module includes: A fully convolutional encoding unit, configured to pass the output current time series input vector through a feature extractor based on a fully convolutional network model to obtain an output current time series feature vector; A first linear processing unit, configured to perform linear processing on the output current time series feature vector to obtain a linearly processed output current time series feature vector; A second linear processing unit, configured to perform linear processing on the sequence of the light intensity local time series feature vectors to obtain a sequence of linearly processed light intensity local time series feature vectors; A vector fusion unit for fusing sequences of the current timing feature vectors output after the linear processing and the local timing feature vectors of the light intensity after the linear processing to obtain a linearly pre-fused vector; A one-dimensional convolution unit for performing one-dimensional convolution processing on the sequence of the local timing feature vectors of the light intensity to obtain a sequence of the neighborhood correlation feature vectors of the light intensity timing; and A splicing fusion unit for fusing the sequence of the neighborhood correlation feature vectors of the light intensity timing and the linearly pre-fused vector in a splicing manner to obtain the light intensity timing embedded output current timing feature vector.

7. The dust monitoring system for a photovoltaic module according to claim 6, wherein, The dust coverage analysis module includes: A feature distribution optimization unit for optimizing the feature distribution of the light intensity timing embedded output current timing feature vector to obtain an optimized light intensity timing embedded output current timing feature vector; and A classification unit for obtaining a classification result by passing the optimized light intensity timing embedded output current timing feature vector through a classifier, where the classification result is used to indicate whether the dust coverage of the monitored photovoltaic module exceeds a predetermined threshold.

8. The dust monitoring system for a photovoltaic module according to claim 7, wherein, The classification unit includes: A fully connected encoding sub-unit for performing fully connected encoding on the optimized light intensity timing embedded output current timing feature vector using the fully connected layer of the classifier to obtain an encoded classification feature vector; and A classification sub-unit for inputting the encoded classification feature vector into the Softmax classification function of the classifier to obtain the classification result.

9. A method for monitoring dust on a photovoltaic module, characterized in that, It includes: Obtaining the light intensity values at multiple predetermined time points within a predetermined time period and the output current values of the monitored photovoltaic module at the multiple predetermined time points; Performing data preprocessing on the light intensity values at the multiple predetermined time points and the output current values of the monitored photovoltaic module at the multiple predetermined time points to obtain a light intensity timing input vector and an output current timing input vector; Performing timing analysis on the light intensity timing input vector to obtain a sequence of local timing feature vectors of the light intensity; Incorporating the output current timing information expressed by the output current timing input vector into the sequence of the local timing feature vectors of the light intensity to obtain a light intensity timing embedded output current timing feature vector; Based on the light intensity timing embedded output current timing feature vector, determining whether the dust coverage of the monitored photovoltaic module exceeds a predetermined threshold.

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 can implement the dust monitoring method according to claim 9.

Images