Abnormality detection control system and method for photovoltaic module

Abstract

The invention discloses an anomaly detection control system and method for a photovoltaic module. The method comprises the following steps: firstly, carrying out feature extraction on a set of photovoltaic panel surface reference images through a photovoltaic panel surface state feature extractor, then, carrying out prototype feature extraction on an obtained set of photovoltaic panel surface reference semantic feature vectors by using a prototype feature extraction network to obtain photovoltaic panel surface reference prototype semantic feature vectors, and then, carrying out prototype feature extraction on the obtained set of the photovoltaic panel surface reference prototype semantic feature vectors. Performing feature extraction on the surface image of the photovoltaic panel through the photovoltaic panel surface state feature extractor to obtain a semantic feature vector of the surface state of the photovoltaic panel to be detected; and determining whether dust is deposited on the surface of the photovoltaic panel to be detected or not based on a calculated surface state differential semantic measurement coefficient between the photovoltaic panel surface reference prototype semantic feature vector and the surface state semantic feature vector of the photovoltaic panel to be detected, and determining whether photovoltaic panel surface cleaning prompt information is generated or not. Therefore, the power generation efficiency and stability of the photovoltaic module can be ensured.

Description

Technical Field

[0001] The present application relates to the field of photovoltaics, and more specifically, to an abnormality detection control system and method for photovoltaic components. Background Art

[0002] Photovoltaic modules are devices that convert solar energy into electrical energy and are widely used in solar power generation systems. Since photovoltaic modules are usually installed outdoors, they are easily affected by environmental factors such as wind, sand, and dust, which causes dust to accumulate on the surface, reducing light transmittance and thus affecting the output power of photovoltaic modules. Therefore, timely testing the surface condition of photovoltaic modules and cleaning them according to the test results are effective measures to improve the power generation efficiency of photovoltaic modules.

[0003] At present, there are two common methods for detecting the surface status of photovoltaic modules: manual inspection and sensor monitoring. The manual inspection method consumes a lot of manpower and material resources, is inefficient, and cannot achieve real-time monitoring; the sensor monitoring method requires the installation of multiple sensors on the photovoltaic module, which increases the cost and maintenance difficulty, and the sensor signal is easily interfered and cannot accurately reflect the surface status of the photovoltaic module. Therefore, these methods have certain limitations and shortcomings.

[0004] Therefore, an optimized abnormality detection control system for photovoltaic modules is desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The present application provides an abnormality detection control system and method for photovoltaic components, which can collect the surface image of the photovoltaic panel by using a camera, and introduce image processing and analysis algorithms at the back end to automatically analyze the surface image, so as to detect whether the surface of the photovoltaic panel is dusty and the degree of dust accumulation.

[0006] According to one aspect of the present application, there is provided an abnormality detection control system for a photovoltaic assembly, comprising:

[0007] A photovoltaic panel surface detection image acquisition module is used to obtain a photovoltaic panel surface image of the photovoltaic panel to be detected collected by a camera;

[0008] A photovoltaic panel surface reference image acquisition module is used to extract a set of photovoltaic panel surface reference images without dust accumulation on the surface from a database;

[0009] A photovoltaic panel surface reference image semantic feature extraction module, used for extracting features from a set of photovoltaic panel surface reference images through a photovoltaic panel surface state feature extractor based on a deep neural network model to obtain a set of photovoltaic panel surface reference semantic feature vectors;

[0010] A photovoltaic panel surface reference prototype semantic expression module is used to use a prototype feature extraction network to perform prototype feature extraction on the set of photovoltaic panel surface reference semantic feature vectors to obtain a photovoltaic panel surface reference prototype semantic feature vector;

[0011] A photovoltaic panel surface detection image semantic feature extraction module, used for extracting features from the photovoltaic panel surface image through the photovoltaic panel surface state feature extractor based on the deep neural network model to obtain a semantic feature vector of the photovoltaic panel surface state to be detected; and

[0012] The photovoltaic panel surface detection module is used to calculate the surface state differentiation semantic measurement coefficient between the semantic feature vector of the photovoltaic panel surface reference prototype and the surface state semantic feature vector of the photovoltaic panel to be detected, and determine whether the surface of the photovoltaic panel to be detected is dusty based on the surface state differentiation semantic measurement coefficient, and determine whether to generate photovoltaic panel surface cleaning prompt information.

[0013] According to another aspect of the present application, there is provided an abnormality detection control method for a photovoltaic assembly, comprising:

[0014] Acquire a photovoltaic panel surface image of the photovoltaic panel to be inspected captured by a camera;

[0015] A set of reference images of photovoltaic panel surfaces without dust accumulation on the surfaces are extracted from a database;

[0016] Extracting features from the set of photovoltaic panel surface reference images using a photovoltaic panel surface state feature extractor based on a deep neural network model to obtain a set of photovoltaic panel surface reference semantic feature vectors;

[0017] Using a prototype feature extraction network to perform prototype feature extraction on the set of reference semantic feature vectors of the photovoltaic panel surface to obtain a reference prototype semantic feature vector of the photovoltaic panel surface;

[0018] Performing feature extraction on the photovoltaic panel surface image by the photovoltaic panel surface state feature extractor based on the deep neural network model to obtain a semantic feature vector of the photovoltaic panel surface state to be detected; and

[0019] Calculate the surface state differentiation semantic measurement coefficient between the semantic feature vector of the photovoltaic panel surface reference prototype and the semantic feature vector of the surface state of the photovoltaic panel to be detected, and determine whether the surface of the photovoltaic panel to be detected is dusty based on the surface state differentiation semantic measurement coefficient, and determine whether to generate photovoltaic panel surface cleaning prompt information.

[0020] Compared with the prior art, the abnormality detection control system and method for photovoltaic modules provided by the present application first extracts features from a set of photovoltaic panel surface reference images through a photovoltaic panel surface state feature extractor, then uses a prototype feature extraction network to extract prototype features from a set of photovoltaic panel surface reference semantic feature vectors to obtain a photovoltaic panel surface reference prototype semantic feature vector, then extracts features from the photovoltaic panel surface image through the photovoltaic panel surface state feature extractor to obtain a photovoltaic panel surface state semantic feature vector to be detected, and finally, determines whether the surface of the photovoltaic panel to be detected is dusty based on the calculation of the surface state difference semantic metric coefficient between the photovoltaic panel surface reference prototype semantic feature vector and the photovoltaic panel surface state semantic feature vector to be detected, and determines whether to generate photovoltaic panel surface cleaning prompt information. In this way, the power generation efficiency and stability of the photovoltaic module can be guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. The following drawings are not intentionally scaled to the actual size, and the focus is on illustrating the main purpose of the present application.

[0022] Figure 1 It is a block diagram of an abnormality detection control system for photovoltaic components according to an embodiment of the present application.

[0023] Figure 2 4 is a flow chart of an abnormality detection control method for a photovoltaic module according to an embodiment of the present application.

[0024] Figure 3 Schematic diagram of the system architecture of the abnormality detection and control method for photovoltaic components according to an embodiment of the present application.

[0025] Figure 4 This is an application scenario diagram of an abnormality detection control system for photovoltaic components according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.

[0027] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0028] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0029] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0030] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0031] In response to the above technical problems, the technical concept of this application is to use a camera to collect the surface image of the photovoltaic panel, and introduce image processing and analysis algorithms in the back end to automatically analyze the surface image, so as to detect whether there is dust accumulation on the surface of the photovoltaic panel and the degree of dust accumulation. In this way, it is possible to realize automated dust accumulation detection on the surface of the photovoltaic panel, improve the efficiency and accuracy of abnormal detection of photovoltaic components, and when dust accumulation is detected on the surface of the photovoltaic panel, the system will automatically generate photovoltaic panel surface cleaning prompt information, so as to clean the surface of the photovoltaic panel in time, thereby ensuring the power generation efficiency and stability of the photovoltaic component.

[0032] Figure 1 FIG. 1 is a block diagram of an abnormality detection control system for a photovoltaic module according to an embodiment of the present application. Figure 1As shown, according to an embodiment of the present application, an abnormality detection control system 100 for photovoltaic components includes: a photovoltaic panel surface detection image acquisition module 110, which is used to obtain a photovoltaic panel surface image of a photovoltaic panel to be detected acquired by a camera; a photovoltaic panel surface reference image acquisition module 120, which is used to extract a set of photovoltaic panel surface reference images whose surfaces are not dusty from a database; a photovoltaic panel surface reference image semantic feature extraction module 130, which is used to extract features from the set of photovoltaic panel surface reference images through a photovoltaic panel surface state feature extractor based on a deep neural network model to obtain a set of photovoltaic panel surface reference semantic feature vectors; a photovoltaic panel surface reference prototype semantic expression module 140, which is used to extract features from the set of photovoltaic panel surface reference images using a prototype feature extraction network; A set of semantic feature vectors is used to perform prototype feature extraction to obtain a reference prototype semantic feature vector of the photovoltaic panel surface; a photovoltaic panel surface detection image semantic feature extraction module 150 is used to perform feature extraction on the photovoltaic panel surface image through the photovoltaic panel surface state feature extractor based on the deep neural network model to obtain a semantic feature vector of the surface state of the photovoltaic panel to be detected; and a photovoltaic panel surface detection module 160 is used to calculate a surface state differentiation semantic measurement coefficient between the photovoltaic panel surface reference prototype semantic feature vector and the photovoltaic panel surface state semantic feature vector to be detected, and determine whether the surface of the photovoltaic panel to be detected is dusty based on the surface state differentiation semantic measurement coefficient, and determine whether to generate photovoltaic panel surface cleaning prompt information.

[0033] Specifically, in the technical solution of the present application, first, the surface image of the photovoltaic panel to be inspected is acquired by a camera, and a set of reference images of the photovoltaic panel surface without dust accumulation on the surface is extracted from a database. It should be understood that the purpose of extracting a set of reference images of the photovoltaic panel surface without dust accumulation on the surface from the database is to establish a reference standard for comparison with the surface image of the photovoltaic panel to be inspected. By comparing the feature differences between the surface image of the photovoltaic panel to be inspected and the reference image of the photovoltaic panel surface without dust accumulation on the surface, it can be determined whether the surface of the photovoltaic panel is dusty.

[0034] Then, a photovoltaic panel surface state feature extractor based on a convolutional neural network model with excellent performance in extracting implicit features of images is used to perform feature mining on each photovoltaic panel surface reference image in the set of photovoltaic panel surface reference images, so as to respectively extract reference semantic feature information about the surface state of the photovoltaic panel in each photovoltaic panel surface reference image, thereby obtaining a set of photovoltaic panel surface reference semantic feature vectors.

[0035] Correspondingly, in the photovoltaic panel surface reference image semantic feature extraction module 130, the deep neural network model is a convolutional neural network model, that is, the photovoltaic panel surface state feature extractor based on the deep neural network model is a photovoltaic panel surface state feature extractor based on the convolutional neural network model.

[0036] It should be understood that in the process of detecting abnormal surface state of photovoltaic panels, the convolutional neural network model is used to extract state features of the photovoltaic panel surface. These state features can represent feature information such as texture, shape, color, etc. of the photovoltaic panel surface, and are used as a reference state semantic database for detecting the surface state of photovoltaic panels. Furthermore, in order to capture the commonality and prototype feature information between the state semantic features of the surface reference images of photovoltaic panels without dust accumulation on the surface of each surface in the database, and to compare and analyze the semantic features of the photovoltaic panel surface image of the photovoltaic panel to be detected, so as to more accurately detect the surface state of the photovoltaic panel, in the technical solution of the present application, a prototype feature extraction network is further used to extract prototype features from the set of reference semantic feature vectors of the photovoltaic panel surface to obtain a reference prototype semantic feature vector of the photovoltaic panel surface. In particular, the prototype feature extraction network can learn the common features between the reference semantic features of the surface state of each photovoltaic panel in the set of reference images of the photovoltaic panel surface, ignoring individual differences, so as to extract a more representative reference prototype semantic feature vector. This reference prototype semantic feature vector can better represent the reference semantic features of the photovoltaic panel surface, and can be used as a reference standard for comparison with the surface state semantic features of the photovoltaic panel to be detected, so as to determine whether dust is accumulated on the surface of the photovoltaic panel, thereby improving the accuracy and reliability of dust accumulation detection on the photovoltaic panel surface.

[0037] Accordingly, the photovoltaic panel surface reference prototype semantic expression module 140 is used to: use a prototype feature extraction network to perform prototype feature extraction on the set of photovoltaic panel surface reference semantic feature vectors using the following prototype feature extraction formula to obtain the photovoltaic panel surface reference prototype semantic feature vector; wherein the prototype feature extraction formula is:

[0038]

[0039]

[0040] Among them, v i is the i-th photovoltaic panel surface reference semantic feature vector in the set of photovoltaic panel surface reference semantic feature vectors, v j is the jth photovoltaic panel surface reference semantic feature vector in the set of photovoltaic panel surface reference semantic feature vectors, ||·|| 1represents the 1-norm of the feature vector, M is the number of reference semantic feature vectors of the photovoltaic panel surface in the set of reference semantic feature vectors of the photovoltaic panel surface - 1, V k is the set of reference semantic feature vectors of the photovoltaic panel surface, D i represents the semantic feature difference coefficient, exp(·) represents the natural exponential function operation, N s represents the total number of the semantic difference coefficients, and V is the semantic feature vector of the reference prototype of the photovoltaic panel surface.

[0041] In the technical solution of the present application, each photovoltaic panel surface reference semantic feature vector in the set of photovoltaic panel surface reference semantic feature vectors expresses the image semantic features of the corresponding photovoltaic panel surface reference image without dust accumulation on the surface, and thus, the image semantic feature distribution of each photovoltaic panel surface reference semantic feature vector will also be inconsistent due to the difference in source semantic distribution between images. In this way, when the set of photovoltaic panel surface reference semantic feature vectors is processed using a prototype feature extraction network, considering that the inconsistency of feature distribution of each photovoltaic panel surface reference semantic feature vector will affect the significance of its respective feature distribution information based on image semantic feature distribution, it makes it difficult for the photovoltaic panel surface reference prototype semantic feature vector to stably focus on the significant local distribution of features during the training process, thereby affecting the expression effect of the photovoltaic panel surface reference prototype semantic feature vector, and further affecting the calculation accuracy between the surface state differentiation semantic measurement coefficient between the photovoltaic panel surface reference prototype semantic feature vector and the photovoltaic panel surface state semantic feature vector to be detected.

[0042] Based on this, in a preferred embodiment of the present application, the process of using a prototype feature extraction network to extract prototype features from the set of reference semantic feature vectors of the photovoltaic panel surface to obtain the reference prototype semantic feature vector of the photovoltaic panel surface includes:

[0043] Inputting the set of reference semantic feature vectors of the photovoltaic panel surface into a feature filtering network based on pseudo-anchor center contribution semantic query to obtain a set of filtered reference semantic feature vectors of the photovoltaic panel surface;

[0044] The set of filtered photovoltaic panel surface reference semantic feature vectors is input into the prototype feature extraction network to obtain the photovoltaic panel surface reference prototype semantic feature vector.

[0045] Specifically, the set of reference semantic feature vectors of the photovoltaic panel surface is input into a feature filtering network based on pseudo-anchor center contribution semantic query to obtain a filtered set of reference semantic feature vectors of the photovoltaic panel surface, including the steps of:

[0046] Performing cluster analysis on the set of reference semantic feature vectors of the photovoltaic panel surface to obtain a cluster center representation vector as a pseudo anchor center; for example, calculating a positional mean vector of the set of reference semantic feature vectors of the photovoltaic panel surface as the cluster center representation vector;

[0047] Calculating a semantic contribution representation matrix of each photovoltaic panel surface reference semantic feature vector in the set of photovoltaic panel surface reference semantic feature vectors relative to the pseudo anchor center to obtain a sequence of semantic contribution representation matrices;

[0048] Performing a contribution pseudo-masked sparse module based on a random dropout mechanism on each semantic contribution representation matrix in the sequence of semantic contribution representation matrices to obtain a sequence of masked sparse semantic contribution representation matrices;

[0049] Each masked sparse semantic contribution representation matrix in the sequence of the masked sparse semantic contribution representation matrices is used as a modulation matrix, and the matrix product between the masked sparse semantic contribution representation matrix and each photovoltaic panel surface reference semantic feature vector is calculated to obtain the set of filtered photovoltaic panel surface reference semantic feature vectors.

[0050] The data processing process of the feature filtering network based on pseudo anchor center contribution semantic query can be expressed as follows:

[0051] X={x 1 , x 2 , …, x n}

[0052]

[0053]

[0054] {x i '}={WF i *x i}

[0055] Where X represents the set of reference semantic feature vectors of the photovoltaic panel surface, x 1 、x 2 、x n represent the first, second, and nth photovoltaic panel surface reference semantic feature vectors in the set of photovoltaic panel surface reference semantic feature vectors, respectively; {·} represents a set of vectors; x i represents the i-th photovoltaic panel surface reference semantic feature vector in the set of photovoltaic panel surface reference semantic feature vectors, x c represents the cluster center representation vector, ||·|| 2represents the binary norm of the vector, N represents the number of vectors in the set of reference semantic feature vectors of the photovoltaic panel surface, T represents the transpose of the vector, Softmax represents the normalized exponential function, dropout represents the random deactivation mechanism, WF i represents the ith masked sparse semantic contribution representation matrix in the sequence of masked sparse semantic contribution representation matrices, x i ' represents the i-th filtered photovoltaic panel surface reference semantic feature vector in the set of filtered photovoltaic panel surface reference semantic feature vectors.

[0056] Accordingly, in the preferred embodiment, a feature filtering network based on pseudo-anchor center contribution semantic query is used to perform feature filtering on the set of photovoltaic panel surface reference semantic feature vectors based on pseudo-anchor center contribution semantic query, so that the photovoltaic panel surface reference semantic feature vectors in the filtered set of photovoltaic panel surface reference semantic feature vectors have a relatively balanced feature distribution, so as to avoid the situation where the image semantic feature distribution of outliers is significantly abnormal when performing the prototype feature extractor. In this way, the calculation accuracy of the surface state differentiation semantic measurement coefficient between the photovoltaic panel surface reference prototype semantic feature vector and the photovoltaic panel surface state semantic feature vector to be detected is improved.

[0057] Next, the photovoltaic panel surface image is subjected to feature mining in the photovoltaic panel surface state feature extractor based on the convolutional neural network model to extract the surface state detection semantic feature information about the photovoltaic panel to be detected in the photovoltaic panel surface image, thereby obtaining the surface state semantic feature vector of the photovoltaic panel to be detected.

[0058] Further, the surface state difference semantic metric coefficient between the reference prototype semantic feature vector of the photovoltaic panel surface and the semantic feature vector of the surface state of the photovoltaic panel to be detected is calculated to determine the semantic difference between the surface state detection semantic feature of the photovoltaic panel to be detected and the reference prototype semantic feature of the photovoltaic panel surface, so as to determine whether the surface of the photovoltaic panel to be detected is dusty. That is to say, based on the comparison between the surface state difference semantic metric coefficient and the predetermined threshold, it is determined whether the surface of the photovoltaic panel to be detected is dusty. If the surface state difference semantic metric coefficient is greater than the predetermined threshold, it is determined that there is dust on the surface of the photovoltaic panel to be detected. In particular, in a specific example of the present application, in response to the dust accumulation on the surface of the photovoltaic panel to be detected, a photovoltaic panel surface cleaning prompt information is generated. In this way, automated photovoltaic panel surface dust detection can be achieved, the efficiency and accuracy of photovoltaic module abnormal detection can be improved, and when dust is detected on the surface of the photovoltaic panel, the system will automatically generate photovoltaic panel surface cleaning prompt information so as to clean the photovoltaic panel surface in time, thereby ensuring the power generation efficiency and stability of the photovoltaic module.

[0059] Correspondingly, the photovoltaic panel surface detection module 160 includes: a surface state differentiation semantic measurement unit, which is used to calculate the surface state differentiation semantic measurement coefficient between the photovoltaic panel surface reference prototype semantic feature vector and the surface state semantic feature vector of the photovoltaic panel to be detected; a photovoltaic panel surface state detection unit, which is used to determine whether the surface of the photovoltaic panel to be detected is dusty based on a comparison between the surface state differentiation semantic measurement coefficient and a predetermined threshold; and a photovoltaic panel surface cleaning prompt unit, which is used to generate photovoltaic panel surface cleaning prompt information in response to dust accumulation on the surface of the photovoltaic panel to be detected.

[0060] In one example, the surface state differentiation semantic measurement unit includes: calculating the surface state differentiation semantic measurement coefficient between the photovoltaic panel surface reference prototype semantic feature vector and the photovoltaic panel surface state semantic feature vector to be detected by the following coefficient calculation formula; wherein the coefficient calculation formula is:

[0061]

[0062] Among them, p(x) and q(x) are the eigenvalues ​​of each position of the semantic feature vector of the reference prototype of the photovoltaic panel surface and the semantic feature vector of the surface state of the photovoltaic panel to be detected, respectively, N is the scale of the semantic feature vector of the reference prototype of the photovoltaic panel surface and the semantic feature vector of the surface state of the photovoltaic panel to be detected, log represents the logarithmic function with base 2, and S is the surface state differentiation semantic measurement coefficient.

[0063] Furthermore, in the technical solution of the present application, the abnormality detection and control system for photovoltaic modules also includes a training module for training the photovoltaic panel surface state feature extractor based on the convolutional neural network model and the prototype feature extraction network.

[0064] In one example, the training module includes: a training data acquisition unit, which is used to acquire training data, wherein the training data includes a training photovoltaic panel surface image of the photovoltaic panel to be detected captured by a camera, a set of training photovoltaic panel surface reference images with no dust accumulated on the surface extracted from a database, and a true value of a surface state differentiation semantic measurement coefficient; a training photovoltaic panel surface reference image semantic feature extraction unit, which is used to perform feature extraction on the set of training photovoltaic panel surface reference images through the photovoltaic panel surface state feature extractor based on the convolutional neural network model to obtain a set of training photovoltaic panel surface reference semantic feature vectors; a training photovoltaic panel surface reference prototype semantic expression unit, which is used to perform prototype feature extraction on the set of training photovoltaic panel surface reference semantic feature vectors using the prototype feature extraction network to obtain a training photovoltaic panel surface reference prototype semantic feature vector; a training photovoltaic panel surface detection image semantic feature extraction unit , used to extract features from the training photovoltaic panel surface image through the photovoltaic panel surface state feature extractor based on the convolutional neural network model to obtain a semantic feature vector of the surface state of the training photovoltaic panel to be detected; a training photovoltaic panel surface detection unit, used to calculate the training surface state differentiation semantic measurement coefficient between the training photovoltaic panel surface reference prototype semantic feature vector and the training photovoltaic panel surface state semantic feature vector to be detected; a difference loss function value calculation unit, used to calculate the difference loss function value between the training surface state differentiation semantic measurement coefficient and the true value of the surface state differentiation semantic measurement coefficient; and a loss training unit, used to train the photovoltaic panel surface state feature extractor based on the convolutional neural network model and the prototype feature extraction network with the difference loss function value, wherein, in each round of iteration of the training, the training photovoltaic panel surface reference prototype semantic feature vector is optimized.

[0065] In the technical solution of the present application, each training photovoltaic panel surface reference semantic feature vector in the set of training photovoltaic panel surface reference semantic feature vectors expresses the image semantic features of the corresponding training photovoltaic panel surface reference image whose surface is not dusty, and thus, the image semantic feature distribution of each training photovoltaic panel surface reference semantic feature vector will also be inconsistent due to the difference in source semantic distribution between images. In this way, when the set of training photovoltaic panel surface reference semantic feature vectors is processed using a prototype feature extraction network, considering that the significance of each training photovoltaic panel surface reference semantic feature vector based on the image semantic feature distribution will be affected due to the inconsistency of feature distribution, the significance of feature distribution information of each training photovoltaic panel surface reference semantic feature vector based on the image semantic feature distribution will be affected, making it difficult for the training photovoltaic panel surface reference prototype semantic feature vector to stably focus on the significant local distribution of features during the training process, thereby affecting the expression effect of the training photovoltaic panel surface reference prototype semantic feature vector and the training speed of the model.

[0066] Based on this, the applicant of the present application optimizes the training photovoltaic panel surface reference prototype semantic feature vector during each model iteration training, for example, when back-propagating the difference loss function value between the inferred training surface state differentiation semantic measurement coefficient and the true surface state differentiation semantic measurement coefficient through the training photovoltaic panel surface reference prototype semantic feature vector.

[0067] Accordingly, in one example, in each iteration of the training, the semantic feature vector of the reference prototype of the training photovoltaic panel surface is optimized by the following optimization formula to obtain the optimized semantic feature vector of the reference prototype of the training photovoltaic panel surface; wherein the optimization formula is:

[0068]

[0069] Among them, V 1 is the semantic feature vector of the reference prototype of the training photovoltaic panel surface, v 1i is the feature value of the i-th position of the semantic feature vector of the reference prototype of the training photovoltaic panel surface, and are the reference prototype semantic feature vectors V of the training photovoltaic panel surface 1 The 1-norm and 2-norm square of the training photovoltaic panel surface reference prototype semantic feature vector V 1 The length of , and ω is the weight hyperparameter, log[·] represents the logarithmic function with base 2, v 1i ′ is the feature value of the i-th position of the semantic feature vector of the reference prototype of the photovoltaic panel surface after optimization training.

[0070] Specifically, by training the photovoltaic panel surface reference prototype semantic feature vector V 1 The scale and structure parameters of the training photovoltaic panel surface reference prototype semantic feature vector V can be used to perform geometric registration of its high-dimensional feature manifold shape. 1 The features with rich feature semantic information in the feature set composed of the feature values ​​of , that is, the distinguishable stable interest features based on the dissimilarity of local context information representation during model training, thereby realizing the training photovoltaic panel surface reference prototype semantic feature vector V 1 The significant annotation of feature information during the model training process improves the expression effect of the semantic feature vector of the reference prototype of the training photovoltaic panel surface and the training speed of the model. In this way, automatic detection of dust accumulation on the photovoltaic panel surface can be achieved, the efficiency and accuracy of abnormal detection of photovoltaic components can be improved, and when dust accumulation on the photovoltaic panel surface is detected, the system will automatically generate photovoltaic panel surface cleaning prompt information, so as to clean the photovoltaic panel surface in time, thereby ensuring the power generation efficiency and stability of the photovoltaic component.

[0071] In summary, the abnormality detection and control system 100 for photovoltaic modules based on the embodiment of the present application is explained, which can ensure the power generation efficiency and stability of the photovoltaic modules.

[0072] As described above, the abnormality detection control system 100 for photovoltaic components based on the embodiment of the present application can be implemented in various terminal devices, such as a server having an abnormality detection control algorithm for photovoltaic components based on the embodiment of the present application. In one example, the abnormality detection control system 100 for photovoltaic components based on the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the abnormality detection control system 100 for photovoltaic components based on the embodiment of the present application can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the abnormality detection control system 100 for photovoltaic components based on the embodiment of the present application can also be one of the many hardware modules of the terminal device.

[0073] Alternatively, in another example, the abnormality detection control system 100 for photovoltaic components based on the embodiment of the present application and the terminal device may also be separate devices, and the abnormality detection control system 100 for photovoltaic components may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0074] Figure 2 4 is a flow chart of an abnormality detection control method for a photovoltaic module according to an embodiment of the present application. Figure 3 FIG. 1 is a schematic diagram of a system architecture of an abnormality detection control method for a photovoltaic module according to an embodiment of the present application. Figure 2 and Figure 3As shown, according to the embodiment of the present application, the abnormality detection control method for photovoltaic components includes: S110, acquiring a photovoltaic panel surface image of a photovoltaic panel to be detected collected by a camera; S120, extracting a set of photovoltaic panel surface reference images without dust accumulation on the surface from a database; S130, performing feature extraction on the set of photovoltaic panel surface reference images by a photovoltaic panel surface state feature extractor based on a deep neural network model to obtain a set of photovoltaic panel surface reference semantic feature vectors; S140, performing prototype feature extraction on the set of photovoltaic panel surface reference semantic feature vectors using a prototype feature extraction network to obtain a set of photovoltaic panel surface reference semantic feature vectors; Obtaining a semantic feature vector of a reference prototype of a photovoltaic panel surface; S150, performing feature extraction on the photovoltaic panel surface image through the photovoltaic panel surface state feature extractor based on the deep neural network model to obtain a semantic feature vector of the surface state of the photovoltaic panel to be detected; and, S160, calculating a surface state differentiation semantic measurement coefficient between the semantic feature vector of the reference prototype of the photovoltaic panel surface and the semantic feature vector of the surface state of the photovoltaic panel to be detected, and determining whether dust is accumulated on the surface of the photovoltaic panel to be detected based on the surface state differentiation semantic measurement coefficient, and determining whether to generate a prompt message for cleaning the photovoltaic panel surface.

[0075] In a specific example, in the above-mentioned abnormality detection and control method for photovoltaic components, the deep neural network model is a convolutional neural network model.

[0076] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned abnormality detection control method for photovoltaic modules have been described in the above reference. Figure 1 The abnormality detection control system 100 for a photovoltaic module has been described in detail, and therefore, its repeated description will be omitted.

[0077] Figure 4 FIG. 1 is an application scenario diagram of an abnormality detection control system for photovoltaic modules according to an embodiment of the present application. Figure 4 As shown, in this application scenario, first, a photovoltaic panel surface image of the photovoltaic panel to be detected is acquired by a camera (for example, Figure 4 D1) and a set of reference images of photovoltaic panel surfaces without dust accumulation extracted from a database (e.g., Figure 4 D2 as shown in the figure), then, the set of the photovoltaic panel surface reference images and the photovoltaic panel surface image are input to a server (for example, Figure 4In S) as shown in, the server is capable of using the abnormality detection control algorithm for photovoltaic components to process the set of photovoltaic panel surface reference images and the photovoltaic panel surface image to obtain a surface state differentiation semantic measurement coefficient, and then, based on a comparison between the surface state differentiation semantic measurement coefficient and a predetermined threshold, determine whether the surface of the photovoltaic panel to be inspected is dusty, and generate photovoltaic panel surface cleaning prompt information in response to dust accumulation on the surface of the photovoltaic panel to be inspected.

[0078] According to another aspect of the present application, a non-volatile computer-readable storage medium is provided, on which computer-readable instructions are stored. When the instructions are executed by a computer, the above-mentioned method can be executed.

[0079] The program part of the technology can be considered as a "product" or "manufactured product" in the form of executable code and / or related data, which is participated in or realized by computer-readable media. Tangible and permanent storage media can include any memory or storage used by any computer, processor, or similar device or related module. For example, various semiconductor memories, tape drives, disk drives or any similar devices that can provide storage functions for software.

[0080] The present application uses specific words to describe the embodiments of the present application. For example, "first / second embodiment", "one embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of the present application. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or multiple times in different positions in this specification does not necessarily refer to the same embodiment. In addition, some features, structures or characteristics in one or more embodiments of the present application can be appropriately combined.

[0081] In addition, it will be appreciated by those skilled in the art that various aspects of the present application may be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, various aspects of the present application may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may all be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". In addition, various aspects of the present application may be represented as a computer product located in one or more computer-readable media, which includes computer-readable program code.

[0082] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art to which this application belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an idealized or extremely formal sense, unless explicitly defined as such herein.

[0083] The above is an explanation of the present application and should not be considered as a limitation thereof. Although several exemplary embodiments of the present application are described, it will be readily understood by those skilled in the art that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present application. Therefore, all of these modifications are intended to be included within the scope of the present application as defined by the claims. It should be understood that the above is an explanation of the present application and should not be considered to be limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present application is defined by the claims and their equivalents.

Claims

1. An abnormality detection control system for photovoltaic modules, characterized in that: include: A photovoltaic panel surface detection image acquisition module is used to obtain a photovoltaic panel surface image of the photovoltaic panel to be detected collected by a camera; A photovoltaic panel surface reference image acquisition module is used to extract a set of photovoltaic panel surface reference images without dust accumulation on the surface from a database; A photovoltaic panel surface reference image semantic feature extraction module, used for extracting features from a set of photovoltaic panel surface reference images through a photovoltaic panel surface state feature extractor based on a deep neural network model to obtain a set of photovoltaic panel surface reference semantic feature vectors; A photovoltaic panel surface reference prototype semantic expression module is used to use a prototype feature extraction network to perform prototype feature extraction on the set of photovoltaic panel surface reference semantic feature vectors to obtain a photovoltaic panel surface reference prototype semantic feature vector; A photovoltaic panel surface detection image semantic feature extraction module, used for extracting features from the photovoltaic panel surface image through the photovoltaic panel surface state feature extractor based on the deep neural network model to obtain a semantic feature vector of the photovoltaic panel surface state to be detected; and The photovoltaic panel surface detection module is used to calculate the surface state differentiation semantic measurement coefficient between the semantic feature vector of the photovoltaic panel surface reference prototype and the surface state semantic feature vector of the photovoltaic panel to be detected, and determine whether the surface of the photovoltaic panel to be detected is dusty based on the surface state differentiation semantic measurement coefficient, and determine whether to generate photovoltaic panel surface cleaning prompt information.

2. The abnormality detection and control system for photovoltaic modules according to claim 1, characterized in that: include: The deep neural network model is a convolutional neural network model.

3. The abnormality detection and control system for photovoltaic modules according to claim 2, characterized in that: The photovoltaic panel surface reference prototype semantic expression module is used to: Inputting the set of reference semantic feature vectors of the photovoltaic panel surface into a feature filtering network based on pseudo-anchor center contribution semantic query to obtain a set of filtered reference semantic feature vectors of the photovoltaic panel surface; The set of filtered photovoltaic panel surface reference semantic feature vectors is input into the prototype feature extraction network to obtain the photovoltaic panel surface reference prototype semantic feature vector.

4. The abnormality detection and control system for photovoltaic modules according to claim 3, characterized in that: Inputting the set of reference semantic feature vectors of the photovoltaic panel surface into a feature filtering network based on pseudo-anchor center contribution semantic query to obtain a filtered set of reference semantic feature vectors of the photovoltaic panel surface, comprising the steps of: Performing cluster analysis on the set of reference semantic feature vectors of the photovoltaic panel surface to obtain a cluster center representation vector as a pseudo anchor center; Calculating a semantic contribution representation matrix of each photovoltaic panel surface reference semantic feature vector in the set of photovoltaic panel surface reference semantic feature vectors relative to the pseudo anchor center to obtain a sequence of semantic contribution representation matrices; Performing a contribution pseudo-masked sparse module based on a random dropout mechanism on each semantic contribution representation matrix in the sequence of semantic contribution representation matrices to obtain a sequence of masked sparse semantic contribution representation matrices; Each masked sparse semantic contribution representation matrix in the sequence of the masked sparse semantic contribution representation matrices is used as a modulation matrix, and the matrix product between the masked sparse semantic contribution representation matrix and each photovoltaic panel surface reference semantic feature vector is calculated to obtain the set of filtered photovoltaic panel surface reference semantic feature vectors.

5. The abnormality detection and control system for photovoltaic modules according to claim 4, characterized in that: The photovoltaic panel surface detection module comprises: A surface state differentiation semantic measurement unit, used to calculate the surface state differentiation semantic measurement coefficient between the photovoltaic panel surface reference prototype semantic feature vector and the photovoltaic panel surface state semantic feature vector to be detected; A photovoltaic panel surface state detection unit, configured to determine whether dust is accumulated on the surface of the photovoltaic panel to be detected based on a comparison between the surface state difference semantic metric coefficient and a predetermined threshold; and The photovoltaic panel surface cleaning prompt unit is used to generate photovoltaic panel surface cleaning prompt information in response to dust accumulation on the surface of the photovoltaic panel to be detected.

6. The abnormality detection and control system for photovoltaic modules according to claim 5, characterized in that: The surface state differentiation semantic measurement unit includes: The surface state difference semantic measurement coefficient between the semantic feature vector of the photovoltaic panel surface reference prototype and the semantic feature vector of the photovoltaic panel surface state to be detected is calculated using the following coefficient calculation formula; wherein the coefficient calculation formula is: Among them, p (x) and q (x) are the eigenvalues ​​of each position of the semantic feature vector of the reference prototype of the photovoltaic panel surface and the semantic feature vector of the surface state of the photovoltaic panel to be detected, N is the scale of the semantic feature vector of the reference prototype of the photovoltaic panel surface and the semantic feature vector of the surface state of the photovoltaic panel to be detected, log represents the logarithmic function with base 2, and S is the surface state differentiation semantic measurement coefficient.

7. The abnormality detection and control system for photovoltaic modules according to claim 6, characterized in that: It also includes a training module for training a photovoltaic panel surface state feature extractor based on a convolutional neural network model and the prototype feature extraction network.

8. The abnormality detection and control system for photovoltaic modules according to claim 7, characterized in that: The training module comprises: A training data acquisition unit, used to acquire training data, wherein the training data includes a training photovoltaic panel surface image of the photovoltaic panel to be detected collected by a camera, a set of training photovoltaic panel surface reference images without dust accumulation extracted from a database, and a true value of a surface state differentiation semantic measurement coefficient; A training photovoltaic panel surface reference image semantic feature extraction unit, used for performing feature extraction on the set of training photovoltaic panel surface reference images through the photovoltaic panel surface state feature extractor based on the convolutional neural network model to obtain a set of training photovoltaic panel surface reference semantic feature vectors; A training photovoltaic panel surface reference prototype semantic expression unit is used to use the prototype feature extraction network to perform prototype feature extraction on the set of training photovoltaic panel surface reference semantic feature vectors to obtain a training photovoltaic panel surface reference prototype semantic feature vector; A training photovoltaic panel surface detection image semantic feature extraction unit is used to extract features from the training photovoltaic panel surface image through the photovoltaic panel surface state feature extractor based on the convolutional neural network model to obtain a training photovoltaic panel surface state semantic feature vector; A training photovoltaic panel surface detection unit is used to calculate a training surface state difference semantic measurement coefficient between the training photovoltaic panel surface reference prototype semantic feature vector and the training photovoltaic panel surface state semantic feature vector to be detected; a difference loss function value calculation unit, used to calculate a difference loss function value between the training surface state differential semantic metric coefficient and a true value of the surface state differential semantic metric coefficient; and A loss training unit is used to train the photovoltaic panel surface state feature extractor based on the convolutional neural network model and the prototype feature extraction network with the difference loss function value, wherein in each round of iteration of the training, the reference prototype semantic feature vector of the training photovoltaic panel surface is optimized.

9. A method for detecting and controlling abnormality of a photovoltaic module, characterized in that: include: Acquire a photovoltaic panel surface image of the photovoltaic panel to be inspected captured by a camera; A set of reference images of photovoltaic panel surfaces without dust accumulation on the surfaces are extracted from a database; Extracting features from the set of photovoltaic panel surface reference images using a photovoltaic panel surface state feature extractor based on a deep neural network model to obtain a set of photovoltaic panel surface reference semantic feature vectors; Using a prototype feature extraction network to perform prototype feature extraction on the set of reference semantic feature vectors of the photovoltaic panel surface to obtain a reference prototype semantic feature vector of the photovoltaic panel surface; Performing feature extraction on the photovoltaic panel surface image by the photovoltaic panel surface state feature extractor based on the deep neural network model to obtain a semantic feature vector of the photovoltaic panel surface state to be detected; as well as Calculate the surface state differentiation semantic measurement coefficient between the semantic feature vector of the photovoltaic panel surface reference prototype and the semantic feature vector of the surface state of the photovoltaic panel to be detected, and determine whether the surface of the photovoltaic panel to be detected is dusty based on the surface state differentiation semantic measurement coefficient, and determine whether to generate photovoltaic panel surface cleaning prompt information.

10. The abnormality detection and control method for photovoltaic modules according to claim 9, characterized in that: include: The deep neural network model is a convolutional neural network model.

Images