Risk assessment method, device, equipment and storage medium for flexible photovoltaic support

Abstract

The present application discloses a risk assessment method, device, equipment and storage medium for a flexible photovoltaic support, which relates to the technical field of electrical digital data processing. The method calibrates the shape of a photovoltaic panel in a windless state as a standard sample, provides an evaluation basis for the real-time state during subsequent use, and avoids using the photovoltaic panels in use as sample references (which already have vibration characteristics or risk deformation characteristics) to cause inaccurate evaluation. The photovoltaic predicted shape obtained through machine learning is then compared with the standard sample to learn about the upcoming risk deformation, making the risk assessment of the flexible photovoltaic support predictable.

Description

Technical Field

[0001] The present application relates to the technical field of electrical digital data processing, and in particular to a risk assessment method, device, equipment and storage medium for a flexible photovoltaic support. Background Art

[0002] ‌Flexible photovoltaic support is a large-span photovoltaic module support structure designed based on a tension structure system and formed by cables as component support members. ‌Compared with traditional rigid supports, it has the characteristics of "large span, high clearance, and long column distance", which enable flexible supports to adapt to more complex and diverse installation environments. ‌The structure of the flexible photovoltaic support adopts prestressed chords (wire ropes) between the two fixed points at both ends. The two fixed points at both ends adopt a rigid structure and an outer inclined steel strand to provide support. It can achieve a large span of 10~30 m and adapt to situations such as undulating mountains and increased vegetation. It only needs to set the foundation at a suitable location and tension the prestressed steel strand or wire rope. Under the condition that the water level remains unchanged, the construction of rigid columns, foundations and flexible supports can be realized in lakes and fish ponds.

[0003] The mainstream flexible photovoltaic brackets are single-layer cable suspension structure, double-layer cable truss structure, fish-belly cable truss structure, and string structure. Among them, the single-layer cable suspension structure is widely used due to its easy installation and easy subsequent maintenance. The single-layer cable suspension structure is generally composed of a main steel frame composed of beams and columns, cable bodies and other main components. The cable body is two cables parallel to the plane of the component. The cable body is used instead of a general tension member. After the component support cable is tensioned, it is fixed by the anchor at the end of the steel beam. The tensioning equipment makes the support cable obtain stress stiffness for supporting the component, and forms a self-balancing system by relying on the end of the cable.

[0004] Since flexible photovoltaic bracket technology only began to appear more than ten years ago, due to early technical and material limitations, the related technology of flexible photovoltaic bracket started late and the market development was relatively slow. Currently, the research and development of flexible photovoltaic brackets is still in the hardware design and optimization stage. The risk assessment of flexible photovoltaic brackets is relatively lacking, and there is currently no clear risk assessment method. Summary of the invention

[0005] The main purpose of the present application is to provide a risk assessment method, device, equipment and storage medium for a flexible photovoltaic bracket, so as to solve the problem that the risk assessment of flexible photovoltaic brackets in the prior art is relatively insufficient and there is currently no clear risk assessment method.

[0006] In order to achieve the above objectives, this application provides the following technical solutions:

[0007] A risk assessment method for a flexible photovoltaic support, the flexible photovoltaic support comprising at least two steel beams fixed to the ground, upper and lower chords erected on adjacent steel beams and in different planes from each other, and a plurality of photovoltaic panels installed on the upper and lower chords, the risk assessment method comprising:

[0008] Step S1, acquiring first standard image data and second standard image data of the photovoltaic panel under windless conditions, wherein a shooting center line of the first standard image data is collinear with a perpendicular midline of the upper chord and the lower chord and is perpendicular to the panel surfaces of all photovoltaic panels, and a shooting center line of the second standard image data is perpendicular to the shooting center lines of all first image data and is parallel to the panel surfaces of all photovoltaic panels;

[0009] Step S2, acquiring a plurality of first image data and a plurality of second image data of the photovoltaic panel based on a preset time period, wherein shooting constraints of all the first image data and all the second image data are respectively the same as shooting constraints of the first standard image data and the second standard image data;

[0010] Step S3, dividing all first image data into a first training set and a first sample set according to a first preset ratio, and dividing all second image data into a second training set and a second sample set according to a second preset ratio;

[0011] Step S4, outputting the first training set to a preset neural network model and training it to obtain a first prediction model, and outputting the second training set to the preset neural network model and training it to obtain a second prediction model;

[0012] Step S5, outputting the first sample set to the first prediction model to obtain first predicted image data, and outputting the second sample set to the second prediction model to obtain second predicted image data;

[0013] Step S6, obtaining a first degree of coincidence between the first predicted image data and the first standard image data, and a second degree of coincidence between the second predicted image data and the second standard image data by using a boundary acquisition algorithm;

[0014] Step S7, determining whether the first overlap degree and the second overlap degree are both less than a preset percentage threshold based on the same preset time period, and if the first overlap degree and the second overlap degree are both less than the preset percentage threshold, executing step S8;

[0015] Step S8, determining whether the flexible photovoltaic bracket has undergone risky deformation.

[0016] As a further improvement of the present application, step S3, dividing all first image data into a first training set and a first sample set according to a first preset ratio, and dividing all second image data into a second training set and a second sample set according to a second preset ratio, includes:

[0017] Step S31, obtaining a first minimum bounding box based on the photovoltaic panel in the first image data and a second minimum bounding box based on the photovoltaic panel in the second image data by using a target detection algorithm;

[0018] Step S32, dividing all first minimum bounding boxes into the first training set and the first sample set according to the first preset ratio;

[0019] Step S33: Divide all second minimum bounding boxes into the second training set and the second sample set according to the second preset ratio.

[0020] As a further improvement of the present application, step S4, outputting the first training set to a preset neural network model and training it to obtain a first prediction model, and outputting the second training set to the preset neural network model and training it to obtain a second prediction model, includes:

[0021] Step S41, defining a topological relationship of the preset neural network model, wherein the topological relationship includes an input layer, a hidden layer, and an output layer that are sequentially signal-connected;

[0022] Step S42, outputting the first training set to the input layer, training the first training set a first preset number of times by using a least squares estimation method, and obtaining a root mean square error between the first sample set and the training result based on each training;

[0023] Step S43, obtaining a first minimum value among all root mean square errors, and obtaining a training result corresponding to the first minimum value as the first prediction model;

[0024] Step S44, outputting the second training set to the input layer, training the second training set a second preset number of times using the least squares estimation method, and obtaining a root mean square error between the second sample set and the training result based on each training;

[0025] Step S45, obtaining a second minimum value in the root mean square error, and obtaining a training result corresponding to the second minimum value as the second prediction model.

[0026] As a further improvement of the present application, step S6, obtaining a first degree of coincidence between the first predicted image data and the first standard image data, and a second degree of coincidence between the second predicted image data and the second standard image data through a boundary acquisition algorithm, includes:

[0027] Step S61, binarizing the first standard image data to obtain first black-and-white standard image data;

[0028] Step S62, dividing the first black-and-white standard image data into a plurality of grids using a grid of a preset density;

[0029] Step S63, respectively obtaining the color of each grid, and assigning a value of 0 to a grid with a black color and a value of 1 to a grid with a white color;

[0030] Step S64, defining the current grid and the surrounding eight grids as a 3×3 window, and traversing all grids through the 3×3 window with one grid as a step;

[0031] Step S65, based on all traversal results, the grids assigned with all values ​​of 1 and all values ​​of 0 in the 3×3 window are deleted, and the remaining grids are defined as the first standard boundary;

[0032] Step S66, repeating steps S61 to S65 with the second standard image data as the main body to obtain a second standard boundary;

[0033] Step S67, repeating steps S61 to S65 based on the first predicted image data to obtain a first predicted boundary;

[0034] Step S68, repeating steps S61 to S65 based on the second predicted image data to obtain a second predicted boundary;

[0035] Step S69: obtain the area ratio of the first predicted boundary to the first standard boundary and define it as the first overlap, and obtain the area ratio of the second predicted boundary to the second standard boundary and define it as the second overlap.

[0036] As a further improvement of the present application, step S7, based on the same preset time period, determines whether the first overlap degree and the second overlap degree are both less than a preset percentage threshold, if the first overlap degree and the second overlap degree are both less than the preset percentage threshold, then executes step S8, including:

[0037] Step S71, predicting a plurality of first predicted image data based on a plurality of different preset step numbers to obtain a plurality of first coincidence degrees;

[0038] Step S72, predicting a plurality of second predicted image data based on a plurality of different preset step numbers to obtain a plurality of second coincidence degrees;

[0039] Step S73, defining at least two consecutive percentage intervals based on the preset percentage threshold;

[0040] Step S74, assigning risk coefficients from high to low based on all percentage intervals from small to large;

[0041] Step S75, classifying all first coincidence degrees into all percentage intervals, and obtaining the risk coefficient of each first coincidence degree respectively;

[0042] Step S76, weighted calculation of risk coefficients of all first coincidence degrees to obtain a first comprehensive risk;

[0043] Step S77, classifying all second coincidence degrees into all percentage intervals, and obtaining the risk coefficient of each second coincidence degree respectively;

[0044] Step S78, weighted calculation of risk coefficients of all second coincidence degrees to obtain a second comprehensive risk;

[0045] Step S79, based on the same preset time period, determine whether the first comprehensive risk and the second comprehensive risk are both less than the preset percentage threshold; if the first comprehensive risk and the second comprehensive risk are both less than the preset percentage threshold, execute step S8.

[0046] As a further improvement of the present application, in step S75, all first overlaps are classified into all percentage intervals, and the risk coefficient of each first overlap is obtained respectively, including:

[0047] Step S751: Integrate all first overlaps into a data set to be classified , For the The first degree of coincidence, is the number of all first degree of coincidence;

[0048] Step S752: define a category set based on the number of all percentage intervals ,in, For the category set The Percentage interval, is the number of all percentage intervals;

[0049] Step S753, calculate the first The conditional probability of the first coincidence degree in each percentage interval is:

[0050] (1);

[0051] in, For all first coincidences in The conditional probability of a percentage interval; For the The marginal probability of a percentage interval; For the In the percentage interval The conditional probability of the first degree of coincidence;

[0052] Step S754, classifying each first coincidence degree into the percentage interval with the highest conditional probability;

[0053] Step S755, respectively obtain the risk coefficient corresponding to each percentage interval with the highest first coincidence degree conditional probability.

[0054] As a further improvement of the present application, in step S77, all second coincidence degrees are classified into all percentage intervals, and the risk coefficient of each second coincidence degree is obtained respectively, including:

[0055] Step S771, repeating steps S751 to S755 based on all the second coincidence degrees, so as to respectively obtain the risk coefficient corresponding to the percentage interval with the highest conditional probability of each second coincidence degree.

[0056] In order to achieve the above objectives, this application also provides the following technical solutions:

[0057] A risk assessment device for a flexible photovoltaic support, the risk assessment device is applied to the risk assessment method as described above, and the risk assessment device comprises:

[0058] A standard image data acquisition module is used to acquire first standard image data and second standard image data of the photovoltaic panel under windless conditions, wherein a shooting center line of the first standard image data is collinear with a perpendicular midline of the upper chord and the lower chord and is perpendicular to the panel surfaces of all photovoltaic panels, and a shooting center line of the second standard image data is perpendicular to the shooting center lines of all first image data and is parallel to the panel surfaces of all photovoltaic panels;

[0059] An image data acquisition module, used for acquiring a plurality of first image data and a plurality of second image data of the photovoltaic panel based on a preset time period, wherein shooting constraints of all the first image data and all the second image data are respectively the same as shooting constraints of the first standard image data and the second standard image data;

[0060] A training set and sample set division module, used to divide all first image data into a first training set and a first sample set according to a first preset ratio, and divide all second image data into a second training set and a second sample set according to a second preset ratio;

[0061] A prediction model acquisition module, used to output the first training set to a preset neural network model and perform training to obtain a first prediction model, and output the second training set to the preset neural network model and perform training to obtain a second prediction model;

[0062] A predicted image data acquisition module, configured to output the first sample set to the first prediction model to obtain first predicted image data, and output the second sample set to the second prediction model to obtain second predicted image data;

[0063] a coincidence acquisition module, configured to acquire a first coincidence between the first predicted image data and the first standard image data, and a second coincidence between the second predicted image data and the second standard image data through a boundary acquisition algorithm;

[0064] A coincidence determination module, used for determining whether the first coincidence degree and the second coincidence degree are both less than a preset percentage threshold value based on the same preset time period;

[0065] The risk deformation determination module is used to determine that the flexible photovoltaic support has a risk deformation if both the first overlap degree and the second overlap degree are less than a preset percentage threshold.

[0066] In order to achieve the above objectives, this application also provides the following technical solutions:

[0067] An electronic device includes a processor and a memory coupled to the processor, wherein the memory stores program instructions executable by the processor; when the processor executes the program instructions stored in the memory, the risk assessment method for a flexible photovoltaic bracket as described above is implemented.

[0068] In order to achieve the above objectives, this application also provides the following technical solutions:

[0069] A storage medium stores program instructions, and when the program instructions are executed by a processor, the risk assessment method for a flexible photovoltaic support as described above can be implemented.

[0070] The present application acquires first standard image data and second standard image data of a photovoltaic panel under windless conditions, wherein a shooting centerline of the first standard image data is collinear with the perpendicular midline of the upper chord and the lower chord and is perpendicular to the panel surfaces of all photovoltaic panels, and a shooting centerline of the second standard image data is perpendicular to the shooting centerlines of all first image data and is parallel to the panel surfaces of all photovoltaic panels; acquires a plurality of first image data and a plurality of second image data of the photovoltaic panel based on a preset time period, wherein shooting constraints of all first image data and all second image data are respectively the same as shooting constraints of the first standard image data and the second standard image data; divides all first image data into a first training set and a first sample set according to a first preset ratio, and divides all second image data into a second training set according to a second preset ratio. training set and second sample set; output the first training set to the preset neural network model and train it to obtain the first prediction model, and output the second training set to the preset neural network model and train it to obtain the second prediction model; output the first sample set to the first prediction model to obtain the first prediction image data, and output the second sample set to the second prediction model to obtain the second prediction image data; obtain the first overlap between the first prediction image data and the first standard image data, and the second overlap between the second prediction image data and the second standard image data through the boundary acquisition algorithm; based on the same preset time period, determine whether the first overlap and the second overlap are both less than the preset percentage threshold, if the first overlap and the second overlap are both less than the preset percentage threshold, then determine that the flexible photovoltaic bracket has risk deformation. This application calibrates the photovoltaic panel morphology in a windless state as a standard sample, provides an evaluation basis for the real-time state in the subsequent use process, avoids using the photovoltaic panel in the use process as a sample reference (already has vibration characteristics or risk deformation characteristics) to cause inaccurate evaluation, and then compares the photovoltaic prediction morphology obtained by machine learning with the standard sample to learn about the upcoming risk deformation, so that the risk evaluation of the flexible photovoltaic bracket is predictable. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 A schematic diagram of the steps of an embodiment of a risk assessment method for a flexible photovoltaic support of the present application;

[0072] Figure 2 This is a functional module diagram of an embodiment of a risk assessment device for a flexible photovoltaic support of the present application;

[0073] Figure 3 This is a schematic diagram of the structure of an embodiment of the electronic device of the present application;

[0074] Figure 4 This is a schematic diagram of the structure of an embodiment of the storage medium of the present application. DETAILED DESCRIPTION

[0075] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0076] The terms "first", "second" and "third" in this application are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined as "first", "second" and "third" can explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. All directional indications (such as up, down, left, right, front, back...) in the embodiments of this application are only used to explain the relative position relationship, movement, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication also changes accordingly. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or devices.

[0077] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0078] like Figure 1 As shown, this embodiment provides an embodiment of a risk assessment method for a flexible photovoltaic bracket. In this embodiment, the flexible photovoltaic bracket includes at least two steel beams fixed to the ground, upper and lower chords erected on adjacent steel beams and in opposite planes, and a plurality of photovoltaic panels installed on the upper and lower chords.

[0079] Preferably, the upper and lower chords also need two anchor cables for fixing. One end of each anchor cable is anchored to the ground around the steel beam frame, and the other end of each anchor cable is fixedly connected to the upper or lower chord cable. The upper, lower and anchor cables all use steel strands with a diameter of 15.2 mm. The photovoltaic panels are then installed on the upper and lower chord cables. Two steel beam frames form one span, and 16 photovoltaic panels are generally installed in each span. The interval between photovoltaic panels is 30 mm, and the interval between photovoltaic panels in the middle of the span is 300 mm. A single photovoltaic panel is 2278 mm long, 1134 mm wide, 30 mm thick, and has an inclination of 15°.

[0080] Specifically, the risk assessment method includes the following steps:

[0081] Step S1, acquiring the first standard image data and the second standard image data of the photovoltaic panel under windless conditions, wherein the shooting center line of the first standard image data is collinear with the perpendicular bisectors of the upper chord and the lower chord and is perpendicular to the surfaces of all photovoltaic panels, and the shooting center line of the second standard image data is perpendicular to the shooting center lines of all the first image data and is parallel to the surfaces of all photovoltaic panels.

[0082] Preferably, since the photovoltaic panels have an inclination angle of 15° and the panels of each span are not necessarily coplanar during the actual installation process, the photovoltaic panel at the mid-span position or the two photovoltaic panels closest to the mid-span position can be taken as reference photovoltaic panels to participate in subsequent image acquisition.

[0083] Preferably, the first standard image data is an image directly taken based on the photovoltaic panel face angle under windless conditions, and the second standard image data is an image directly taken based on the photovoltaic panel side angle under windless conditions. Since the shooting angle is fixed, the data training below is more inclined to the pixel coordinate position of the photovoltaic panel in the image, and since the image taken under windless conditions does not shake, it is used as a standard for subsequent calibration. The windless conditions can be simulated by building a scene indoors and shooting, or directly simulated by software.

[0084] Step S2, acquiring a plurality of first image data and a plurality of second image data of the photovoltaic panel based on a preset time period, wherein shooting constraints of all the first image data and all the second image data are respectively the same as the shooting constraints of the first standard image data and the second standard image data.

[0085] Preferably, the first image data is an image directly captured based on the photovoltaic panel surface angle during actual use (windy conditions), and the shooting angle is the same as the shooting angle of the first standard image data; the second image data is an image directly captured based on the side angle of the photovoltaic panel during actual use (windy conditions), and the shooting angle is the same as the shooting angle of the second standard image data. At this time, the first image data is compared with the first standard image data, so that the first standard image data can calibrate the first image data; the second image data is compared with the second standard image data, so that the second standard image data can calibrate the second image data.

[0086] It is worth noting that the shooting angle of view and shooting distance of the first image data need to be the same as those of the first standard image data, and the same is true for the second image data and the second standard image data.

[0087] Step S3: divide all first image data into a first training set and a first sample set according to a first preset ratio, and divide all second image data into a second training set and a second sample set according to a second preset ratio.

[0088] Preferably, the first preset ratio may be the same as the second preset ratio, so as to divide the image data into a training set and a sample set in a ratio of 8:2.

[0089] Preferably, in actual application, a certain proportion of validation sets are required to verify the accuracy of the model, that is, verification is required after training is completed, and the sample set prediction is started only after the verification is successful. Usually, the image data is divided into training set, validation set, and sample set in a ratio of 70%:15%:15%, that is, 70% of the data is the training set, 15% of the data is the validation set, and 15% of the data is the sample set.

[0090] Step S4: output the first training set to a preset neural network model and perform training to obtain a first prediction model, and output the second training set to a preset neural network model and perform training to obtain a second prediction model.

[0091] Step S5: output the first sample set to the first prediction model to obtain first predicted image data, and output the second sample set to the second prediction model to obtain second predicted image data.

[0092] Preferably, the image data is specifically the coordinate value of each pixel.

[0093] Step S6, obtaining a first degree of coincidence between the first predicted image data and the first standard image data, and a second degree of coincidence between the second predicted image data and the second standard image data through a boundary obtaining algorithm.

[0094] Preferably, the overlap degree can be set as an area ratio.

[0095] Step S7, based on the same preset time period, determine whether the first overlap degree and the second overlap degree are both less than the preset percentage threshold. If the first overlap degree and the second overlap degree are both less than the preset percentage threshold, execute step S8.

[0096] Step S8, determining whether the flexible photovoltaic bracket has undergone risky deformation.

[0097] Furthermore, step S3, dividing all first image data into a first training set and a first sample set according to a first preset ratio, and dividing all second image data into a second training set and a second sample set according to a second preset ratio, includes:

[0098] Step S31 , obtaining a first minimum bounding box based on the photovoltaic panel in the first image data and a second minimum bounding box based on the photovoltaic panel in the second image data by using a target detection algorithm.

[0099] Preferably, the target detection algorithm can be implemented by target detection algorithms VJ, HOG, DPMDetector; deep learning two-stage target detection algorithms RCNN, SPPNet, FastRCNN, FasterRCNN; target detection trick algorithms FPN, CascadeRCNN; deep learning one-stage target detection algorithms Yolo, X, SSD, RetinaNet; deep learning anchor-free target detection algorithms CornerNet, CenterNet, FCOS; Transformer-based target detection algorithm DETR, etc.

[0100] Preferably, the minimum bounding box can be set as an AABB box or an OBB box. In this embodiment, the OBB box is preferred to reduce the blank space of the detection frame.

[0101] Step S32: Divide all first minimum bounding boxes into a first training set and a first sample set according to a first preset ratio.

[0102] Step S33: Divide all second minimum bounding boxes into a second training set and a second sample set according to a second preset ratio.

[0103] Further, step S4, outputting the first training set to a preset neural network model and training it to obtain a first prediction model, and outputting the second training set to a preset neural network model and training it to obtain a second prediction model, including:

[0104] Step S41, defining the topological relationship of the preset neural network model, the topological relationship includes an input layer, a hidden layer, and an output layer that are sequentially signal-connected.

[0105] Preferably, the preset neural network model is represented by the following formula:

[0106] .

[0107] in, To preset the neural network model; is the input layer and , The input layer input nodes, each input node corresponds to a first training set or a second training set, The input layer The first input node to the hidden layer The weights of the input nodes; is connected to the hidden layer The bias of the input node; is the transfer function, and .

[0108] Preferably, the numbers in the brackets of the symbol superscripts in formula (1) of this embodiment are the number of layers, for example, superscript (1) is the first layer, that is, the input layer, and superscript (1, 2) is from the first layer to the second layer, that is, the input layer to the hidden layer.

[0109] It should be noted that the above formula is a principle explanation, and its symbolic meaning is not interchangeable with other formulas in this embodiment.

[0110] Preferably, in the actual training process, the image data needs to be normalized to prevent the large difference in data magnitude during the training process from resulting in a lack of significance. In this embodiment, the image data can be first coordinate-converted, and the boundary of the image data can be defined as 0 to 1 (in this case, the image data can be cut into squares at the same position), so that the coordinates of the four corner points of the minimum bounding box can be obtained, and the coordinate values ​​are all between 0 and 1. Training can be performed based on the four coordinate values ​​of a single minimum bounding box.

[0111] Step S42: output the first training set to the input layer, train the first training set for a first preset number of times using a least squares estimation method, and obtain a root mean square error between the first sample set and the training result based on each training.

[0112] Preferably, the root mean square error is characterized by the following formula:

[0113] .

[0114] in, is the root mean square error, is the number of the first training set or the number of the second training set, For the The true value of the first training set or The true value of the second training set, For the The first training set or The training results after the second training set training is completed.

[0115] It should be noted that the above formula is a principle explanation, and its symbolic meaning is not interchangeable with other formulas in this embodiment.

[0116] Preferably, the training model training a neural network usually requires providing a large amount of data, namely a data set; the data set is generally divided into three categories, namely the above-mentioned training set (training set), validation set (validation set) and test set (test set).

[0117] Among them, one epoch is the process of training once with all the samples in the training set. The so-called training once refers to one forward pass and one back pass. When the number of samples in an epoch (i.e., training set) is too large, training once may consume too much time, and it is not necessary to use all the data in the training set for each training. In this case, the entire training set needs to be divided into multiple small blocks, that is, divided into multiple batches for training. An epoch consists of one or more batches, where a batch is a part of the training set. Each training process only uses a part of the data, i.e., a batch. The process of training a batch is an iteration.

[0118] Preferably, the neural network training specifically includes a perceptron, which is composed of two layers of neurons. The input layer receives external input signals and transmits them to the output layer. The output layer is MP neurons, and the step function is .

[0119] Preferably, given a training data set, the weights ( =1,2,...,n) and training bias It can be obtained through learning, It can be understood as the weight corresponding to a fixed value of -1,0. .

[0120] It should be noted that the step function and weights here This is for the purpose of explanation of the principle and has no interchangeable symbolic meaning with other formulas in the embodiment. This step function is only for the purpose of explanation of the principle and does not participate in the calculation of other formulas.

[0121] Preferably, in this embodiment, the number of neural network training times can be set to 1000 times.

[0122] Preferably, the learning rate from the 1st to the 500th epoch can be set to 0.01, the learning rate from the 501st to the 750th epoch can be set to 0.001, and the learning rate from the 751st to the 1000th epoch can be set to 0.0001.

[0123] It can be understood that the neural network training of this embodiment mainly includes the following ideas:

[0124] ① Initialize the weights and bias items in the network.

[0125] Initializing parameter values ​​(output unit weights, bias terms and hidden unit weights, bias terms are all model parameters) is to activate forward propagation, obtain the output value of each layer element, and then obtain the value of the loss function.

[0126] ②Activate forward propagation to obtain the output value of each layer and the expected value of the loss function of each layer.

[0127] ③According to the loss function, calculate the error term of the output unit and the error term of the hidden unit.

[0128] Calculate various errors, calculate the gradient of the parameters with respect to the loss function, or calculate partial derivatives according to the chain rule of calculus. For partial derivatives of vectors or matrices in a composite function, the partial derivative of the function inside the composite function is always multiplied on the left; for partial derivatives of scalars in a composite function, the partial derivative of the function inside the composite function can be multiplied on the left or on the right.

[0129] ④Update the weights and bias items in the neural network.

[0130] ⑤ Repeat ②~④ until the loss function is less than the preset bias or the number of iterations is used up, and the parameters output at this time are the current optimal parameters.

[0131] Step S43, obtaining the first minimum value among all root mean square errors, and obtaining the training result corresponding to the first minimum value as the first prediction model.

[0132] Step S44, outputting the second training set to the input layer, training the second training set for a second preset number of times using a least squares estimation method, and obtaining a root mean square error between the second sample set and the training result based on each training.

[0133] Preferably, the least squares estimation method is a nonlinear least squares method, which is used to solve nonlinear optimization problems, such as the neural network model in this embodiment. The method uses the gradient information of the objective function to adjust the parameters in each iteration to find the best model parameter estimate. The method uses an adjustment factor called least squares estimation, which balances the gradient descent and Gauss-Newton method according to the current iteration step. In each iteration, the method calculates a parameter increment and then decides whether to accept the parameter increment by comparing the size of the current model residual and the model residual estimated using the new parameter.

[0134] Preferably, the iteration step size of the least squares estimation method is given by Representation.

[0135] in, is the iteration step size of the least squares estimation method, is the current model residual, is the Jacobian matrix of the current model residual, is the transposed matrix of the Jacobian matrix of the current model residual, is the least squares estimated adjustment factor mentioned above, is the identity matrix, is the covariance matrix.

[0136] Preferably, the order of the identity matrix is ​​the same as the order of the Jacobian matrix.

[0137] Preferably, when hour, is a positive number, Prefer the Gauss-Newton method; when hour, Tends to balance gradient descent.

[0138] It should be noted that the additional content of the least squares estimation method mentioned above is a principle explanation, and its symbolic meaning is not interchangeable with other formulas in this implementation.

[0139] Step S45, obtaining a second minimum value in the root mean square error, and obtaining a training result corresponding to the second minimum value as a second prediction model.

[0140] Furthermore, step S6, obtaining a first degree of coincidence between the first predicted image data and the first standard image data, and a second degree of coincidence between the second predicted image data and the second standard image data through a boundary acquisition algorithm, includes:

[0141] Step S61 , binarizing the first standard image data to obtain first black-and-white standard image data.

[0142] Step S62, dividing the first black-and-white standard image data into a plurality of grids using a grid of a preset density.

[0143] Step S63, respectively obtaining the color of each grid, and assigning a value of 0 to a grid with a black color and a value of 1 to a grid with a white color.

[0144] Step S64, define the current grid and the surrounding eight grids as a 3×3 window, and traverse all grids through the 3×3 window with one grid as a step.

[0145] Step S65 , based on all traversal results, the grids in the 3×3 window that are all assigned values ​​of 1 and all assigned values ​​of 0 are deleted, and the remaining grids are defined as the first standard boundary.

[0146] Step S66, repeating steps S61 to S65 based on the second standard image data to obtain a second standard boundary.

[0147] Step S67, repeating steps S61 to S65 based on the first predicted image data to obtain a first predicted boundary.

[0148] Step S68, repeating steps S61 to S65 based on the second predicted image data to obtain a second predicted boundary.

[0149] Step S69, obtaining an area ratio of the first predicted boundary to the first standard boundary and defining it as a first overlap, and obtaining an area ratio of the second predicted boundary to the second standard boundary and defining it as a second overlap.

[0150] Further, step S7, based on the same preset time period, determines whether the first overlap degree and the second overlap degree are both less than a preset percentage threshold. If the first overlap degree and the second overlap degree are both less than the preset percentage threshold, step S8 is executed, including:

[0151] Step S71 , predicting a plurality of first predicted image data based on a plurality of different preset step numbers to obtain a plurality of first coincidence degrees.

[0152] Step S72: predicting a plurality of second predicted image data based on a plurality of different preset step numbers to obtain a plurality of second coincidence degrees.

[0153] Step S73: defining at least two consecutive percentage intervals based on a preset percentage threshold.

[0154] Step S74, assigning risk coefficients from high to low based on all percentage intervals from small to large.

[0155] Step S75: classify all first overlap degrees into all percentage intervals, and obtain the risk coefficient of each first overlap degree respectively.

[0156] Step S76, weighted calculation of risk coefficients of all first overlap degrees to obtain a first comprehensive risk.

[0157] Step S77: classify all second overlaps into all percentage intervals, and obtain the risk coefficient of each second overlap respectively.

[0158] Step S78, weighted calculation of the risk coefficients of all second coincidence degrees to obtain a second comprehensive risk.

[0159] Step S79, based on the same preset time period, determine whether the first comprehensive risk and the second comprehensive risk are both less than the preset percentage threshold. If the first comprehensive risk and the second comprehensive risk are both less than the preset percentage threshold, execute step S8.

[0160] Furthermore, in step S75, all first overlaps are classified into all percentage intervals, and the risk coefficient of each first overlap is obtained respectively, including:

[0161] Step S751: Integrate all first overlaps into a data set to be classified , For the The first degree of coincidence, is the number of all first degrees of coincidence.

[0162] Step S752: define a category set based on the number of all percentage intervals ,in, For category collection The Percentage interval, is the number of all percentage intervals.

[0163] Step S753, calculate the first The conditional probability of the first coincidence degree in each percentage interval is:

[0164] (1).

[0165] in, For all first coincidences in The conditional probability of a percentage interval; For the The marginal probability of a percentage interval; For the In the percentage interval The conditional probability of the first degree of coincidence.

[0166] Step S754: classify each first degree of overlap into a percentage interval with the highest conditional probability.

[0167] Step S755, respectively obtain the risk coefficient corresponding to each percentage interval with the highest first coincidence degree conditional probability.

[0168] Furthermore, in step S77, all second coincidence degrees are classified into all percentage intervals, and the risk coefficient of each second coincidence degree is obtained respectively, including:

[0169] Step S771, repeating steps S751 to S755 based on all the second coincidence degrees, so as to respectively obtain the risk coefficient corresponding to the percentage interval with the highest conditional probability of each second coincidence degree.

[0170] Preferably, the classification algorithm of this embodiment is preferably naive Bayes classification.

[0171] Preferably, the definition of Naive Bayes is as follows:

[0172] ① Let x={a1, a2, a3,…, an} be an item to be classified, and each a is a feature of x.

[0173] ②There is a category set c={y1, y2, y3, ..., y m}.

[0174] ③ Calculate P(y1|x), P(y2|x), ..., P(y m |x).

[0175] ④If P(y k |x)=max{P(y1|x),P(y2|x),...,P(y m |x)}, then x∈y k .

[0176] Then calculate the conditional probabilities in step ③ through the following steps:

[0177] 1. Find a set of items to be classified with known classifications. This set is called a training sample set.

[0178] 2. Statistically obtain the conditional probability estimate of each feature attribute under each category. That is:

[0179] P(a1|y1),P(a2|y1),……,P(a n |y1)

[0180] P(a1|y2),P(a2|y2),……,P(a n |y2);

[0181] …

[0182] P(a1|y m),P(a2|y m ),……,P(a n |y m );

[0183] 3. Assuming that each feature attribute is conditionally independent, according to the Bayesian principle:

[0184] P(y i |x)=P(x|y i )P(y i ) / p(x).

[0185] Since the denominator is a constant for all categories, we only need to maximize the numerator. And because each feature attribute is conditionally independent, then:

[0186] P(x|y i )P(y i )=P(a1|y i )P(a2|y i )……P(a n |y i )P(y i ).

[0187] It should be noted that the above preferred contents are also explanations of the principles, and the meanings of their symbols are not interchangeable with the meanings of the symbols of other formulas in this embodiment.

[0188] In this embodiment, the first standard image data and the second standard image data of the photovoltaic panel are obtained under windless conditions, the shooting center line of the first standard image data is collinear with the perpendicular midline of the upper chord and the lower chord and is perpendicular to the panel surfaces of all photovoltaic panels, and the shooting center line of the second standard image data is perpendicular to the shooting center lines of all the first image data and is parallel to the panel surfaces of all photovoltaic panels; a plurality of first image data and a plurality of second image data of the photovoltaic panel are obtained based on a preset time period, and the shooting constraints of all the first image data and all the second image data are respectively the same as the shooting constraints of the first standard image data and the second standard image data; all the first image data are divided into a first training set and a first sample set according to a first preset ratio, and all the second image data are divided into a second training set according to a second preset ratio. training set and a second sample set; output the first training set to a preset neural network model and train it to obtain a first prediction model, and output the second training set to a preset neural network model and train it to obtain a second prediction model; output the first sample set to the first prediction model to obtain first prediction image data, and output the second sample set to the second prediction model to obtain second prediction image data; obtain a first overlap between the first prediction image data and the first standard image data, and a second overlap between the second prediction image data and the second standard image data through a boundary acquisition algorithm; determine whether the first overlap and the second overlap are both less than a preset percentage threshold based on the same preset time period, and if the first overlap and the second overlap are both less than the preset percentage threshold, it is determined that the flexible photovoltaic bracket has risk deformation. In this embodiment, the photovoltaic panel shape in a windless state is calibrated as a standard sample, providing an evaluation basis for the real-time state in the subsequent use process, avoiding the photovoltaic panel in the use process as a sample reference (already having vibration characteristics or risk deformation characteristics) to cause inaccurate evaluation, and then comparing the photovoltaic prediction shape obtained by machine learning with the standard sample to learn about the upcoming risk deformation, so that the risk evaluation of the flexible photovoltaic bracket is predictable.

[0189] like Figure 2 As shown, this embodiment provides an embodiment of a risk assessment device for a flexible photovoltaic support. In this embodiment, the risk assessment device is applied to the risk assessment method in the above embodiment.

[0190] Specifically, the risk assessment device includes a standard image data acquisition module 1, an image data acquisition module 2, a training set and sample set division module 3, a prediction model acquisition module 4, a prediction image data acquisition module 5, an overlap acquisition module 6, an overlap judgment module 7, and a risk deformation judgment module 8, which are electrically connected in sequence.

[0191] Among them, the standard image data acquisition module 1 is used to acquire the first standard image data and the second standard image data of the photovoltaic panel under windless conditions, the shooting center line of the first standard image data is collinear with the perpendicular midline of the upper chord and the lower chord and is perpendicular to the panel surface of all photovoltaic panels, and the shooting center line of the second standard image data is perpendicular to the shooting center line of all the first image data and is parallel to the panel surface of all photovoltaic panels; the image data acquisition module 2 is used to acquire a plurality of first image data and a plurality of second image data of the photovoltaic panel based on a preset time period, and the shooting constraints of all the first image data and all the second image data are respectively the same as the shooting constraints of the first standard image data and the second standard image data; the training set and sample set division module 3 is used to divide all the first image data into a first training set and a first sample set according to a first preset ratio, and divide all the second image data into a second training set and a second sample set according to a second preset ratio; The prediction model acquisition module 4 is used to output the first training set to the preset neural network model and train it to obtain the first prediction model, and output the second training set to the preset neural network model and train it to obtain the second prediction model; the prediction image data acquisition module 5 is used to output the first sample set to the first prediction model to obtain the first prediction image data, and output the second sample set to the second prediction model to obtain the second prediction image data; the coincidence acquisition module 6 is used to obtain the first coincidence between the first prediction image data and the first standard image data, and the second coincidence between the second prediction image data and the second standard image data through a boundary acquisition algorithm; the coincidence judgment module 7 is used to judge whether the first coincidence and the second coincidence are both less than the preset percentage threshold based on the same preset time period; the risk deformation judgment module 8 is used to judge whether the flexible photovoltaic bracket has risk deformation if the first coincidence and the second coincidence are both less than the preset percentage threshold.

[0192] Furthermore, the training set and sample set division module 3 specifically includes a first training set and sample set division submodule, a second training set and sample set division submodule, and a third training set and sample set division submodule, which are electrically connected in sequence; the first training set and sample set division submodule is electrically connected to the image data acquisition module 2, and the third training set and sample set division submodule is electrically connected to the prediction model acquisition module 4.

[0193] Among them, the first training set and sample set division submodule is used to obtain the first minimum bounding box based on the photovoltaic panel in the first image data and the second minimum bounding box based on the photovoltaic panel in the second image data through the target detection algorithm; the second training set and sample set division submodule is used to divide all the first minimum bounding boxes into the first training set and the first sample set according to the first preset ratio; the third training set and sample set division submodule is used to divide all the second minimum bounding boxes into the second training set and the second sample set according to the second preset ratio.

[0194] Furthermore, the prediction model acquisition module 4 specifically includes a first prediction model acquisition submodule, a second prediction model acquisition submodule, a third prediction model acquisition submodule, a fourth prediction model acquisition submodule, and a fifth prediction model acquisition submodule, which are electrically connected in sequence; the first prediction model acquisition submodule is electrically connected to the third training set and sample set division submodule, and the fifth prediction model acquisition submodule is electrically connected to the prediction image data acquisition module 5.

[0195] Among them, the first prediction model acquisition submodule is used to define the topological relationship of the preset neural network model, and the topological relationship includes an input layer, a hidden layer, and an output layer that are signal-connected in sequence; the second prediction model acquisition submodule is used to output the first training set to the input layer, train the first training set for a first preset number of times through the least squares estimation method, and obtain the root mean square error between the first sample set and the training result based on each training; the third prediction model acquisition submodule is used to obtain the first minimum value among all root mean square errors, and obtain the training result corresponding to the first minimum value as the first prediction model; the fourth prediction model acquisition submodule is used to output the second training set to the input layer, train the second training set for a second preset number of times through the least squares estimation method, and obtain the root mean square error between the second sample set and the training result based on each training; the fifth prediction model acquisition submodule is used to obtain the second minimum value in the root mean square error, and obtain the training result corresponding to the second minimum value as the second prediction model.

[0196] Furthermore, the overlap acquisition module 6 specifically includes a first overlap acquisition submodule, a second overlap acquisition submodule, a third overlap acquisition submodule, a fourth overlap acquisition submodule, a fifth overlap acquisition submodule, a sixth overlap acquisition submodule, a seventh overlap acquisition submodule, and an eighth overlap acquisition submodule, which are electrically connected in sequence; the first overlap acquisition submodule is electrically connected to the predicted image data acquisition module 5, and the eighth overlap acquisition submodule is electrically connected to the overlap judgment module 7.

[0197] Among them, the first coincidence acquisition submodule is used to binarize the first standard image data to obtain the first black-and-white standard image data; the second coincidence acquisition submodule is used to divide the first black-and-white standard image data into a plurality of grids through a grid of preset density; the third coincidence acquisition submodule is used to respectively obtain the color of each grid, and assign a value of 0 to the black grid and a value of 1 to the white grid; the fourth coincidence acquisition submodule is used to define the current grid and the surrounding eight grids as a 3×3 window, and traverse all grids through the 3×3 window with one grid as a step on all grids; the fifth coincidence acquisition submodule is used to delete the grids with all values ​​1 and all values ​​0 in the 3×3 window based on all traversal results, and define the retained grids The sixth coincidence acquisition submodule is used to repeatedly execute the first coincidence acquisition submodule to the fifth coincidence acquisition submodule with the second standard image data as the main body to obtain the second standard boundary; the seventh coincidence acquisition submodule is used to execute the first coincidence acquisition submodule to the fifth coincidence acquisition submodule with the first predicted image data as the main body to obtain the first predicted boundary; the eighth coincidence acquisition submodule is used to repeatedly execute the first coincidence acquisition submodule to the fifth coincidence acquisition submodule with the second predicted image data as the main body to obtain the second predicted boundary; the ninth coincidence acquisition submodule is used to obtain the area ratio of the first predicted boundary to the first standard boundary and define it as the first coincidence, and obtain the area ratio of the second predicted boundary to the second standard boundary and define it as the second coincidence.

[0198] Furthermore, the overlap judgment module 7 specifically includes a first overlap judgment submodule, a second overlap judgment submodule, a third overlap judgment submodule, a fourth overlap judgment submodule, a fifth overlap judgment submodule, a sixth overlap judgment submodule, a seventh overlap judgment submodule, an eighth overlap judgment submodule, and a ninth overlap judgment submodule, which are electrically connected in sequence; the first overlap judgment submodule is electrically connected to the eighth overlap acquisition submodule, and the ninth overlap judgment submodule is electrically connected to the risk deformation judgment module 8.

[0199] Among them, the first overlap judgment submodule is used to predict a number of first predicted image data based on a number of different preset step numbers to obtain a number of first overlaps; the second overlap judgment submodule is used to predict a number of second predicted image data based on a number of different preset step numbers to obtain a number of second overlaps; the third overlap judgment submodule is used to define at least two continuous percentage intervals based on a preset percentage threshold; the fourth overlap judgment submodule is used to assign risk factors from high to low in order from small to large based on all percentage intervals; the fifth overlap judgment submodule is used to classify all first overlaps into all percentage intervals, and obtain each first overlap respectively. The sixth overlap judgment submodule is used for weighted calculation of the risk coefficients of all first overlaps to obtain a first comprehensive risk; the seventh overlap judgment submodule is used for classifying all second overlaps into all percentage intervals, and obtaining the risk coefficient of each second overlap respectively; the eighth overlap judgment submodule is used for weighted calculation of the risk coefficients of all second overlaps to obtain a second comprehensive risk; the ninth overlap judgment submodule is used for judging whether the first comprehensive risk and the second comprehensive risk are both less than a preset percentage threshold based on the same preset time period, and if the first comprehensive risk and the second comprehensive risk are both less than the preset percentage threshold, the risk deformation judgment module 8 is executed.

[0200] Furthermore, the fifth overlap judgment submodule specifically includes a first overlap judgment unit, a second overlap judgment unit, a third overlap judgment unit, a fourth overlap judgment unit, and a fifth overlap judgment unit which are electrically connected in sequence; the first overlap judgment unit is electrically connected to the fourth overlap judgment submodule, and the fifth overlap judgment unit is electrically connected to the sixth overlap judgment submodule.

[0201] The first coincidence judgment unit is used to integrate all first coincidences into a data set to be classified , For the The first degree of coincidence, is the number of all first degrees of coincidence.

[0202] The second coincidence judgment unit is used to define a category set according to the number of all percentage intervals. ,in, For category collection The Percentage interval, is the number of all percentage intervals.

[0203] The third coincidence judgment unit is used to calculate the first The conditional probability of the first coincidence degree in each percentage interval is:

[0204] (1).

[0205] in, For all first coincidences in The conditional probability of a percentage interval; For the The marginal probability of a percentage interval; For the In the percentage interval The conditional probability of the first degree of coincidence.

[0206] The fourth coincidence degree judgment unit is used to classify each first coincidence degree into a percentage interval with the highest conditional probability.

[0207] The fifth coincidence degree judgment unit is used to respectively obtain the risk coefficient corresponding to each percentage interval with the highest probability of the first coincidence degree condition.

[0208] Furthermore, the seventh overlap judgment submodule is specifically used to repeatedly execute the first overlap judgment unit to the fifth overlap judgment unit based on all the second overlaps, so as to respectively obtain the risk coefficient corresponding to the percentage interval with the highest conditional probability of each second overlap.

[0209] It should be noted that this embodiment is a functional module item embodiment based on the above-mentioned method embodiment. For additional contents such as the preferences, examples, and principle explanations of this embodiment, please refer to the above-mentioned embodiments, and this embodiment will not be repeated herein.

[0210] In this embodiment, the first standard image data and the second standard image data of the photovoltaic panel are obtained under windless conditions, the shooting center line of the first standard image data is collinear with the perpendicular midline of the upper chord and the lower chord and is perpendicular to the panel surfaces of all photovoltaic panels, and the shooting center line of the second standard image data is perpendicular to the shooting center lines of all the first image data and is parallel to the panel surfaces of all photovoltaic panels; a plurality of first image data and a plurality of second image data of the photovoltaic panel are obtained based on a preset time period, and the shooting constraints of all the first image data and all the second image data are respectively the same as the shooting constraints of the first standard image data and the second standard image data; all the first image data are divided into a first training set and a first sample set according to a first preset ratio, and all the second image data are divided into a second training set according to a second preset ratio. training set and a second sample set; output the first training set to a preset neural network model and train it to obtain a first prediction model, and output the second training set to a preset neural network model and train it to obtain a second prediction model; output the first sample set to the first prediction model to obtain first prediction image data, and output the second sample set to the second prediction model to obtain second prediction image data; obtain a first overlap between the first prediction image data and the first standard image data, and a second overlap between the second prediction image data and the second standard image data through a boundary acquisition algorithm; determine whether the first overlap and the second overlap are both less than a preset percentage threshold based on the same preset time period, and if the first overlap and the second overlap are both less than the preset percentage threshold, it is determined that the flexible photovoltaic bracket has risk deformation. In this embodiment, the photovoltaic panel shape in a windless state is calibrated as a standard sample, providing an evaluation basis for the real-time state in the subsequent use process, avoiding the photovoltaic panel in the use process as a sample reference (already having vibration characteristics or risk deformation characteristics) to cause inaccurate evaluation, and then comparing the photovoltaic prediction shape obtained by machine learning with the standard sample to learn about the upcoming risk deformation, so that the risk evaluation of the flexible photovoltaic bracket is predictable.

[0211] like Figure 3 As shown, this embodiment provides an embodiment of an electronic device. In this embodiment, the electronic device 9 includes a processor 91 and a memory 92 coupled to the processor 91.

[0212] The memory 92 stores program instructions for implementing the risk assessment method of the flexible photovoltaic support of any of the above embodiments.

[0213] The processor 91 is used to execute program instructions stored in the memory 92 to perform risk assessment on the flexible photovoltaic support.

[0214] The processor 91 may also be referred to as a CPU (Central Processing Unit). The processor 91 may be an integrated circuit chip having data processing capabilities. The processor 91 may also be a general-purpose processor, a digital data processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0215] Further, Figure 4 The schematic diagram of the structure of the storage medium of an embodiment of the present application is that the storage medium 10 of the embodiment of the present application stores program instructions 101 that can implement all the above methods, wherein the program instructions 101 can be stored in the above storage medium in the form of a software product, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, a server, a mobile phone, and a tablet.

[0216] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0217] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of software functional units. The above is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the specification and drawings of this application, or directly or indirectly used in other related technical fields, is also included in the patent protection scope of the present application.

[0218] The specific implementation methods of the present application are described in detail above, but they are only examples, and the present application is not limited to the specific implementation methods described above. For those skilled in the art, any equivalent modification or substitution of the present application is also within the scope of the present application, and therefore, the equalization, modification, and improvement made without departing from the spirit and principle of the present application should be included in the scope of the present application.

Claims

1. A risk assessment method for a flexible photovoltaic support, wherein the flexible photovoltaic support comprises at least two steel beams fixed to the ground, upper and lower chords erected on adjacent steel beams and in different planes from each other, and a plurality of photovoltaic panels installed on the upper and lower chords, wherein: The risk assessment method includes: Step S1, acquiring first standard image data and second standard image data of the photovoltaic panel under windless conditions, wherein a shooting center line of the first standard image data is collinear with a perpendicular midline of the upper chord and the lower chord and is perpendicular to the panel surfaces of all photovoltaic panels, and a shooting center line of the second standard image data is perpendicular to the shooting center lines of all first image data and is parallel to the panel surfaces of all photovoltaic panels; Step S2, acquiring a plurality of first image data and a plurality of second image data of the photovoltaic panel based on a preset time period, wherein shooting constraints of all the first image data and all the second image data are respectively the same as shooting constraints of the first standard image data and the second standard image data; Step S3, dividing all first image data into a first training set and a first sample set according to a first preset ratio, and dividing all second image data into a second training set and a second sample set according to a second preset ratio; Step S4, outputting the first training set to a preset neural network model and training it to obtain a first prediction model, and outputting the second training set to the preset neural network model and training it to obtain a second prediction model; Step S5, outputting the first sample set to the first prediction model to obtain first predicted image data, and outputting the second sample set to the second prediction model to obtain second predicted image data; Step S6, obtaining a first degree of coincidence between the first predicted image data and the first standard image data, and a second degree of coincidence between the second predicted image data and the second standard image data by using a boundary acquisition algorithm; Step S7, determining whether the first overlap degree and the second overlap degree are both less than a preset percentage threshold based on the same preset time period, and if the first overlap degree and the second overlap degree are both less than the preset percentage threshold, executing step S8; Step S8, determining whether the flexible photovoltaic bracket has risky deformation; Step S6, obtaining a first degree of coincidence between the first predicted image data and the first standard image data, and a second degree of coincidence between the second predicted image data and the second standard image data through a boundary acquisition algorithm, comprising: Step S61, binarizing the first standard image data to obtain first black-and-white standard image data; Step S62, dividing the first black-and-white standard image data into a plurality of grids using a grid of a preset density; Step S63, respectively obtaining the color of each grid, and assigning a value of 0 to a grid with a black color and a value of 1 to a grid with a white color; Step S64, defining the current grid and the surrounding eight grids as a 3×3 window, and traversing all grids through the 3×3 window with one grid as a step; Step S65, based on all the traversal results, the grids assigned with all values ​​of 1 and all values ​​of 0 in the 3×3 window are deleted, and the remaining grids are defined as the first standard boundary; Step S66, repeating steps S61 to S65 with the second standard image data as the main body to obtain a second standard boundary; Step S67, repeating steps S61 to S65 based on the first predicted image data to obtain a first predicted boundary; Step S68, repeating steps S61 to S65 based on the second predicted image data to obtain a second predicted boundary; Step S69: obtain the area ratio of the first predicted boundary to the first standard boundary and define it as the first overlap, and obtain the area ratio of the second predicted boundary to the second standard boundary and define it as the second overlap.

2. The risk assessment method according to claim 1, characterized in that: Step S3, dividing all first image data into a first training set and a first sample set according to a first preset ratio, and dividing all second image data into a second training set and a second sample set according to a second preset ratio, including: Step S31, obtaining a first minimum bounding box based on the photovoltaic panel in the first image data and a second minimum bounding box based on the photovoltaic panel in the second image data by using a target detection algorithm; Step S32, dividing all first minimum bounding boxes into the first training set and the first sample set according to the first preset ratio; Step S33: Divide all second minimum bounding boxes into the second training set and the second sample set according to the second preset ratio.

3. The risk assessment method according to claim 1, characterized in that: Step S4, outputting the first training set to a preset neural network model and training it to obtain a first prediction model, and outputting the second training set to the preset neural network model and training it to obtain a second prediction model, including: Step S41, defining a topological relationship of the preset neural network model, wherein the topological relationship includes an input layer, a hidden layer, and an output layer that are sequentially signal-connected; Step S42, outputting the first training set to the input layer, training the first training set a first preset number of times by using a least squares estimation method, and obtaining a root mean square error between the first sample set and the training result based on each training; Step S43, obtaining a first minimum value among all root mean square errors, and obtaining a training result corresponding to the first minimum value as the first prediction model; Step S44, outputting the second training set to the input layer, training the second training set a second preset number of times using the least squares estimation method, and obtaining a root mean square error between the second sample set and the training result based on each training; Step S45, obtaining a second minimum value in the root mean square error, and obtaining a training result corresponding to the second minimum value as the second prediction model.

4. The risk assessment method according to claim 1, characterized in that: Step S7, based on the same preset time period, determining whether the first overlap degree and the second overlap degree are both less than a preset percentage threshold, if the first overlap degree and the second overlap degree are both less than the preset percentage threshold, executing step S8, including: Step S71, predicting a plurality of first predicted image data based on a plurality of different preset step numbers to obtain a plurality of first coincidence degrees; Step S72, predicting a plurality of second predicted image data based on a plurality of different preset step numbers to obtain a plurality of second coincidence degrees; Step S73, defining at least two consecutive percentage intervals based on the preset percentage threshold; Step S74, assigning risk coefficients from high to low based on all percentage intervals from small to large; Step S75, classifying all first coincidence degrees into all percentage intervals, and obtaining the risk coefficient of each first coincidence degree respectively; Step S76, weighted calculation of risk coefficients of all first coincidence degrees to obtain a first comprehensive risk; Step S77, classifying all second coincidence degrees into all percentage intervals, and obtaining the risk coefficient of each second coincidence degree respectively; Step S78, weighted calculation of risk coefficients of all second coincidence degrees to obtain a second comprehensive risk; Step S79, based on the same preset time period, determine whether the first comprehensive risk and the second comprehensive risk are both less than the preset percentage threshold; if the first comprehensive risk and the second comprehensive risk are both less than the preset percentage threshold, execute step S8.

5. The risk assessment method according to claim 4, characterized in that: Step S75, classifying all first overlaps into all percentage intervals, and obtaining the risk factor of each first overlap respectively, including: Step S751: Integrate all first overlaps into a data set to be classified , For the The first degree of coincidence, is the number of all first degree of coincidence; Step S752: define a category set based on the number of all percentage intervals ,in, For the category set The Percentage interval, is the number of all percentage intervals; Step S753, calculate the first The conditional probability of the first coincidence degree in each percentage interval is: （1）； in, For all first coincidences in The conditional probability of a percentage interval; For the The marginal probability of a percentage interval; For the In the percentage interval The conditional probability of the first degree of coincidence; Step S754, classifying each first coincidence degree into the percentage interval with the highest conditional probability; Step S755, respectively obtain the risk coefficient corresponding to each percentage interval with the highest first coincidence degree conditional probability.

6. The risk assessment method according to claim 5, characterized in that: Step S77, classifying all second overlaps into all percentage intervals, and obtaining the risk factor of each second overlap respectively, including: Step S771, repeating steps S751 to S755 based on all the second coincidence degrees, so as to respectively obtain the risk coefficient corresponding to the percentage interval with the highest conditional probability of each second coincidence degree.

7. A risk assessment device for a flexible photovoltaic support, the risk assessment device being applied to the risk assessment method as claimed in any one of claims 1 to 6, characterized in that: The risk assessment device comprises: A standard image data acquisition module is used to acquire first standard image data and second standard image data of the photovoltaic panel under windless conditions, wherein a shooting center line of the first standard image data is collinear with a perpendicular midline of the upper chord and the lower chord and is perpendicular to the panel surfaces of all photovoltaic panels, and a shooting center line of the second standard image data is perpendicular to the shooting center lines of all first image data and is parallel to the panel surfaces of all photovoltaic panels; An image data acquisition module, used for acquiring a plurality of first image data and a plurality of second image data of the photovoltaic panel based on a preset time period, wherein shooting constraints of all the first image data and all the second image data are respectively the same as shooting constraints of the first standard image data and the second standard image data; A training set and sample set division module, used to divide all first image data into a first training set and a first sample set according to a first preset ratio, and divide all second image data into a second training set and a second sample set according to a second preset ratio; A prediction model acquisition module, used to output the first training set to a preset neural network model and perform training to obtain a first prediction model, and output the second training set to the preset neural network model and perform training to obtain a second prediction model; A predicted image data acquisition module, configured to output the first sample set to the first prediction model to obtain first predicted image data, and output the second sample set to the second prediction model to obtain second predicted image data; a coincidence acquisition module, configured to acquire a first coincidence between the first predicted image data and the first standard image data, and a second coincidence between the second predicted image data and the second standard image data through a boundary acquisition algorithm; A coincidence determination module, used for determining whether the first coincidence degree and the second coincidence degree are both less than a preset percentage threshold value based on the same preset time period; A risk deformation determination module, configured to determine that the flexible photovoltaic support has a risk deformation if both the first overlap degree and the second overlap degree are less than a preset percentage threshold; The overlap acquisition module specifically includes a first overlap acquisition submodule, a second overlap acquisition submodule, a third overlap acquisition submodule, a fourth overlap acquisition submodule, a fifth overlap acquisition submodule, a sixth overlap acquisition submodule, a seventh overlap acquisition submodule, and an eighth overlap acquisition submodule which are electrically connected in sequence; the first overlap acquisition submodule is electrically connected to the predicted image data acquisition module, and the eighth overlap acquisition submodule is electrically connected to the overlap judgment module; Among them, the first coincidence acquisition submodule is used to binarize the first standard image data to obtain the first black-and-white standard image data; the second coincidence acquisition submodule is used to divide the first black-and-white standard image data into a plurality of grids through a grid of preset density; the third coincidence acquisition submodule is used to respectively obtain the color of each grid, and assign a value of 0 to the black grid and a value of 1 to the white grid; the fourth coincidence acquisition submodule is used to define the current grid and the surrounding eight grids as a 3×3 window, and traverse all grids through the 3×3 window with one grid as a step on all grids; the fifth coincidence acquisition submodule is used to delete the grids with all values ​​1 and all values ​​0 in the 3×3 window based on all traversal results, and define the retained grids The sixth coincidence acquisition submodule is used to repeatedly execute the first coincidence acquisition submodule to the fifth coincidence acquisition submodule with the second standard image data as the main body to obtain the second standard boundary; the seventh coincidence acquisition submodule is used to execute the first coincidence acquisition submodule to the fifth coincidence acquisition submodule with the first predicted image data as the main body to obtain the first predicted boundary; the eighth coincidence acquisition submodule is used to repeatedly execute the first coincidence acquisition submodule to the fifth coincidence acquisition submodule with the second predicted image data as the main body to obtain the second predicted boundary; the ninth coincidence acquisition submodule is used to obtain the area ratio of the first predicted boundary to the first standard boundary and define it as the first coincidence, and obtain the area ratio of the second predicted boundary to the second standard boundary and define it as the second coincidence.

8. An electronic device, characterized in that: It includes a processor and a memory coupled to the processor, wherein the memory stores program instructions that can be executed by the processor; when the processor executes the program instructions stored in the memory, the risk assessment method of the flexible photovoltaic bracket as described in any one of claims 1 to 6 is implemented.

9. A storage medium, characterized in that: The storage medium stores program instructions, and when the program instructions are executed by the processor, the risk assessment method for the flexible photovoltaic bracket according to any one of claims 1 to 6 can be implemented.

Images