Photovoltaic power station mechanical reliability prediction method and related device
By applying deep learning-based video motion amplification model in photovoltaic power stations, real-time video data of mechanical structures is extracted and analyzed, the problems of large errors and low efficiency of prediction results in the prior art are solved, and accurate prediction and early warning of the reliability of mechanical structure of photovoltaic power stations are achieved, and operation and maintenance efficiency is improved.
Patent Information
- Application Number
- CN202411703856.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-26
- Publication Date
- 2025-03-14
AI Technical Summary
The existing mechanical reliability prediction methods for photovoltaic power plants have problems such as large errors in prediction results and low efficiency, and cannot effectively process the complex data generated during the operation of photovoltaic power plants.
Using a video motion amplification model based on deep learning, we use real-time video data of the mechanical structure of the photovoltaic power station to extract vibration displacement characteristics, and compare them with preset thresholds to achieve mechanical reliability prediction.
It realizes accurate prediction and early warning of the reliability of mechanical structure of photovoltaic power stations, improves operation and maintenance efficiency, reduces operation and maintenance costs, and ensures the stable operation of the power station.
Smart Images
Figure CN119624907A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of reliability prediction of photovoltaic power station equipment, and in particular relates to a photovoltaic power station mechanical reliability prediction method and related devices. Background Art
[0002] With the acceleration of global energy transformation and the growth of demand for renewable energy, the stability and reliability of photovoltaic power stations, as an important part of clean energy, are crucial to ensuring the continuity and stability of power supply; the mechanical part of a photovoltaic power station, including photovoltaic modules, brackets, inverters, etc., is the core component of the power station operation, and its reliability directly affects the overall performance and power generation efficiency of the power station; therefore, accurate prediction of the mechanical reliability of photovoltaic power stations is of great significance to improving the operational stability of the power station, reducing operation and maintenance costs, and extending the service life of the power station.
[0003] At present, the prediction methods for the mechanical reliability of photovoltaic power stations mainly include: methods based on fault tree analysis (FTA), methods based on Markov models, and methods based on fuzzy logic; however, due to the huge and complex amount of data generated during the operation of photovoltaic power stations, the existing prediction methods cannot effectively and accurately process the data; secondly, the existing prediction models have the problems of low accuracy and generalization ability, which leads to large errors in the prediction results and low prediction efficiency, posing a serious threat to the safe operation of photovoltaic power stations. Summary of the invention
[0004] In view of the technical problems existing in the prior art, the present invention provides a photovoltaic power station mechanical reliability prediction method and related devices to solve the technical problems that the existing photovoltaic power station mechanical reliability prediction method has large prediction error and low efficiency.
[0005] In order to achieve the above object, the technical solution adopted by the present invention is: The present invention provides a photovoltaic power station mechanical reliability prediction method, comprising: Collecting real-time video data of a target to be predicted; wherein the target to be predicted is a predetermined mechanical structure in a photovoltaic power station; Input the real-time video data of the target to be predicted into a pre-trained video motion magnification model, and output a motion magnification result of the target to be measured; wherein the pre-trained motion magnification model is a video motion magnification model based on deep learning; Extracting vibration displacement characteristics of the target to be predicted based on the motion amplification result of the target to be predicted; The vibration displacement characteristic of the target to be predicted is compared with the vibration displacement threshold of the target to be predicted to obtain a reliability prediction result of the target to be predicted.
[0006] Furthermore, the deep learning-based video motion magnification model includes a first encoder, a second encoder, a modulator and a decoder; The first encoder is used to extract shape features and texture features of the first video frame to obtain shape features of the first video frame; The second encoder is used to extract shape features and texture features of the second video frame to obtain the second video frame shape features and the second video frame texture features; wherein the first video frame and the second video frame are the previous and next video frames in the real-time video data of the target to be predicted; The modulator is used to perform bandpass filtering, motion amplification and weighted superposition operations on the shape features of the first video frame and the shape features of the second video frame, and output a shape amplification feature; The decoder is used to fuse and reconstruct the shape magnification feature and the texture feature of the second video frame to obtain a motion magnification frame, that is, to obtain a motion magnification result of the target to be measured.
[0007] Furthermore, the first encoder and the second encoder have the same structure, both including: a trunk, a texture extraction branch and a shape extraction branch, the trunk including a first convolutional layer, a first activation function layer, a second convolutional layer, a second activation function layer and three first residual blocks connected in sequence, the texture extraction branch including a third convolutional layer, a third activation function layer and three second residual blocks connected in sequence, and the shape extraction branch including a fourth convolutional layer, a fourth activation function layer and three third residual blocks connected in sequence; wherein the third convolutional layer and the fourth convolutional layer are respectively connected to the last first residual block of the trunk.
[0008] Furthermore, the encoder includes a feature connection layer, three fourth residual blocks, a downsampling layer, a fifth convolution layer, a fifth activation function layer, a sixth convolution layer and a sixth activation function layer connected in sequence; wherein the input of the feature connection layer includes the shape magnification feature and the second video frame texture feature, and the three fourth residual blocks are connected in series in sequence.
[0009] Furthermore, the input of the modulator includes the first video frame shape feature, the second video frame shape feature and a preset magnification coefficient; the modulator performs bandpass filtering, motion amplification and weighted superposition operations through a preset nonlinear objective function; Wherein, the preset nonlinear objective function is:
[0010] in, amplify the features for the shape of the modulator output; is the shape feature of the first video frame; is the shape feature of the second video frame; is the magnification factor; A 3×3 convolutional layer followed by a 3×3 residual block; It is a 3×3 convolutional layer.
[0011] Furthermore, when training the deep learning-based video motion magnification model, L1 distance is used as a metric to determine the loss function; Wherein, the loss function is:
[0012] in, is the loss function; is the true value after motion amplification; Output values for the model; is the texture feature of the perturbation frame; is the shape feature of the perturbation frame; is the regularization weight; is the texture feature of the first video frame; is the texture feature of the second video frame; is the shape feature of the second video frame.
[0013] The present invention also provides a photovoltaic power station mechanical reliability prediction system, comprising: A data acquisition module, used to acquire real-time video data of a target to be predicted; wherein the target to be predicted is a predetermined mechanical structure in a photovoltaic power station; A motion magnification module, used for inputting the real-time video data of the target to be predicted into a pre-trained video motion magnification model, and outputting a motion magnification result of the target to be measured; wherein the pre-trained motion magnification model is a video motion magnification model based on deep learning; A displacement feature extraction module, used to extract the vibration displacement feature of the target to be predicted based on the motion amplification result of the target to be predicted; The comparison prediction module is used to compare the vibration displacement characteristics of the target to be predicted with the vibration displacement threshold of the target to be predicted to obtain a reliability prediction result of the target to be predicted.
[0014] The present invention also provides a photovoltaic power station mechanical reliability prediction device, comprising: a processor suitable for executing a computer program; A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by the processor, the photovoltaic power station mechanical reliability prediction method is executed.
[0015] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the photovoltaic power station mechanical reliability prediction method is implemented.
[0016] The present invention also provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the method for predicting the mechanical reliability of a photovoltaic power station is implemented.
[0017] Compared with the prior art, the present invention has the following beneficial effects: The photovoltaic power station reliability prediction method provided by the present invention collects real-time video data of a predetermined mechanical structure in the photovoltaic power station, analyzes the real-time video data of the predetermined mechanical structure based on a pre-trained video motion amplification model, obtains a motion amplification result of a target to be measured, and compares a vibration displacement feature extracted based on the video motion amplification result with a preset vibration displacement threshold, so as to accurately judge the reliability state of the mechanical structure, realize accurate prediction and early warning of the reliability of the photovoltaic power station mechanical structure, and thus ensure the operation and maintenance efficiency and stable operation of the photovoltaic power station; the present invention realizes real-time monitoring of key components in the photovoltaic power station, which helps to timely discover potential mechanical failures or anomalies, so as to take measures in advance to prevent or repair them, and avoid power station shutdown or efficiency reduction caused by failures.
[0018] The photovoltaic power station mechanical reliability prediction system, photovoltaic power station mechanical reliability prediction device, computer-readable storage medium and computer program product provided by the present invention have all the advantages of the photovoltaic power station mechanical reliability prediction method mentioned above. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0020] Figure 1 A flow chart of the photovoltaic power station mechanical reliability prediction method provided in Example 1; Figure 2 This is a structural diagram of the video motion amplification model based on deep learning in Example 1; Figure 3 A structural block diagram of a photovoltaic power station mechanical reliability prediction system provided in Example 2; Figure 4 This is a structural block diagram of the photovoltaic power station mechanical reliability prediction device provided in Example 3. DETAILED DESCRIPTION
[0021] In order to make the technical problems, technical solutions and beneficial effects solved by this application clearer, the technical solutions in the embodiments of this application will be described clearly and completely in combination with the drawings in the embodiments of this application; obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0022] Example 1 As attached Figure 1 As shown, this embodiment 1 provides a photovoltaic power station mechanical reliability prediction method, including the following steps: Step 1: Collect real-time video data of the target to be measured; wherein the target to be predicted is a predetermined mechanical structure in a photovoltaic power station. Specifically, the process of collecting real-time video data of the target to be measured is as follows: using a pre-installed camera to collect real-time video data of a predetermined mechanical structure in a photovoltaic power station to obtain real-time video data of the target to be measured; preferably, the predetermined mechanical structure includes a photovoltaic module, a bracket and an inverter.
[0023] Step 2: input the real-time data of the target to be predicted into a pre-trained video motion magnification model, and output the motion magnification result of the target to be measured; wherein the pre-trained motion magnification model is a video motion magnification model based on deep learning.
[0024] As attached Figure 2 As shown, the video motion magnification model based on deep learning includes a first encoder, a second encoder, a modulator and a decoder; the first encoder is used to extract the shape features and texture features of the first video frame to obtain the shape features of the first video frame; the second encoder is used to extract the shape features and texture features of the second video frame to obtain the shape features of the second video frame and the texture features of the second video frame; wherein the first video frame and the second video frame are the two previous and next video frames in the real-time video data of the target to be predicted; the modulator is used to perform bandpass filtering, motion magnification and weight superposition operations on the shape features of the first video frame and the shape features of the second video frame, and output the shape magnification features; the decoder is used to fuse and reconstruct the shape magnification features with the texture features of the second video frame to obtain the motion magnification frame, that is, to obtain the motion magnification result of the target to be measured.
[0025] It should be noted that the first encoder and the second encoder have the same structure and share weights, and are both used to receive corresponding video frames and output required texture features and shape features; the modulator is used to perform motion amplification based on the relative interpolation of the shape features output by the first encoder and the second encoder; the decoder is used to reconstruct the amplified motion features output by the modulator and the texture features output by the second encoder to obtain a motion amplified video frame, that is, to obtain the motion amplification result of the target to be measured.
[0026] In this embodiment 1, the first encoder and the second encoder are both spatial decomposition filters composed of a multi-layer convolutional neural network, and both include: a trunk, a texture extraction branch, and a shape extraction branch; the trunk includes a first convolution layer, a first activation function layer, a second convolution layer, a second activation function layer, and three first residual blocks connected in sequence, and the three first residual blocks are connected in series in sequence; the texture extraction branch includes a third convolution layer, a third activation function layer, and three second residual blocks connected in sequence, and the three second residual blocks are connected in series in sequence; the shape extraction branch includes a fourth convolution layer, a fourth activation function layer, and three third residual blocks connected in sequence, and the three third residual blocks are connected in series in sequence. The blocks are connected in series in sequence; wherein the third convolution layer and the fourth convolution layer are respectively connected to the last first residual block of the main path; it should be noted that the video frames input to the first encoder or the second encoder can extract texture features through the main path and the texture extraction branch in sequence, and the video frames can extract shape features through the main path and the shape extraction branch in sequence; wherein the first encoder outputs shape features, and the second encoder outputs texture features and shape features; that is, through the first encoder and the second encoder, it is possible to extract the shape features of the moving subject from the video frames obtained by video decomposition, and then obtain motion features and encode to achieve the representation of the motion.
[0027] In this embodiment 1, the modulator amplifies the motion by acquiring motion features and multiplying the relative difference between two frames, and includes an adder, a subtractor, a multiplier, a convolutional layer and a modulator residual block; the input of the modulator includes the motion features and amplification factors output by two encoders, that is, the shape features of the first video frame, the shape features of the second video frame and a preset amplification factor; the output of the modulator is a shape amplification feature.
[0028] Specifically, the modulator performs bandpass filtering, motion amplification and weighted superposition operations through a preset nonlinear objective function; wherein the preset nonlinear objective function is:
[0029] in, amplify the features for the shape of the modulator output; is the shape feature of the first video frame; is the shape feature of the second video frame; is the magnification factor; A 3×3 convolutional layer followed by a 3×3 residual block; A 3×3 convolutional layer That is, in the modulator, the shape features of the first video frame output by the first encoder and the shape features of the second video frame output by the second encoder are input into a subtractor for subtraction, and then the difference is multiplied by the amplification factor, and the product is successively input into the convolution layer and the residual block for processing to obtain an intermediate feature, and the intermediate feature is summed with the shape features of the previous frame output by the first encoder, and the sum result is the modulation feature after the amplification motion.
[0030] In this embodiment 1, the encoder includes a feature connection layer, three fourth residual blocks, a downsampling layer, a fifth convolution layer, a fifth activation function layer, a sixth convolution layer and a sixth activation function layer connected in sequence; wherein the input of the feature connection layer includes the shape magnification feature and the second video frame texture feature, and the three fourth residual blocks are connected in series in sequence. It should be noted that, in the first encoder, the second encoder and the decoder, residual blocks are used to generate high-quality output; when the first encoder and the second encoder start working, the activated image is downsampled by 2 times with the help of convolution and activation functions; when the decoder ends, 2 times upsampling is performed; wherein, downsampling is completed using a convolutional network with a step size of 2, and upsampling is performed using nearest neighbor interpolation, and then a convolutional layer is used to avoid artifacts.
[0031] In this embodiment 1, when training the deep learning-based video motion magnification model, a preset video training set is used for training; wherein the preset video training set includes a plurality of video frames, and the preset video training set is obtained by data synthesis using textures generated after segmentation of real images; secondly, during training, L1 distance is used as a metric to determine a loss function; wherein the loss function is:
[0032] in, is the loss function; is the true value after motion amplification; Output values for the model; is the texture feature of the perturbation frame; is the shape feature of the perturbation frame; is the regularization weight; is the texture feature of the first video frame; is the texture feature of the second video frame; is the shape feature of the second video frame.
[0033] Step 3: extracting the vibration displacement characteristics of the target to be predicted based on the motion amplification result of the target to be predicted. Specifically, the process of extracting the vibration displacement characteristics of the target to be predicted based on the motion amplification result of the target to be predicted comprises the following steps: Step 31: Using computer vision technology, extract feature points from the motion magnification result of the target to be predicted; wherein the computer vision technology includes a corner point detection method or a feature matching method.
[0034] Step 32: perform optical flow tracking on the feature points to obtain optical flow tracking results; wherein the optical flow tracking results are used to obtain displacement information of the feature points between consecutive video frames; the optical flow tracking algorithm can calculate the motion vector of the feature points between adjacent frames, thereby reflecting the vibration displacement of the target.
[0035] Step 33: Calculate the displacement changes of the feature points at different time points according to the optical flow tracking results to obtain a vibration displacement sequence; wherein the vibration displacement sequence contains the vibration displacement characteristics of the target to be predicted.
[0036] Step 34, filtering the vibration displacement sequence to obtain a filtered vibration displacement sequence to eliminate the influence of noise and interference signals; wherein low-pass filtering or Kalman filtering is used during the filtering process.
[0037] Step 35: extracting useful characteristic indicators from the vibration displacement sequence after filtering; wherein the useful characteristic indicators include vibration frequency, amplitude or phase to reflect the vibration characteristics and state of the target to be predicted.
[0038] Step 4: Compare the vibration displacement characteristics of the target to be predicted with the vibration displacement threshold of the target to be predicted to obtain a reliability prediction result of the target to be predicted.
[0039] Specifically, the vibration displacement characteristics of the target to be predicted are compared with the vibration displacement threshold of the target to be predicted; if the vibration displacement characteristics of the target to be predicted exceeds the threshold, the mechanical reliability prediction result of the target to be predicted is abnormal, and an abnormal warning signal is generated to prompt the operation and maintenance personnel to take corresponding measures; otherwise, the mechanical reliability prediction result of the target to be predicted is normal.
[0040] The photovoltaic power station mechanical reliability prediction method described in this embodiment 1 realizes accurate prediction and early warning of the reliability of the photovoltaic power station mechanical structure through real-time collection of real-time video data of the target to be predicted, application of the video motion amplification model, and extraction of vibration displacement characteristics and reliability prediction steps. It is of great significance to improving the operation and maintenance efficiency of photovoltaic power stations, ensuring stable operation, and promoting intelligent development.
[0041] Example 2 As attached Figure 3 As shown, this embodiment 2 provides a photovoltaic power station mechanical reliability prediction system, including a data acquisition module, a motion amplification module, a displacement feature extraction module and a comparison prediction module. The data acquisition module is used to collect real-time video data of the target to be predicted; wherein the target to be predicted is a predetermined mechanical structure in the photovoltaic power station; the motion amplification module is used to input the real-time video data of the target to be predicted into a pre-trained video motion amplification model, and output the motion amplification result of the target to be measured; wherein the pre-trained motion amplification model is a video motion amplification model based on deep learning; the displacement feature extraction module is used to extract the vibration displacement feature of the target to be predicted based on the motion amplification result of the target to be predicted; the comparison prediction module is used to compare the vibration displacement feature of the target to be predicted with the vibration displacement threshold of the target to be predicted, and obtain the reliability prediction result of the target to be predicted.
[0042] Example 3 As attached Figure 4 As shown, this embodiment 4 provides a photovoltaic power station mechanical reliability prediction device, including: a memory for storing a computer program; a processor for implementing the steps of a photovoltaic power station mechanical reliability prediction method when executing the computer program.
[0043] When the processor executes the computer program, the steps of the photovoltaic power station mechanical reliability prediction method are implemented, for example: Collect real-time video data of a target to be predicted; wherein the target to be predicted is a predetermined mechanical structure in a photovoltaic power station; input the real-time video data of the target to be predicted into a pre-trained video motion amplification model, and output a motion amplification result of the target to be measured; wherein the pre-trained motion amplification model is a video motion amplification model based on deep learning; based on the motion amplification result of the target to be predicted, extract the vibration displacement characteristics of the target to be predicted; compare the vibration displacement characteristics of the target to be predicted with the vibration displacement threshold of the target to be predicted, and obtain a reliability prediction result of the target to be predicted.
[0044] Alternatively, when the processor executes the computer program, the functions of each module in the photovoltaic power station mechanical reliability prediction system are realized, for example: A data acquisition module is used to collect real-time video data of a target to be predicted; wherein the target to be predicted is a predetermined mechanical structure in a photovoltaic power station; a motion amplification module is used to input the real-time video data of the target to be predicted into a pre-trained video motion amplification model, and output a motion amplification result of the target to be measured; wherein the pre-trained motion amplification model is a video motion amplification model based on deep learning; a displacement feature extraction module is used to extract the vibration displacement feature of the target to be predicted based on the motion amplification result of the target to be predicted; a comparison prediction module is used to compare the vibration displacement feature of the target to be predicted with the vibration displacement threshold of the target to be predicted, and obtain a reliability prediction result of the target to be predicted.
[0045] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of completing preset functions, and the instruction segments are used to describe the execution process of the computer program in the photovoltaic power plant mechanical reliability prediction device.
[0046] For example, the computer program can be divided into a data acquisition module, a motion amplification module, a displacement feature extraction module and a comparison prediction module, and the specific functions of each module are as follows: a data acquisition module, used to collect real-time video data of the target to be predicted; wherein the target to be predicted is a predetermined mechanical structure in a photovoltaic power station; a motion amplification module, used to input the real-time video data of the target to be predicted into a pre-trained video motion amplification model, and output the motion amplification result of the target to be measured; wherein the pre-trained motion amplification model is a video motion amplification model based on deep learning; a displacement feature extraction module, used to extract the vibration displacement characteristics of the target to be predicted based on the motion amplification result of the target to be predicted; a comparison prediction module, used to compare the vibration displacement characteristics of the target to be predicted with the vibration displacement threshold of the target to be predicted, to obtain a reliability prediction result of the target to be predicted.
[0047] The photovoltaic power station mechanical reliability prediction device can be a computing device such as a desktop computer, a notebook, a PDA, and a cloud server. The photovoltaic power station mechanical reliability prediction device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the above is an example of a photovoltaic power station mechanical reliability prediction device, and does not constitute a limitation on the photovoltaic power station mechanical reliability prediction device, and may include more components than the above, or a combination of certain components, or different components. For example, the photovoltaic power station mechanical reliability prediction device may also include input and output devices, network access devices, buses, etc.
[0048] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc. The processor is the control center of the photovoltaic power station mechanical reliability prediction device, and uses various interfaces and lines to connect various parts of the entire photovoltaic power station mechanical reliability prediction device.
[0049] The memory can be used to store the computer program and / or module, and the processor implements various functions of the photovoltaic power station mechanical reliability prediction device by running or executing the computer program and / or module stored in the memory and calling the data stored in the memory.
[0050] The memory may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (SmartMediaCard, SMC), a secure digital (SecureDigital, SD) card, a flash card (FlashCard), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.
[0051] Example 4 This embodiment 4 further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the photovoltaic power station mechanical reliability prediction method are implemented, for example: The invention relates to obtaining junction temperature data of an IGBT module in a photovoltaic inverter to be tested; determining the number of junction temperature thermal cycles of the IGBT module in the photovoltaic inverter to be tested according to the junction temperature data of the IGBT module in the photovoltaic inverter to be tested; calculating the cumulative damage of the IGBT module in the photovoltaic inverter to be tested according to the number of junction temperature thermal cycles of the IGBT module in the photovoltaic inverter to be tested; and calculating the remaining life prediction result of the photovoltaic inverter to be tested according to the cumulative damage of the IGBT module in the photovoltaic inverter to be tested.
[0052] If the modules / units integrated in the photovoltaic power station mechanical reliability prediction system are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.
[0053] Based on such understanding, the present invention implements all or part of the processes in the above-mentioned photovoltaic power station mechanical reliability prediction method, and can also be completed by instructing related hardware through a computer program, and the computer program can be stored in a computer-readable storage medium, and when the computer program is executed by a processor, the steps of the above-mentioned photovoltaic power station mechanical reliability prediction method can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or preset intermediate form, etc.
[0054] The computer-readable storage medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium, etc.
[0055] It should be noted that the content contained in the computer-readable storage medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable storage media do not include electrical carrier signals and telecommunication signals.
[0056] Example 5 This embodiment 5 provides a computer product, which includes a computer program, and the computer program is stored in a computer-readable storage medium; the processor of the photovoltaic power station mechanical reliability prediction device reads the computer program from the computer-readable storage medium, and the processor executes the computer program, so that the photovoltaic power station mechanical reliability prediction device can execute the photovoltaic power station mechanical reliability prediction method described in embodiment 1, which will not be repeated here.
[0057] It should be noted that a person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing related hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods.
[0058] The photovoltaic power station mechanical reliability prediction method described in the present invention realizes real-time monitoring of key components in the photovoltaic power station by collecting real-time video data of predetermined mechanical structures in the photovoltaic power station, which is helpful to timely discover potential mechanical failures or abnormalities; through the application of the video motion amplification model, the tiny vibrations of the mechanical structure in the video can be amplified, so that the originally imperceptible vibrations become clearly visible, which is very useful for detecting early failures or abnormal vibrations of the mechanical structure, because tiny vibrations are often a precursor to failures, and thus the detection sensitivity of the mechanical structure vibration can be significantly improved; by comparing the extracted vibration displacement characteristics with the preset vibration displacement threshold, the vibration displacement characteristics extracted based on the video motion amplification results can more accurately reflect the actual vibration conditions of the mechanical structure, and then can scientifically judge the reliability status of the mechanical structure, which is helpful to timely discover and deal with mechanical failures or abnormalities in the photovoltaic power station, thereby improving operation and maintenance efficiency and reducing operation and maintenance costs.
[0059] The above embodiment is only one of the implementation methods that can realize the technical solution of the present invention. The scope of protection claimed by the present invention is not limited only to this embodiment, but also includes changes, replacements and other implementation methods that can be easily thought of by any technician familiar with the technical field within the technical scope disclosed by the present invention.
Claims
1. A photovoltaic power station mechanical reliability prediction method, characterized in that: include: Collecting real-time video data of a target to be predicted; wherein the target to be predicted is a predetermined mechanical structure in a photovoltaic power station; Input the real-time video data of the target to be predicted into a pre-trained video motion magnification model, and output a motion magnification result of the target to be measured; wherein the pre-trained motion magnification model is a video motion magnification model based on deep learning; Extracting vibration displacement characteristics of the target to be predicted based on the motion amplification result of the target to be predicted; The vibration displacement characteristic of the target to be predicted is compared with the vibration displacement threshold of the target to be predicted to obtain a reliability prediction result of the target to be predicted.
2. A photovoltaic power station mechanical reliability prediction method according to claim 1, characterized in that: The deep learning-based video motion magnification model includes a first encoder, a second encoder, a modulator and a decoder; The first encoder is used to extract shape features and texture features of the first video frame to obtain shape features of the first video frame; The second encoder is used to extract shape features and texture features of the second video frame to obtain the second video frame shape features and the second video frame texture features; wherein the first video frame and the second video frame are the previous and next video frames in the real-time video data of the target to be predicted; The modulator is used to perform bandpass filtering, motion amplification and weighted superposition operations on the shape features of the first video frame and the shape features of the second video frame, and output a shape amplification feature; The decoder is used to fuse and reconstruct the shape magnification feature and the texture feature of the second video frame to obtain a motion magnification frame, that is, to obtain a motion magnification result of the target to be measured.
3. A photovoltaic power station mechanical reliability prediction method according to claim 2, characterized in that: The first encoder and the second encoder have the same structure, and both include: a trunk, a texture extraction branch and a shape extraction branch, the trunk includes a first convolution layer, a first activation function layer, a second convolution layer, a second activation function layer and three first residual blocks connected in sequence, the texture extraction branch includes a third convolution layer, a third activation function layer and three second residual blocks connected in sequence, and the shape extraction branch includes a fourth convolution layer, a fourth activation function layer and three third residual blocks connected in sequence; wherein the third convolution layer and the fourth convolution layer are respectively connected to the last first residual block of the trunk.
4. A photovoltaic power station mechanical reliability prediction method according to claim 2, characterized in that: The encoder includes a feature connection layer, three fourth residual blocks, a downsampling layer, a fifth convolution layer, a fifth activation function layer, a sixth convolution layer and a sixth activation function layer connected in sequence; wherein the input of the feature connection layer includes the shape magnification feature and the second video frame texture feature, and the three fourth residual blocks are connected in series in sequence.
5. A photovoltaic power station mechanical reliability prediction method according to claim 2, characterized in that: The input of the modulator includes the shape feature of the first video frame, the shape feature of the second video frame and a preset magnification coefficient; the modulator performs bandpass filtering, motion amplification and weighted superposition operations through a preset nonlinear objective function; Wherein, the preset nonlinear objective function is: in, amplify the features for the shape of the modulator output; is the shape feature of the first video frame; is the shape feature of the second video frame; is the magnification factor; A 3×3 convolutional layer followed by a 3×3 residual block; It is a 3×3 convolutional layer.
6. A photovoltaic power station mechanical reliability prediction method according to claim 2, characterized in that: When training the deep learning-based video motion magnification model, L1 distance is used as a metric to determine the loss function; Wherein, the loss function is: in, is the loss function; is the true value after motion amplification; Output values for the model; is the texture feature of the perturbation frame; is the shape feature of the perturbation frame; is the regularization weight; is the texture feature of the first video frame; is the texture feature of the second video frame; is the shape feature of the second video frame.
7. A photovoltaic power station mechanical reliability prediction system, characterized in that: include: A data acquisition module, used to acquire real-time video data of a target to be predicted; wherein the target to be predicted is a predetermined mechanical structure in a photovoltaic power station; A motion magnification module, used for inputting the real-time video data of the target to be predicted into a pre-trained video motion magnification model, and outputting a motion magnification result of the target to be measured; wherein the pre-trained motion magnification model is a video motion magnification model based on deep learning; A displacement feature extraction module, used to extract the vibration displacement feature of the target to be predicted based on the motion amplification result of the target to be predicted; The comparison prediction module is used to compare the vibration displacement characteristics of the target to be predicted with the vibration displacement threshold of the target to be predicted to obtain a reliability prediction result of the target to be predicted.
8. A photovoltaic power station mechanical reliability prediction device, characterized in that: include: a processor suitable for executing a computer program; A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by the processor, the photovoltaic power station mechanical reliability prediction method according to any one of claims 1 to 6 is executed.
9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the photovoltaic power station mechanical reliability prediction method according to any one of claims 1 to 6 is implemented.
10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method for predicting mechanical reliability of a photovoltaic power station according to any one of claims 1 to 6 is implemented.