Equipment monitoring and maintenance method and system for smart power plant

Abstract

The invention discloses an equipment monitoring and maintenance method and system for an intelligent power plant, and relates to the technical field of equipment cleaning, and the method comprises the steps: reading a current maintenance period of a photovoltaic cell panel, querying a dirt factor timetable, and determining a real-time dirt factor; acquiring a dirt fluctuation factor in the current maintenance period, performing dirt accumulation prediction according to the real-time dirt factor and the dirt fluctuation factor in combination with the attribute characteristics of the battery panel, and generating a dirt prediction distribution diagram; and inputting the dirt prediction distribution diagram into a performance attenuation analysis plug-in to carry out performance attenuation analysis on the photovoltaic cell panel, determining a performance attenuation coefficient, and if the performance attenuation coefficient is greater than an expected attenuation scalar, carrying out surface cleaning operation on the photovoltaic cell panel according to a preset cleaning scheme. The technical problems that in the prior art, real-time environment changes are difficult to respond in time, and the actual pollution situation cannot be dynamically handled are solved, the dirt influence of the battery panel is accurately predicted, and the technical effects of improving the cleaning efficiency and the real-time performance are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment cleaning, and in particular to an equipment monitoring and maintenance method and system for a smart power plant. Background Art

[0002] With the rapid development of the new energy industry and the continuous growth of electricity demand, smart power plants have been widely used in modern power production. However, the performance of key equipment such as photovoltaic panels in smart power plants is easily affected by environmental factors, such as dirt in the air, wind speed changes, rainfall, dust accumulation, and industrial pollutants, which will gradually reduce the power generation efficiency of the panels. Traditional equipment maintenance methods mainly rely on regular cleaning and manual inspections, which are difficult to respond to real-time environmental changes in a timely manner and cannot dynamically respond to actual pollution conditions. Therefore, it is difficult to meet the efficient and accurate operation and maintenance needs of modern power production. Summary of the invention

[0003] The present application provides a method and system for equipment monitoring and maintenance of a smart power plant, which is used to solve the technical problems that the existing technology is difficult to respond to real-time environmental changes in a timely manner and cannot dynamically cope with actual pollution conditions.

[0004] In view of the above problems, the present application provides a method and system for equipment monitoring and maintenance of a smart power plant.

[0005] In a first aspect of the present application, a method for monitoring and maintaining equipment in a smart power plant is provided, the method comprising:

[0006] Read the current maintenance cycle of the photovoltaic panels in the smart power plant, query the fouling factor timetable, and determine the real-time fouling factor according to the current maintenance cycle; obtain the fouling fluctuation factor in the current maintenance cycle, and predict the fouling accumulation according to the real-time fouling factor and the fouling fluctuation factor in combination with the attribute characteristics of the panel to generate a fouling prediction distribution map; input the fouling prediction distribution map into the performance attenuation analysis plug-in to perform performance attenuation analysis of the photovoltaic panels and determine the performance attenuation coefficient. If the performance attenuation coefficient is greater than the expected attenuation scalar, perform the photovoltaic panel surface cleaning operation according to the predetermined cleaning plan.

[0007] A second aspect of the present application provides an equipment monitoring and maintenance system for a smart power plant, the system comprising:

[0008] A real-time fouling factor determination module, which reads the current maintenance cycle of the photovoltaic panels in the smart power plant, queries the fouling factor timetable, and determines the real-time fouling factor according to the current maintenance cycle; a fouling accumulation prediction module, which obtains the fouling fluctuation factor in the current maintenance cycle, predicts fouling accumulation according to the real-time fouling factor and the fouling fluctuation factor, and combines the attribute characteristics of the panel to generate a fouling prediction distribution map; a performance attenuation analysis module, which inputs the fouling prediction distribution map into the performance attenuation analysis plug-in to perform performance attenuation analysis of the photovoltaic panel and determine the performance attenuation coefficient. If the performance attenuation coefficient is greater than the expected attenuation scalar, the photovoltaic panel surface cleaning operation is performed according to the predetermined cleaning plan.

[0009] The third aspect of an embodiment of the present application provides an electronic device, comprising: a memory for storing executable instructions; and a processor for executing the executable instructions stored in the memory to implement a method for monitoring and maintaining equipment in a smart power plant provided in the present application.

[0010] The fourth aspect of the embodiments of the present application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements a method for monitoring and maintaining equipment in a smart power plant provided in the present application.

[0011] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0012] This application reads the current maintenance cycle of the photovoltaic panels in the smart power plant, queries the fouling factor schedule, and determines the real-time fouling factor according to the current maintenance cycle match; obtains the fouling fluctuation factor in the current maintenance cycle, and predicts the fouling accumulation according to the real-time fouling factor and the fouling fluctuation factor, combined with the attribute characteristics of the panel, and generates a fouling prediction distribution map; inputs the fouling prediction distribution map into the performance attenuation analysis plug-in to perform performance attenuation analysis of the photovoltaic panel, and determines the performance attenuation coefficient. If the performance attenuation coefficient is greater than the expected attenuation scalar, the photovoltaic panel surface cleaning operation is performed according to the predetermined cleaning scheme. The present invention solves the technical problem that the prior art is difficult to respond to real-time environmental changes in a timely manner and cannot dynamically respond to actual pollution conditions. By reading the current maintenance cycle, combining the fouling factor schedule, the fouling fluctuation factor and the attribute characteristics of the panel, the fouling accumulation prediction is performed, the fouling distribution map is generated, the performance attenuation analysis plug-in is input, and the performance attenuation coefficient is calculated. If it exceeds the preset threshold, the cleaning operation is triggered, and the impact of fouling on the panel is accurately predicted, so as to achieve the technical effect of improving cleaning efficiency and real-time performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0014] Figure 1 A schematic diagram of a process flow of a method for monitoring and maintaining equipment in a smart power plant provided in an embodiment of the present application;

[0015] Figure 2 A schematic diagram of the structure of an equipment monitoring and maintenance system for a smart power plant provided in an embodiment of the present application;

[0016] Figure 3 This is a schematic diagram of the structure of an exemplary electronic device of the present application.

[0017] Explanation of reference numerals: real-time fouling factor determination module 11 , fouling accumulation prediction module 12 , performance degradation analysis module 13 , bus 300 , receiver 301 , processor 302 , transmitter 303 , memory 304 , bus interface 305 . DETAILED DESCRIPTION

[0018] The present application provides an equipment monitoring and maintenance method and system for a smart power plant, aiming to solve the technical problem that the existing technology is difficult to respond to real-time environmental changes in a timely manner and cannot dynamically cope with actual pollution conditions. By reading the current maintenance cycle, combining the fouling factor schedule, fouling fluctuation factor and solar panel attribute characteristics, fouling accumulation prediction is performed, a fouling distribution map is generated, and a performance attenuation analysis plug-in is input to calculate the performance attenuation coefficient. If the preset threshold is exceeded, the cleaning operation is triggered, thereby achieving accurate prediction of the impact of fouling on the solar panel and achieving the technical effect of improving cleaning efficiency and real-time performance.

[0019] 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, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0020] It should be noted that any variations of the terms "include" and "have" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0021] Embodiment 1, as Figure 1 As shown, the present application provides a method for equipment monitoring and maintenance of a smart power plant, the method comprising:

[0022] Step S100: Read the current maintenance cycle of the photovoltaic panels in the smart power plant, query the fouling factor schedule, and determine the real-time fouling factor according to the current maintenance cycle.

[0023] In the embodiment of the present application, the time node of the last cleaning operation, i.e., the last cleaning node, is first obtained from the equipment management platform. The current maintenance cycle refers to the time period from the last cleaning node to the current time period. By calculating the length of this time period, the current maintenance cycle of the photovoltaic panels in the smart power plant is obtained.

[0024] Next, we query a pre-built pollution factor schedule, which records typical environmental pollution conditions around the power plant at different times, including possible pollution types such as dust, sand, bird droppings, pollen, and industrial pollution emissions. Each pollution factor has its own time period in the schedule.

[0025] Then, according to the time range of the current maintenance cycle, the corresponding dirt type is extracted from the dirt factor timetable to generate the real-time dirt factor. If the current maintenance cycle spans two or more time intervals in the timetable, a union determination is performed, that is, the high-frequency dirt factors in all overlapping time intervals are combined to obtain the real-time dirt factor in the current maintenance cycle. The real-time dirt factor specifically includes a variety of possible dirt types such as dust, sand, bird droppings, pollen, industrial pollution, etc.

[0026] Furthermore, in the method provided in the embodiment of the application, constructing the fouling factor schedule also includes:

[0027] Retrieve the environmental monitoring log of the area where the smart power plant is located, extract the fouling factors according to the environmental monitoring log, and determine the fouling factor type set in multiple time intervals, wherein each fouling factor is identified by a factor duration ratio set and a factor intensity set; configure the gradient weight to perform weighted calculation on the factor duration ratio set and the factor intensity set to determine the duration ratio and the intensity coefficient; filter out the factor type whose duration ratio is greater than a predetermined ratio threshold and whose intensity coefficient is greater than a predetermined intensity scalar and set it as a high-frequency fouling factor, obtain multiple high-frequency fouling factors in multiple time intervals, and map and construct the fouling factor timetable.

[0028] In an embodiment of the present application, the environmental monitoring log of the area where the smart power plant is located is first retrieved, which contains information such as air quality, weather, and pollutant concentration for many years. Based on this data, the dirt factors are extracted and classified to identify the main types of dirt factors in multiple time periods. The time is divided into multiple segments, for example, one week as an interval, such as "January 1 to January 7", and the dirt factor data in each time interval is extracted from the monitoring log within 10 years. Each type of dirt factor (such as dust, pollen, sand, etc.) records its factor duration ratio set and factor intensity set in the interval. Among them, the factor duration ratio set reflects the occurrence time ratio of a specific dirt factor in the time interval. For example, the appearance time of dust in the 10 years from January 1 to 7 is 40%. The factor intensity set represents the occurrence density, or concentration, of the dirt factor in each time period. For example, the intensity of sand is lower in the same time period, but the intensity of dust is relatively higher.

[0029] Next, we apply the gradient weights to the factor duration ratio set and factor intensity set for weighted calculation. The weight configuration follows the principle of "the closer the time, the greater the weight", that is, the data closer to the current time is given a higher weight, so as to more accurately reflect the pollution impact of newer environmental conditions. For example, when calculating the dirt factor from January 1 to 7, the data within the last year is given a higher weight, while the data from 10 years ago is given a lower weight. These gradient weights are pre-set by technical experts.

[0030] Then, the duration ratio and intensity coefficient of each fouling factor are screened to determine the factor type that has the greatest impact on the solar panel in each time period. The screening criteria are that the duration ratio is greater than the predetermined ratio threshold, and the intensity coefficient is greater than the set intensity scalar. Only fouling factors that meet both conditions are considered high-frequency fouling factors.

[0031] Finally, the high-frequency fouling factors screened out in different time intervals are mapped to construct a fouling factor timetable.

[0032] Step S200: Obtain the fouling fluctuation factor in the current maintenance cycle, perform fouling accumulation prediction based on the real-time fouling factor and the fouling fluctuation factor in combination with the property characteristics of the solar panel, and generate a fouling prediction distribution map.

[0033] In an embodiment of the present application, the dirt fluctuation factor in the current maintenance cycle is first obtained through a preset environmental sensor, such as an anemometer, that is, a factor reflecting environmental changes, such as wind speed, wind direction, humidity, etc. These fluctuation factors are then combined with the real-time dirt factor, and the attribute characteristics of the solar panel are used as retrieval conditions to obtain similar samples from historical data, including dirt factor, fluctuation factor, cycle duration, and dirt distribution image. Using these sample data, a dirt accumulation predictor is constructed through an adversarial network, and a reliable model is generated through repeated training. Finally, the real-time dirt factor, dirt fluctuation factor, and current maintenance cycle are input into the predictor, and a dirt prediction distribution map is output.

[0034] Furthermore, in the method provided in the embodiment of the application, according to the real-time fouling factor and the fouling fluctuation factor, the fouling accumulation prediction is performed in combination with the property characteristics of the solar panel to generate a fouling prediction distribution map, and the method also includes:

[0035] A positive sample retrieval is performed using the panel attribute characteristics as retrieval constraints to obtain a sample fouling factor set, a sample fluctuation factor set, a sample cycle duration set, and a sample fouling distribution image set; an adversarial network is used to construct a fouling accumulation predictor, wherein the fouling accumulation predictor includes an accumulation generator and an accumulation discriminator, the accumulation generator inputs are the fouling factor, the fluctuation factor, and the cycle duration, and the output is a fouling prediction distribution map, the accumulation discriminator inputs are the fouling prediction distribution map output by the accumulation generator, and the output is the prediction credibility; the sample fouling factor set, the sample fluctuation factor set, the sample cycle duration set, and the sample fouling distribution image set are used as training data, and adversarial training is performed on the accumulation generator and the accumulation discriminator until the prediction credibility meets the expected conditions, thereby obtaining a converging fouling accumulation predictor; the real-time fouling factor, the fouling fluctuation factor, and the current maintenance cycle are input into the fouling accumulation prediction, and the fouling prediction distribution map is output.

[0036] In the embodiment of the present application, the attribute characteristics of the photovoltaic panel, such as surface material, tilt angle, etc., are first used as the retrieval conditions to retrieve positive samples from the historical database. Specifically, the positive sample conditions for retrieval are first defined according to the attribute characteristics of the panel, and these attribute characteristics include the surface material, tilt angle and geographical location of the panel. At the same time, a similarity threshold is set by a technical expert to control the accuracy of the retrieval results. After defining the positive sample conditions, key features are extracted from each sample data from the historical database, including the fouling factor, fluctuation factor and cycle duration. By extracting these features for each sample, the sample is compared in detail with the conditions of the current panel. Next, the similarity between each historical sample and the current panel is calculated using the Euclidean distance, a similarity measurement algorithm. Specifically, the distance between the attribute feature value of each sample and the feature value of the current panel is calculated by the Euclidean distance. If the distance is smaller, it means that the conditions of the sample and the current panel are more similar. The calculated Euclidean distance is compared with the preset similarity threshold, and those samples whose distance is less than the threshold are retained. Through this process, a set of sample dirt factors, a set of sample fluctuation factors, a set of sample cycle durations, and a set of sample dirt distribution images are obtained.

[0037] After obtaining positive sample data, a generative adversarial network is used to construct a dirt accumulation predictor, which contains an accumulation generator and an accumulation discriminator, and the prediction accuracy is gradually optimized through adversarial training.

[0038] The task of the cumulative generator is to generate a predicted dirt distribution map that meets the input conditions. Its input is the dirt factor, fluctuation factor and cycle duration. The generator gradually adjusts the image features through a series of convolution operations to convert the input conditions into a dirt distribution map, which is used to predict the dirt accumulation pattern that may form on the surface of the solar panel under this environment. In each training iteration, the generator generates a preliminary dirt distribution map from the input conditions. The initially generated distribution map may have a large deviation from the actual dirt distribution. For example, the location and density of the dirt distribution may not be consistent with the actual situation. According to the feedback from the discriminator, the generator continuously fine-tunes the parameters to make the generated map gradually approach the actual situation. The task of the cumulative discriminator is to evaluate the prediction credibility of the predicted dirt distribution map generated by the generator and feed it back to the generator so that the generator can adjust the output. Specifically, the discriminator compares the distribution map output by the generator with the actual historical distribution map, calculates the structural similarity index (SSIM) of the two, and obtains a similarity score. SSIM is a metric for measuring image similarity. It captures the subtle differences between the generated image and the real image by considering multiple features such as brightness, contrast and structure. In each training iteration, the discriminator calculates the SSIM score by comparing the brightness, contrast and structural features of the generated image and the real image pixel by pixel. Specifically, the brightness comparison value, contrast comparison value and structural comparison value between the generated image and the real image are first calculated. Among them, the brightness comparison value is obtained by comparing the average brightness of the local area; the contrast comparison value is calculated by the standard deviation of the pixels in the local area; and the structural comparison value measures the texture and shape similarity of the image by the covariance of the local area of ​​the image. Then, the three comparison values ​​are weighted according to the weights preset by technical experts to obtain the final SSIM score. The value of SSIM ranges from 0 to 1, and the closer it is to 1, the higher the similarity between the generated image and the real situation. For example, if the generated dirt distribution map shows that the dirt is concentrated in the lower half of the panel, and the dirt distribution of the real data also shows a similar dirt distribution area, the SSIM score will be close to 1; if the dirt location or concentration of the generated image is very different from the actual distribution map, the SSIM score will be low. According to the SSIM score feedback of the discriminator, the generator further adjusts the parameters in the next iteration to gradually generate a distribution map that is closer to the actual situation.

[0039] During the training process, the cumulative generator and the cumulative discriminator are mutually optimized through adversarial training, and the generator and the discriminator learn independently. The generator continuously generates more realistic dirt prediction distribution maps based on the SSIM score feedback information of the discriminator, while the discriminator continuously improves its ability to identify real data and generated data. As the training progresses, the SSIM score of the generated map gradually approaches 1, indicating that the generator gradually generates a dirt distribution map that conforms to the actual situation. When the SSIM score of the generated map reaches the set expected condition, that is, the generated distribution map is close enough to the real map, it is considered that the predictor has reached a convergence state, and the training stops. At this time, the output of the generator can be used for practical applications.

[0040] When the model converges, the real-time fouling factor, fouling fluctuation factor and current maintenance cycle are used as inputs to the trained fouling accumulation predictor, which generates a predicted fouling distribution map based on these input conditions.

[0041] Furthermore, in the method provided in the embodiment of the application, a dirt prediction distribution map is generated, and then the method further includes:

[0042] If there is rainfall in the current maintenance cycle, a rainfall feature sequence is obtained, wherein the rainfall features include segmented rainfall intensity, segmented rainfall duration, and segmented rainfall amount; a simulated cleaning network is constructed based on the attribute characteristics of the solar panel, and the predicted fouling distribution map is sequentially simulated and cleaned according to the segmented rainfall intensity, segmented rainfall duration, and segmented rainfall amount to generate a simulated fouling distribution map, and the predicted fouling distribution map is updated and replaced using the simulated fouling distribution map.

[0043] In an embodiment of the present application, firstly, a rainfall feature sequence in the current cycle is extracted from the environmental monitoring data, including segmented rainfall intensity, segmented rainfall duration, and segmented rainfall amount. At this time, by using a time series analysis method, the rainfall data is divided into time periods, and the rainfall intensity (such as light rain, moderate rain, or heavy rain), rainfall duration, and the corresponding total precipitation of each section are recorded. For example, a rainfall event starts with 15 minutes of heavy rain, then drops to moderate rain in 30 minutes, and then gradually turns to light rain. This event is divided into multiple sections to accurately record the dynamic change characteristics of rainfall.

[0044] After obtaining the rainfall characteristics, a simulated cleaning network is constructed based on the attribute characteristics of the solar panels. The simulated cleaning network evaluates the cleaning effect of rainfall based on the surface material and inclination of the solar panels, combined with the rainfall feature sequence. For example, solar panels with larger inclinations have a significant effect on dirt loss under heavy rainfall conditions, while the cleaning effect on horizontal surfaces is relatively weak. The simulated cleaning network uses multi-dimensional features such as rainfall intensity, rainfall amount, and rainfall duration as input, and the attribute characteristics of the solar panels as parameter conditions, thereby generating a simulation effect model of rainfall cleaning.

[0045] Next, the values ​​of the segmented rainfall intensity, rainfall duration, and rainfall amount are input into the simulated cleaning network, and the existing predicted dirt distribution map is simulated for cleaning operations segment by segment, thereby generating a simulated dirt distribution map. During the simulated cleaning process, the rainfall characteristics of different time periods are processed segment by segment according to the rainfall intensity and duration. For example, during heavy rainfall, the network will perform a strong cleaning simulation on the dirt-dense areas to reflect the effect of rainfall scouring; while during light rain, the cleaning effect is weak, with only slight scouring of the dirt surface. The system recursively processes the cleaning effect of each rainfall segment, so that the cumulative impact of the entire rainfall event on the dirt can be superimposed segment by segment, thereby obtaining a simulated dirt distribution map.

[0046] Once the dirt simulation distribution map is generated, the distribution map is used to directly replace the original dirt prediction distribution map, thereby completing the new replacement of the dirt prediction distribution map.

[0047] Furthermore, in the method provided in the embodiment of the application, constructing a simulated cleaning network also includes:

[0048] Taking the attribute characteristics of the solar panel as similarity comparison constraints, positive sample retrieval is performed based on power big data to obtain a sample rainfall feature set, a sample initial dirt distribution atlas set and a sample dirt distribution atlas set after cleaning; twin modeling of photovoltaic panels is performed according to the attribute characteristics of the solar panel to construct a solar panel twin model, a simulation cleaning network is built based on the solar panel twin model, and the sample rainfall feature set, the sample initial dirt distribution atlas set and the sample dirt distribution atlas set are used to train the simulation cleaning network until convergence.

[0049] In an embodiment of the present application, firstly, a similarity comparison is performed based on the attribute characteristics of the battery panel, such as surface material, tilt angle, geographical location, etc., and a positive sample retrieval is performed in the power big data database to find historical data that matches the current battery panel status. This process uses the Euclidean distance algorithm to calculate the similarity between each historical sample and the current battery panel. Specifically, the rainfall conditions, initial dirt distribution and cleaning effect data of each sample are extracted from the historical data, and the matching degree of each sample is calculated by the Euclidean distance. For example, if the inclination angle of the current battery panel is 30 degrees and the surface material is smooth glass, and a historical sample has similar conditions, it is retained as positive sample data. Through positive sample retrieval, a sample rainfall feature set, a sample initial dirt distribution atlas, and a sample dirt distribution atlas after cleaning are obtained.

[0050] After the positive sample data is prepared, the 3D modeling technology is used to perform twin modeling of the photovoltaic panels in the 3D simulation space according to the property characteristics of the panels to build a panel twin model. Twin modeling is a technology that builds virtual images in a 3D simulation environment. Through geometric modeling and physical property definition in 3D space, the inclination angle, surface material and environmental factors of the panels are realistically presented in the virtual environment. This model uses 3D rendering and physical engine simulation to accurately simulate the physical properties of the panels and the environmental interaction effects, and can reflect the accumulation and cleaning changes of dirt during actual rainfall.

[0051] After completing the construction of the twin model in the three-dimensional simulation space, the simulated cleaning network was constructed based on the model. The simulated cleaning network adopts a convolutional neural network architecture, and processes rainfall characteristics and dirt distribution maps through multi-layer convolution and pooling operations to extract the cleaning effect of rainfall on dirt distribution. At the network input, the sample rainfall feature set, including segmented information such as rainfall intensity and duration, is used as the rainfall condition input, and the sample initial dirt distribution map set is introduced as the benchmark input to ensure that the simulated cleaning operation of the network has a real initial state reference. The network transforms the initial dirt distribution map into the dirt distribution map after cleaning through learning, thereby simulating the flushing effect of different rainfall conditions on dirt. During the training process, the simulated cleaning network is trained for multiple rounds of iterations using the sample rainfall feature set, the sample initial dirt distribution map set, and the sample dirt distribution map set after cleaning. Specifically, the initial dirt distribution map and rainfall characteristics of each sample are input into the network to generate a preliminary prediction map of the dirt distribution after cleaning. Then, the similarity between the generated map and the real distribution map after cleaning is calculated through the structural similarity index. According to the SSIM score feedback, the parameters of the convolutional layer and the pooling layer are adjusted to optimize the cleaning simulation effect of the network. Through gradual adjustments, the network's prediction gradually approaches the actual situation and eventually reaches a convergence state.

[0052] When the training is completed and reaches convergence, the construction of the simulated cleaning network is completed.

[0053] Step S300: Input the predicted dirt distribution map into the performance attenuation analysis plug-in to perform performance attenuation analysis on the photovoltaic panel and determine the performance attenuation coefficient. If the performance attenuation coefficient is greater than the expected attenuation scalar, perform the photovoltaic panel surface cleaning operation according to the predetermined cleaning plan.

[0054] In the embodiment of the present application, the predicted dirt distribution map is input into the pre-built performance decay analysis plug-in, and the performance decay analysis of the photovoltaic panel is performed based on the distribution map, thereby calculating a performance decay coefficient. If the performance decay coefficient exceeds the set expected decay scalar, that is, the performance decay threshold allowed for the normal operation of the panel, the predetermined cleaning solution is automatically started to clean the surface of the panel.

[0055] Furthermore, in the method provided in the embodiment of the application, constructing a performance degradation analysis plug-in also includes:

[0056] The solar panel attribute characteristics are used as equipment constraints, and the light information in the current maintenance cycle is used as a conditional constraint to perform retrieval, obtain a sample dirt distribution atlas, and perform performance attenuation analysis based on the sample dirt distribution atlas to determine a sample performance attenuation coefficient set; the sample dirt distribution atlas and the sample performance attenuation coefficient set are used to perform supervised training on a convolutional neural network to generate a performance attenuation analysis plug-in.

[0057] In an embodiment of the present application, firstly, the attribute characteristics of the solar panel, such as surface material, tilt angle, geographical location, etc., are used as equipment constraints, and the lighting information in the current maintenance cycle is used as a conditional constraint. Sample data that meets the conditions is retrieved from the historical database to obtain a sample dirt distribution atlas.

[0058] After obtaining the sample fouling distribution atlas set, the performance of the solar panel is analyzed based on the distribution atlas set to calculate the performance attenuation coefficient corresponding to each fouling distribution map and generate a sample performance attenuation coefficient set.

[0059] Next, the convolutional neural network is supervised and trained using the sample dirt distribution atlas and the sample performance decay coefficient set. During the training process, the sample dirt distribution atlas is used as input data, and the sample performance decay coefficient set corresponding to each dirt distribution map is used as label data, which is input into the convolutional neural network model for supervised training. Through multiple rounds of iterations, the CNN model learns the relationship between dirt distribution and performance decay, and finally outputs a performance decay analysis plug-in that can predict the performance decay coefficient.

[0060] Furthermore, in the method provided in the embodiment of the application, the performance degradation analysis is performed according to the sample dirt distribution atlas, and further includes:

[0061] A first sample dirt distribution map is randomly selected, and a performance attenuation search is performed based on the first sample dirt distribution map to obtain a first sample performance attenuation coefficient set; the mean of the first sample performance attenuation coefficient set is calculated, and deviation calculation is performed on the first sample performance attenuation coefficient set according to the first coefficient mean, and the first sample performance attenuation coefficient whose coefficient deviation is less than the deviation threshold is extracted to construct a first standard performance attenuation coefficient set; the mean calculation is performed on the first standard performance attenuation coefficient set to determine the first sample performance attenuation coefficient, and the first sample performance attenuation coefficient is added to the sample performance attenuation coefficient set.

[0062] In the embodiment of the present application, the first sample fouling distribution map is first randomly selected as the starting point for the performance degradation analysis. The distribution map shows the coverage under specific fouling conditions, and the distribution map is used as input to perform performance degradation retrieval, extract relevant historical records from the database, obtain performance degradation data under the fouling conditions, and obtain the first sample performance degradation coefficient set.

[0063] After obtaining the first sample performance attenuation coefficient set, the coefficient set is averaged to determine the central trend of the set, that is, the first coefficient mean. The mean is calculated by arithmetic mean, that is, all coefficients are added and divided by the total number of items. The mean obtained in this way represents the typical performance attenuation level of the entire panel under the current fouling conditions. Next, the deviation of each performance attenuation coefficient is calculated according to the first coefficient mean to identify and filter outliers in the data. The deviation calculation uses the absolute deviation method to subtract each performance attenuation coefficient from the mean and take its absolute value to obtain the difference value of each data point relative to the mean. These deviation values ​​are then compared with the preset deviation threshold, and the performance attenuation coefficients with deviation values ​​less than the threshold are screened out. It is believed that these data points are within the normal range and can truly reflect the typical impact of fouling distribution on the performance of the panel. The coefficients that meet the deviation conditions are retained to construct a first standard performance attenuation coefficient set.

[0064] Finally, the first standard performance attenuation coefficient set is averaged to obtain the first sample performance attenuation coefficient, and the first sample performance attenuation coefficient is added to the overall sample performance attenuation coefficient set.

[0065] In the embodiments of the present application, in summary, the embodiments of the present application have at least the following technical effects:

[0066] This application reads the current maintenance cycle of the photovoltaic panels in the smart power plant, queries the fouling factor schedule, and determines the real-time fouling factor according to the current maintenance cycle match; obtains the fouling fluctuation factor in the current maintenance cycle, and predicts the fouling accumulation according to the real-time fouling factor and the fouling fluctuation factor, combined with the attribute characteristics of the panel, and generates a fouling prediction distribution map; inputs the fouling prediction distribution map into the performance attenuation analysis plug-in to perform performance attenuation analysis of the photovoltaic panel, and determines the performance attenuation coefficient. If the performance attenuation coefficient is greater than the expected attenuation scalar, the photovoltaic panel surface cleaning operation is performed according to the predetermined cleaning scheme. The present invention solves the technical problem that the prior art is difficult to respond to real-time environmental changes in a timely manner and cannot dynamically respond to actual pollution conditions. By reading the current maintenance cycle, combining the fouling factor schedule, the fouling fluctuation factor and the attribute characteristics of the panel, the fouling accumulation prediction is performed, the fouling distribution map is generated, the performance attenuation analysis plug-in is input, and the performance attenuation coefficient is calculated. If it exceeds the preset threshold, the cleaning operation is triggered, and the impact of fouling on the panel is accurately predicted, so as to achieve the technical effect of improving cleaning efficiency and real-time performance.

[0067] Embodiment 2 is based on the same inventive concept as the equipment monitoring and maintenance method of a smart power plant in the above embodiment. Figure 2 As shown, the present application provides an equipment monitoring and maintenance system for a smart power plant, and the system and method embodiments in the present application are based on the same inventive concept. The system includes:

[0068] A real-time fouling factor determination module 11, the real-time fouling factor determination module 11 reads the current maintenance cycle of the photovoltaic panels in the smart power plant, queries the fouling factor timetable, and determines the real-time fouling factor according to the current maintenance cycle; a fouling accumulation prediction module 12, the fouling accumulation prediction module 12 obtains the fouling fluctuation factor in the current maintenance cycle, and predicts fouling accumulation according to the real-time fouling factor and the fouling fluctuation factor in combination with the attribute characteristics of the panel, and generates a fouling prediction distribution map; a performance attenuation analysis module 13, the performance attenuation analysis module 13 inputs the fouling prediction distribution map into the performance attenuation analysis plug-in to perform performance attenuation analysis of the photovoltaic panel and determine the performance attenuation coefficient. If the performance attenuation coefficient is greater than the expected attenuation scalar, the photovoltaic panel surface cleaning operation is performed according to a predetermined cleaning plan.

[0069] Furthermore, the system is also used to implement the following functions:

[0070] If there is rainfall in the current maintenance cycle, a rainfall feature sequence is obtained, wherein the rainfall features include segmented rainfall intensity, segmented rainfall duration, and segmented rainfall amount; a simulated cleaning network is constructed based on the attribute characteristics of the solar panel, and the predicted fouling distribution map is sequentially simulated and cleaned according to the segmented rainfall intensity, segmented rainfall duration, and segmented rainfall amount to generate a simulated fouling distribution map, and the predicted fouling distribution map is updated and replaced using the simulated fouling distribution map.

[0071] Furthermore, the system is also used to implement the following functions:

[0072] Taking the attribute characteristics of the solar panel as similarity comparison constraints, positive sample retrieval is performed based on power big data to obtain a sample rainfall feature set, a sample initial dirt distribution atlas set and a sample dirt distribution atlas set after cleaning; twin modeling of photovoltaic panels is performed according to the attribute characteristics of the solar panel to construct a solar panel twin model, a simulation cleaning network is built based on the solar panel twin model, and the sample rainfall feature set, the sample initial dirt distribution atlas set and the sample dirt distribution atlas set are used to train the simulation cleaning network until convergence.

[0073] Furthermore, the system is also used to implement the following functions:

[0074] A positive sample retrieval is performed using the panel attribute characteristics as retrieval constraints to obtain a sample fouling factor set, a sample fluctuation factor set, a sample cycle duration set, and a sample fouling distribution image set; an adversarial network is used to construct a fouling accumulation predictor, wherein the fouling accumulation predictor includes an accumulation generator and an accumulation discriminator, the accumulation generator inputs are the fouling factor, the fluctuation factor, and the cycle duration, and the output is a fouling prediction distribution map, the accumulation discriminator inputs are the fouling prediction distribution map output by the accumulation generator, and the output is the prediction credibility; the sample fouling factor set, the sample fluctuation factor set, the sample cycle duration set, and the sample fouling distribution image set are used as training data, and adversarial training is performed on the accumulation generator and the accumulation discriminator until the prediction credibility meets the expected conditions, thereby obtaining a converging fouling accumulation predictor; the real-time fouling factor, the fouling fluctuation factor, and the current maintenance cycle are input into the fouling accumulation prediction, and the fouling prediction distribution map is output.

[0075] Furthermore, the system is also used to implement the following functions:

[0076] The solar panel attribute characteristics are used as equipment constraints, and the light information in the current maintenance cycle is used as a conditional constraint to perform retrieval, obtain a sample dirt distribution atlas, and perform performance attenuation analysis based on the sample dirt distribution atlas to determine a sample performance attenuation coefficient set; the sample dirt distribution atlas and the sample performance attenuation coefficient set are used to perform supervised training on a convolutional neural network to generate a performance attenuation analysis plug-in.

[0077] Furthermore, the system is also used to implement the following functions:

[0078] A first sample dirt distribution map is randomly selected, and a performance attenuation search is performed based on the first sample dirt distribution map to obtain a first sample performance attenuation coefficient set; the mean of the first sample performance attenuation coefficient set is calculated, and deviation calculation is performed on the first sample performance attenuation coefficient set according to the first coefficient mean, and the first sample performance attenuation coefficient whose coefficient deviation is less than the deviation threshold is extracted to construct a first standard performance attenuation coefficient set; the mean calculation is performed on the first standard performance attenuation coefficient set to determine the first sample performance attenuation coefficient, and the first sample performance attenuation coefficient is added to the sample performance attenuation coefficient set.

[0079] Furthermore, the system is also used to implement the following functions:

[0080] Retrieve the environmental monitoring log of the area where the smart power plant is located, extract the fouling factors according to the environmental monitoring log, and determine the fouling factor type set in multiple time intervals, wherein each fouling factor is identified by a factor duration ratio set and a factor intensity set; configure the gradient weight to perform weighted calculation on the factor duration ratio set and the factor intensity set to determine the duration ratio and the intensity coefficient; filter out the factor type whose duration ratio is greater than a predetermined ratio threshold and whose intensity coefficient is greater than a predetermined intensity scalar and set it as a high-frequency fouling factor, obtain multiple high-frequency fouling factors in multiple time intervals, and map and construct the fouling factor timetable.

[0081] Embodiment three, based on the inventive concept of a method for equipment monitoring and maintenance of a smart power plant in the aforementioned embodiment, the present application also provides an electronic device, comprising: at least one processor; a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the steps of any one of the methods described in the aforementioned embodiment one.

[0082] Figure 3 This is a schematic diagram of the structure of an exemplary electronic device of this application. Figure 3 In the embodiment, the bus architecture is represented by bus 300, which may include any number of interconnected buses and bridges, and bus 300 connects various circuits including one or more processors represented by processor 302 and memory represented by memory 304. Bus 300 may also connect various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. Bus interface 305 provides an interface between bus 300 and receiver 301 and transmitter 303. Receiver 301 and transmitter 303 may be the same element, namely a transceiver, providing a unit for communicating with various other devices over a transmission medium. Processor 302 is responsible for managing bus 300 and general processing, while memory 304 may be used to store data used by processor 302 when performing operations.

[0083] Embodiment 4, based on the same inventive concept as the equipment monitoring and maintenance method of a smart power plant in the aforementioned embodiment, the present application also provides a computer-readable storage medium, on which a computer program is stored, and the computer program implements the steps of any one of the methods described in the aforementioned embodiment 1 when executed.

[0084] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0085] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0086] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A method for monitoring and maintaining equipment in a smart power plant, characterized in that: include: Read the current maintenance cycle of the photovoltaic panels in the smart power plant, query the fouling factor schedule, and determine the real-time fouling factor according to the current maintenance cycle; Obtaining the fouling fluctuation factor in the current maintenance cycle, performing fouling accumulation prediction based on the real-time fouling factor and the fouling fluctuation factor in combination with the property characteristics of the solar panel, and generating a fouling prediction distribution map; The predicted dirt distribution map is input into the performance attenuation analysis plug-in to perform performance attenuation analysis on the photovoltaic panel to determine the performance attenuation coefficient. If the performance attenuation coefficient is greater than the expected attenuation scalar, the photovoltaic panel surface cleaning operation is performed according to a predetermined cleaning plan.

2. The equipment monitoring and maintenance method of a smart power plant according to claim 1 is characterized in that: Generates a predicted fouling distribution map, followed by: If there is rainfall in the current maintenance period, obtain a rainfall feature sequence, wherein the rainfall features include segmented rainfall intensity, segmented rainfall duration, and segmented rainfall amount; A simulated cleaning network is constructed based on the attribute characteristics of the battery panel, and the predicted dirt distribution map is simulated and cleaned in sequence according to the segmented rainfall intensity, segmented rainfall duration and segmented rainfall amount to generate a simulated dirt distribution map, and the predicted dirt distribution map is updated and replaced using the simulated dirt distribution map.

3. The equipment monitoring and maintenance method of a smart power plant according to claim 2 is characterized in that: Build a simulated cleaning network, including: Taking the attribute characteristics of the solar panel as similarity comparison constraints, positive sample retrieval is performed based on the power big data to obtain a sample rainfall feature set, a sample initial dirt distribution atlas set, and a sample dirt distribution atlas set after cleaning; The photovoltaic panel is twinned and modeled according to the panel property characteristics to construct a panel twin model. A simulation cleaning network is built based on the panel twin model. The simulation cleaning network is trained to convergence using the sample rainfall feature set, the sample initial dirt distribution atlas set and the sample dirt distribution atlas set.

4. The equipment monitoring and maintenance method of a smart power plant according to claim 1 is characterized in that: According to the real-time fouling factor and the fouling fluctuation factor, fouling accumulation is predicted in combination with the property characteristics of the solar panel to generate a fouling prediction distribution map, including: Perform positive sample retrieval using the solar panel attribute features as retrieval constraints to obtain a sample fouling factor set, a sample fluctuation factor set, a sample cycle duration set, and a sample fouling distribution image set; A dirt accumulation predictor is constructed by using an adversarial network, wherein the dirt accumulation predictor includes an accumulation generator and an accumulation discriminator, the input of the accumulation generator is a dirt factor, a fluctuation factor and a cycle duration, and the output is a dirt prediction distribution map, the input of the accumulation discriminator is the dirt prediction distribution map output by the accumulation generator, and the output is the prediction credibility; Using the sample dirt factor set, the sample fluctuation factor set, the sample cycle duration set and the sample dirt distribution image set as training data, the accumulation generator and the accumulation discriminator are subjected to adversarial training until the prediction credibility meets the expected conditions, thereby obtaining a converging dirt accumulation predictor; The real-time fouling factor, fouling fluctuation factor and current maintenance cycle are input into the fouling accumulation prediction, and the fouling prediction distribution map is output.

5. The equipment monitoring and maintenance method of a smart power plant according to claim 1 is characterized in that: Build performance degradation analysis plug-in, including: Taking the solar panel attribute characteristics as equipment constraints and taking the illumination information in the current maintenance cycle as conditional constraints to perform retrieval, obtain a sample dirt distribution atlas, and perform performance attenuation analysis based on the sample dirt distribution atlas to determine a sample performance attenuation coefficient set; The sample dirt distribution atlas and the sample performance decay coefficient set are used to perform supervised training on the convolutional neural network to generate a performance decay analysis plug-in.

6. The equipment monitoring and maintenance method of a smart power plant according to claim 5 is characterized in that: A performance degradation analysis is performed based on the sample fouling distribution atlas, including: Randomly selecting a first sample dirt distribution map, performing a performance attenuation search based on the first sample dirt distribution map, and obtaining a first sample performance attenuation coefficient set; Calculate the mean of the first sample performance decay coefficient set, and perform deviation calculation on the first sample performance decay coefficient set according to the first coefficient mean, extract the first sample performance decay coefficient whose coefficient deviation is less than the deviation threshold to construct a first standard performance decay coefficient set; The first standard performance decay coefficient set is averaged to determine a first sample performance decay coefficient, which is added to the sample performance decay coefficient set.

7. The equipment monitoring and maintenance method of a smart power plant according to claim 1 is characterized in that: Construct a fouling factor schedule, including: Retrieve the environmental monitoring log of the area where the smart power plant is located, extract the fouling factors according to the environmental monitoring log, and determine the fouling factor type set in multiple time intervals, wherein each fouling factor is marked with a factor duration ratio set and a factor intensity set; Configure the gradual weight to perform weighted calculation on the factor duration ratio set and the factor intensity set to determine the duration ratio and intensity coefficient; The factor type whose screening duration ratio is greater than a predetermined ratio threshold and whose intensity coefficient is greater than a predetermined intensity scalar is set as a high-frequency pollution factor, and multiple high-frequency pollution factors in multiple time intervals are obtained, and the pollution factor timetable is mapped and constructed.

8. A smart power plant equipment monitoring and maintenance system, characterized in that: The system comprises: A real-time fouling factor determination module, which reads the current maintenance cycle of the photovoltaic panels in the smart power plant, queries the fouling factor schedule, and determines the real-time fouling factor according to the current maintenance cycle; A fouling accumulation prediction module, wherein the fouling accumulation prediction module obtains a fouling fluctuation factor in the current maintenance cycle, performs fouling accumulation prediction based on the real-time fouling factor and the fouling fluctuation factor in combination with the property characteristics of the solar panel, and generates a fouling prediction distribution map; A performance attenuation analysis module inputs the predicted dirt distribution diagram into a performance attenuation analysis plug-in to perform a performance attenuation analysis on the photovoltaic panel and determine a performance attenuation coefficient. If the performance attenuation coefficient is greater than an expected attenuation scalar, the photovoltaic panel surface cleaning operation is performed according to a predetermined cleaning plan.

9. An electronic device, characterized in that: The electronic device comprises: A memory for storing executable instructions; The processor is used to implement a method for monitoring and maintaining equipment in a smart power plant according to any one of claims 1 to 7 when executing executable instructions stored in the memory.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, an equipment monitoring and maintenance method for a smart power plant as described in any one of claims 1 to 7 is implemented.

Images