Wind-resistant early warning method and system for flexible photovoltaic support

Abstract

The invention relates to the technical field of photovoltaic power generation, in particular to a wind-resistant early warning method and system for a flexible photovoltaic support, and the method comprises the steps: determining the stability of each connection part according to the vibration data of each connection part of the photovoltaic support; acquiring a predicted wind speed and a predicted wind direction at the next moment; taking the orientation of the photovoltaic support and the inclination angle of the cell panel at the current moment as attitude information, and calculating the attention of each connection part according to the attitude information and the predicted wind direction; calculating a predicted wind load at the next moment according to the predicted wind speed, the predicted wind direction and the attitude information, and calculating a local risk value of each connection part; performing weighted summation on the risk values of the connection parts by using the normalized attention to obtain an overall risk value; and wind-resistant early warning is realized according to the overall risk value. According to the technical scheme of the invention, the accuracy of wind-resistant early warning of the flexible photovoltaic support can be improved.

Description

Technical Field

[0001] This application relates to the field of photovoltaic power generation technology, and in particular to a wind-resistant early warning method and system for flexible photovoltaic supports. Background Technology

[0002] Flexible photovoltaic (PV) brackets, as a crucial component of solar photovoltaic (PV) systems, are gradually gaining prominence in the PV field due to their adaptability to diverse environmental conditions and installation requirements. During operation, flexible PV brackets often adjust their angle according to the sun's position to maximize solar energy collection efficiency and maintain maximum power generation. However, ensuring the safety and stability of these brackets under adverse weather conditions such as strong winds remains a critical issue that needs to be addressed.

[0003] Currently, patent application CN118609325A discloses a method and system for providing strong wind warning protection for photovoltaic tracking brackets. The method includes: acquiring historical wind speed data with a length that is an integer multiple of the prediction period according to the configured prediction period; dividing the acquired wind speed data into multiple data segments according to the prediction period; calculating the mean data corresponding to each data segment to obtain the mean dataset M; using the data in the mean dataset M as input to the Kalman filter model to calculate the mean wind speed M2 for the prediction period; calculating the predicted mean M3 based on the acquired mean wind speed M2, the power plant turbulence level parameter Ir, and the turbulence standard deviation std using the formula M3=M2+Ir×(0.75×M2+5.6)×2; and calculating the predicted extreme wind speed MS according to the formula MS=M2+3×M3. If the predicted extreme wind speed MS is greater than the wind speed warning threshold, a strong wind warning process is executed.

[0004] The above method uses a Kalman filter model to obtain the average wind speed during the prediction period, and then determines the predicted extreme wind speed. By comparing the predicted extreme wind speed with the wind speed warning threshold, a strong wind warning is achieved. However, the above method ignores the influence of the stability of the photovoltaic support structure on its wind resistance capability, which may cause false alarms and thus lead to inaccurate wind resistance warning results. Summary of the Invention

[0005] To address the technical problem of inaccurate wind resistance warning results, this application provides a wind resistance warning method and system for flexible photovoltaic supports, which can improve the accuracy of wind resistance warnings for flexible photovoltaic supports.

[0006] In a first aspect, this application provides a wind resistance early warning method for a flexible photovoltaic (PV) support. The method includes: determining the stability of each connection point based on vibration data from vibration sensors at each connection point of the PV support; obtaining the predicted wind speed and direction for the next moment using a time-series prediction model; using the orientation of the PV support and the tilt angle of the solar panels at the current moment as attitude information, and calculating the attention level of each connection point based on the attitude information and the predicted wind direction, including: inputting the attitude information and the predicted wind direction into a ranking model, outputting a ranking result of the force magnitude of each connection point, where the attention level of each connection point is negatively correlated with its ranking in the ranking result; calculating the predicted wind load for the next moment based on the predicted wind speed, predicted wind direction, and attitude information, and calculating the local risk value of each connection point, where the local risk value is positively correlated with the predicted wind load and negatively correlated with stability; weighted summing of the risk values ​​of each connection point using normalized attention levels to obtain an overall risk value; and issuing an early warning in response to the overall risk value exceeding a risk threshold.

[0007] Preferably, determining the stability of each connection part based on the vibration data of the vibration sensors at each connection part of the photovoltaic support includes: performing a Fourier transform on the vibration data to obtain the energy value of each connection part at different vibration frequencies; using the normalized energy value as a weight to calculate the center vibration frequency of each connection part, and the stability of each connection part is negatively correlated with the center vibration frequency.

[0008] When the stability of the connection is high, the energy value is concentrated in the low-frequency region, and the central vibration frequency is a small value. When the connection becomes loose, cracked, or changes in stiffness, the energy value diffuses to the mid-to-high frequency region, and the central vibration frequency will increase. The stability of each connection can be accurately measured based on the central vibration frequency of each connection, providing a data basis for subsequent wind resistance early warning.

[0009] Preferably, the connection part stability for: ; For connection parts The central vibration frequency, This represents the maximum vibration frequency.

[0010] Preferably, the training method of the time series prediction model includes: collecting wind speed and wind direction sequences within a historical time period, and using the wind speed and direction at the next adjacent moment of the historical time period as numerical labels; inputting the wind speed and wind direction sequences within the historical time period into the time series prediction model to obtain the output results; and iterating the time series prediction model multiple times based on the mean squared error loss between the output results and the numerical labels until the number of iterations is greater than a preset number, or the mean squared error loss is less than a preset loss, thus completing the training.

[0011] Preferably, the connection part Attention level Satisfying the relation: ; For connection parts Rank in the sorting results This refers to the number of each connection part.

[0012] The higher a connection point ranks in the ranking results, the greater the force exerted by the wind on that connection point, and the more likely that connection point is to tip over due to strong winds. Therefore, higher-ranked connection points should be given more attention.

[0013] Preferably, the training method of the sorting model includes: collecting multiple training samples, the training samples including attitude information and predicted wind direction, and the force values ​​of each connection part under the attitude information and predicted wind direction; arranging each connection part in descending order of force value to obtain sorting labels; inputting the training samples into the sorting model to obtain sorting results; iteratively training the sorting model based on the sorting loss between the sorting results and the sorting labels until the sorting loss is less than the loss threshold, or the number of iterations is greater than the number of iterations threshold, to obtain the trained sorting model.

[0014] The ranking model is trained so that it can learn the ranking results under different combinations of attitude information and predicted wind direction, ensuring the accuracy of the attention of each connection part.

[0015] Preferably, the predicted wind load at the next moment Satisfying the relation: ;in, air density, For windward area, The drag coefficient of the solar panel. To predict wind speed, the windward area for , The area of ​​the solar panel. and These are the normal vector of the solar panel and the predicted wind direction, respectively. The normal vector of the solar panel is related to the attitude information.

[0016] Preferably, the connection part Local risk value for: ; To predict wind loads, For connection parts Stability.

[0017] Preferably, the overall risk value for: ; This represents the total number of connection points. For connection parts attention, This is the sum of the attention given to each connecting part. For connection parts The local risk value.

[0018] Since the predicted wind load on the solar panel is used as the force exerted by the wind on each connection part in the process of calculating the local risk value, the difference between the attitude information and the magnitude of the force on each connection part under the predicted wind is not considered. Furthermore, the risk value of each connection part is weighted and summed by normalized attention to obtain the overall risk value, so that the overall risk value can truly and accurately reflect the impact of wind on the photovoltaic support.

[0019] In a second aspect, this application also provides a wind resistance early warning system for a flexible photovoltaic support, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the wind resistance early warning method for a flexible photovoltaic support according to the first aspect of this application.

[0020] The technical solution of this application has the following beneficial technical effects: Photovoltaic (PV) support systems comprise multiple connection points. As the system ages, these connection points loosen to varying degrees, leading to differences in their wind resistance. Therefore, vibration data from each connection point is used to determine its stability for accurate early warning. Furthermore, the current posture of the PV support system and the predicted wind direction for the next moment are input into a sorting model to determine the relative forces acting on each connection point, allocating greater attention to those with higher forces. Based on the predicted wind load for the next moment and the stability of each connection point, a local risk value is calculated for each connection point. This local risk value characterizes the impact of the predicted wind load on each connection point. The risk values ​​of each connection point are then weighted and summed using normalized attention values ​​to obtain the overall risk value. Wind resistance warnings are then implemented based on this overall risk value. By comprehensively considering the wind resistance of each connection point and accurately quantifying the overall risk value under the influence of high forces on each connection point in the next moment, the accuracy of wind resistance warnings for flexible PV support systems is improved. Attached Figure Description

[0021] Figure 1 This is a flowchart of a wind-resistant early warning method for a flexible photovoltaic support according to an embodiment of this application.

[0022] Figure 2 This is a structural block diagram of a wind-resistant early warning system for a flexible photovoltaic bracket according to an embodiment of this application. Detailed Implementation

[0023] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] According to a first aspect of this application, this application provides a wind-resistant early warning method for flexible photovoltaic supports. Figure 1 This is a flowchart of a wind resistance early warning method for a flexible photovoltaic support according to an embodiment of this application. Figure 1 As shown, the wind resistance early warning method for the flexible photovoltaic support includes steps S101 to S106, which are described in detail below.

[0025] S101, determine the stability of each connection part based on the vibration data of the vibration sensor at each connection part of the photovoltaic bracket.

[0026] In one embodiment, the photovoltaic support is a whole formed by connecting parts such as bearings, hinges and screws. Vibration sensors are deployed at each connecting part to collect vibration data of each connecting part within a preset time period. The degree of looseness of each connecting part can be determined based on the vibration data. If a vibration data shows a large vibration, it indicates that the connecting part is loose and the stability of the connecting part is worse.

[0027] Specifically, determining the stability of each connection part based on the vibration data of the vibration sensors at each connection part of the photovoltaic support includes: performing a Fourier transform on the vibration data to obtain the energy value of each connection part at different vibration frequencies; using the normalized energy value as a weight to calculate the center vibration frequency of each connection part, and the stability of each connection part is negatively correlated with the center vibration frequency.

[0028] Among them, the connection part central vibration frequency Satisfying the relation: ; This represents the maximum vibration frequency. Vibration frequency Energy value, This is the sum of the energy values ​​at all vibration frequencies. When the connection is relatively stable, the energy value is concentrated in the low-frequency region, and the central vibration frequency is a relatively small value. When the connection becomes loose, cracked, or its stiffness changes, the energy value diffuses to the mid-to-high-frequency region, and the central vibration frequency increases. Therefore, the higher the central vibration frequency, the more unstable the corresponding connection. In other words, the stability of each connection is negatively correlated with the central vibration frequency. Specifically, the connection... stability for: ; For connection parts The central vibration frequency, This represents the maximum vibration frequency.

[0029] In this way, the stability of each connection part can be accurately measured based on the vibration data of the vibration sensors at each connection part. The worse the stability, the greater the possibility of loosening, cracking or stiffness change in the corresponding connection part, and the weaker the wind resistance of the corresponding connection part, thus providing a data basis for subsequent wind resistance early warning.

[0030] S102, using a time-series forecasting model to obtain the predicted wind speed and direction for the next moment.

[0031] In one embodiment, wind speed and wind direction sequences within a preset time period are collected, where the preset time period includes the current moment and multiple historical moments preceding the current moment. The wind speed and wind direction sequences are then input into a time-series prediction model to obtain the predicted wind speed and predicted wind direction for the next moment. In this embodiment, the preset time period includes 15 moments.

[0032] The time-series prediction model can be a recurrent neural network such as LSTM or GRU.

[0033] It should be noted that before using the time-series prediction model to obtain the predicted wind speed and direction for the next moment, the time-series prediction model needs to be trained. The training method for the time-series prediction model includes: collecting wind speed and direction sequences within a historical time period, and using the wind speed and direction at the next adjacent moment of the historical time period as numerical labels; inputting the wind speed and direction sequences within the historical time period into the time-series prediction model to obtain the output results; iterating the time-series prediction model multiple times based on the mean squared error loss between the output results and the numerical labels until the number of iterations is greater than a preset number, or the mean squared error loss is less than a preset loss, thus completing the training. The preset number of iterations is 300; the preset loss is 0.01.

[0034] In this way, by using the time-series prediction model to learn the changes in wind speed and direction, the predicted wind speed and direction for the next moment can be accurately obtained.

[0035] S103, the orientation of the photovoltaic bracket and the tilt angle of the solar panel at the current moment are used as attitude information, and the attention of each connection part is calculated based on the attitude information and the predicted wind direction.

[0036] In one embodiment, the orientation of the photovoltaic support can be adjusted by rotation, and the tilt angle of the solar panel mounted on the photovoltaic support can be adjusted by adjusting the pitch angle. The orientation of the photovoltaic support and the tilt angle of the solar panel at the current moment are used as attitude information.

[0037] Understandably, with the wind direction remaining constant, the stress on the photovoltaic support structure will differ under different attitude information, which in turn will lead to different stress magnitudes at each connection point. Connection points with greater stress are more likely to tip over due to stronger winds. Therefore, the stress magnitude at each connection point is related to the attitude information and the predicted wind direction, and greater attention should be paid to connection points with greater stress.

[0038] Specifically, calculating the attention level of each connection part based on the attitude information and predicted wind direction includes: inputting the attitude information and predicted wind direction into the ranking model, outputting the ranking results of the force magnitude of each connection part, and the attention level of each connection part is negatively correlated with the ranking in the ranking results.

[0039] Among them, the connection part Attention level Satisfying the relation: ; For connection parts Rank in the sorting results This refers to the number of each connection part.

[0040] For example, if the photovoltaic support system includes five connection points (A, B, C, D, and E), and these are arranged in descending order of stress, the sorting result is as follows: That is to say, D ranks 1 and C ranks 5; therefore, the attention given to the connecting part D is... The attention given to the connection part C is More attention is allocated to the top-ranked connection points.

[0041] The ranking model employs a fully connected neural network, comprising an input layer, a hidden layer, and an output layer. The input layer receives the current attitude information of the photovoltaic support and the predicted wind direction for the next moment. The hidden layer extracts features from the attitude information and the predicted wind direction, and inputs the feature extraction results into the output layer to obtain the ranking score of each connection part. The ranking result is obtained in descending order of ranking scores.

[0042] In one embodiment, before using the ranking model to obtain the attention level of each connection part, the ranking model needs to be trained to ensure the accuracy of the attention level for each connection part. The training method of the ranking model includes: collecting multiple training samples, which include attitude information and predicted wind direction, as well as the force values ​​of each connection part under the attitude information and predicted wind direction; arranging each connection part in descending order of force values ​​to obtain ranking labels; inputting the training samples into the ranking model to obtain ranking results; iteratively training the ranking model based on the ranking loss between the ranking results and the ranking labels until the ranking loss is less than a loss threshold, or the number of iterations is greater than a number threshold, to obtain the trained ranking model. The number threshold is set to 300; the loss threshold is set to 0.01. The sorting loss can be any existing sorting loss such as ListNet loss or listMLE loss, and this application does not impose any restrictions.

[0043] Thus, based on the current attitude information of the photovoltaic support and the predicted wind direction for the next moment, the attention level of each connection part is predicted. The greater the attention level, the greater the force on the corresponding connection part under the attitude information and predicted wind direction, and the more likely the connection part is to tilt.

[0044] S104. Calculate the predicted wind load for the next moment based on the predicted wind speed, predicted wind direction, and attitude information, and calculate the local risk value of each connection point. The local risk value is positively correlated with the predicted wind load and negatively correlated with stability.

[0045] In one embodiment, the local risk value is used to characterize the impact of the predicted wind load at the next moment on each connection point. Compared to the solar panels mounted on the photovoltaic support, the surface area of ​​the photovoltaic support and the wind is much smaller than the surface area of ​​the solar panels and the wind. The force of wind on the photovoltaic support is mainly caused by the contact surface between the solar panels and the wind. Therefore, to reduce the amount of calculation, when considering the force of wind on the photovoltaic support, the predicted wind load of the solar panels is regarded as the force of wind on the photovoltaic support.

[0046] Predicted wind load at the next moment Satisfying the relation: ;in, air density, For windward area, The drag coefficient of the solar panel. To predict wind speed, the windward area for , The area of ​​the solar panel. and These are the normal vector of the solar panel and the predicted wind direction, respectively. The normal vector of the solar panel is related to the attitude information.

[0047] The drag coefficient of the solar panel can be obtained by referring to a table based on the model of the solar panel, and will not be elaborated here.

[0048] Thus, the predicted wind load for the next moment is obtained, which can be regarded as the force exerted by the wind on the photovoltaic support at the next moment.

[0049] In one embodiment, the greater the predicted wind load, the greater the force borne by each connection part, and the greater the local risk value of each connection part. Therefore, the local risk value of each connection part is positively correlated with the predicted wind load. If the stability of the connection part is greater, it means that the wind resistance of the connection part is greater, and the local risk value of the connection part is smaller. Therefore, the local risk value of each connection part is positively correlated with the stability.

[0050] Connection parts Local risk value for: ; To predict wind loads, For connection parts Stability.

[0051] Thus, by combining the current attitude information of the photovoltaic support structure with the predicted wind speed and direction for the next moment, the local risk value of each connection part for the next moment is obtained. This local risk value is used to characterize the degree of influence of the predicted wind load for the next moment on each connection part.

[0052] S105 uses the normalized attention level to weighted summation of the risk values ​​of each connection part to obtain the overall risk value.

[0053] In one embodiment, since the predicted wind load on the solar panel is directly used as the force exerted by the wind on each connection point during the calculation of the local risk value, without considering the difference between attitude information and the magnitude of the force on each connection point under the predicted wind direction, the risk values ​​of each connection point are further weighted and summed using normalized attention levels to obtain the overall risk value. The overall risk value can accurately reflect the impact of wind on the photovoltaic support structure.

[0054] Overall risk value for: ; This represents the total number of connection points. For connection parts attention, This is the sum of the attention given to each connecting part. For connection parts The local risk value.

[0055] S106, issue an early warning when the overall risk value exceeds the risk threshold.

[0056] In one embodiment, the risk threshold is set to 0.8 times the maximum safe wind load.

[0057] According to a second aspect of this application, a wind-resistant early warning system for a flexible photovoltaic support is also provided. Figure 2 This is a structural block diagram of a wind-resistant early warning system for a flexible photovoltaic support according to an embodiment of this application. Figure 2 As shown, the system 50 includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a wind-resistant early warning method for a flexible photovoltaic support according to the first aspect of this application. The system also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface. Their configuration and functions are known in the art and will not be described further here.

[0058] It should be noted that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of this application, and these all fall within the scope of protection of this application.

Claims

1. A wind resistance early warning method for a flexible photovoltaic support, characterized in that, The early warning method includes: The stability of each connection part is determined based on the vibration data of the vibration sensors at each connection part of the photovoltaic support; the predicted wind speed and predicted wind direction at the next moment are obtained using a time series prediction model. The orientation of the photovoltaic support and the tilt angle of the solar panels at the current moment are used as attitude information. Based on the attitude information and the predicted wind direction, the attention of each connection part is calculated, including: inputting the attitude information and the predicted wind direction into the ranking model, and outputting the ranking result of the force magnitude of each connection part. The attention of each connection part is negatively correlated with the ranking in the ranking result. The predicted wind load for the next moment is calculated based on the predicted wind speed, predicted wind direction, and attitude information. The local risk value of each connection part is also calculated. The local risk value is positively correlated with the predicted wind load and negatively correlated with stability. The overall risk value is obtained by weighted summing of the risk values ​​of each connection point using normalized attention levels. An early warning is issued in response to the overall risk value exceeding the risk threshold.

2. The wind resistance early warning method for a flexible photovoltaic support according to claim 1, characterized in that, The stability of each connection point is determined based on vibration data from vibration sensors at each connection point of the photovoltaic support system, including: The vibration data is subjected to Fourier transform to obtain the energy values ​​of each connection part at different vibration frequencies; the normalized energy values ​​are used as weights to calculate the center vibration frequency of each connection part, and the stability of each connection part is negatively correlated with the center vibration frequency.

3. The wind resistance early warning method for a flexible photovoltaic support according to claim 2, characterized in that, Connection parts stability for: ; For connection parts The central vibration frequency, This represents the maximum vibration frequency.

4. The wind resistance early warning method for a flexible photovoltaic support according to claim 1, characterized in that, The training method for the time series prediction model includes: Collect wind speed and wind direction sequences within a historical time period, and use the wind speed and direction at the next adjacent moment of the historical time period as numerical labels; Input the wind speed and wind direction sequences from the historical time period into the time series prediction model to obtain the output results; perform multiple iterations on the time series prediction model based on the mean squared error loss between the output results and the numerical labels until the number of iterations is greater than the preset number, or the mean squared error loss is less than the preset loss, and the training is completed.

5. The wind resistance early warning method for a flexible photovoltaic support according to claim 1, characterized in that, Connection parts Attention level Satisfying the relation: ; For connection parts Rank in the sorting results This refers to the number of each connection part.

6. The wind resistance early warning method for a flexible photovoltaic support according to claim 1, characterized in that, The training method for the ranking model includes: Multiple training samples are collected, including attitude information and predicted wind direction, as well as the force values ​​of each connection part under the attitude information and predicted wind direction. The connection parts are arranged in descending order of force value to obtain sorting labels. The training samples are input into the ranking model to obtain the ranking results. The ranking model is then iteratively trained based on the ranking loss between the ranking results and the ranking labels until the ranking loss is less than the loss threshold or the number of iterations is greater than the number of iterations threshold, at which point the trained ranking model is obtained.

7. The wind resistance early warning method for a flexible photovoltaic support according to claim 1, characterized in that, Predicted wind load at the next moment Satisfying the relation: ;in, air density, For windward area, The drag coefficient of the solar panel. To predict wind speed, the windward area for , The area of ​​the solar panel. and These are the normal vector of the solar panel and the predicted wind direction, respectively. The normal vector of the solar panel is related to the attitude information.

8. The wind resistance early warning method for a flexible photovoltaic support according to claim 1, characterized in that, Connection parts Local risk value for: ; To predict wind loads, For connection parts Stability.

9. The wind resistance early warning method for a flexible photovoltaic support according to claim 1, characterized in that, Overall risk value for: ; This represents the total number of connection points. For connection parts attention, This is the sum of the attention given to each connecting part. For connection parts The local risk value.

10. A wind-resistant early warning system for a flexible photovoltaic support, characterized in that, It includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement a wind-resistant early warning method for a flexible photovoltaic support according to any one of claims 1 to 9.

Images