Tracking type photovoltaic support optimization method and system based on artificial intelligence

Abstract

The invention provides a tracking type photovoltaic support optimization method and system based on artificial intelligence, and the method comprises the steps: collecting environment illumination data in real time through a photovoltaic sensor array, inputting the environment illumination data into an illumination distribution prediction model, and generating prediction illumination distribution information; inputting the bracket state parameters into a bracket dynamic response model to generate bracket response characteristics; the optimization control module executes two types of model operation in a time-sharing manner, analyzes prediction information to obtain an optimal tracking angle, calculates a bracket adjusting amount according to a current position parameter and drives an angle adjusting mechanism to adjust the orientation of a photovoltaic panel; and the data acquisition unit monitors the actual power generation efficiency in real time and performs feedback correction, and generates updated parameters for next period optimization to complete the closed-loop optimization process. The photovoltaic power generation efficiency can be improved, and the adjustment precision and the system stability are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation, and more specifically, to an optimization method and system for a tracking type photovoltaic bracket based on artificial intelligence. Background Art

[0002] With the continuous growth of global energy demand and the increasing emphasis on environmental protection, photovoltaic power generation, as a clean and renewable energy technology, has received extensive attention. In a photovoltaic power generation system, the installation method of photovoltaic panels has a significant impact on their power generation efficiency. To improve the power generation efficiency, a tracking type photovoltaic bracket has emerged, which adjusts the angle of the photovoltaic panel to always face the sun, thereby maximizing the reception of solar radiation. Most of the existing tracking type photovoltaic brackets adopt simple mechanical tracking methods, such as single-axis or double-axis tracking systems. Although these systems can improve the power generation efficiency to a certain extent, they have some limitations. For example, they usually can only be adjusted according to a preset trajectory or simple sensor feedback, unable to accurately sense the changes in ambient light in real time, nor can they consider factors such as the dynamic state of the photovoltaic panel and historical adjustment records. In addition, the performance of these traditional tracking systems is often not ideal under complex weather conditions (such as cloudy days), unable to quickly adapt to the rapid changes in light intensity and direction, resulting in a reduction in power generation efficiency.

[0003] In the process of implementing the embodiments of the present invention, the inventors found that there are at least the following problems or defects in the prior art: the existing tracking type photovoltaic brackets cannot make full use of ambient light data and photovoltaic panel state parameters for real-time optimization, lacking accurate prediction of light distribution and comprehensive consideration of the dynamic response of the bracket, resulting in limited improvement in power generation efficiency in complex environments and unable to achieve intelligent optimization control. Summary of the Invention

[0004] The present invention provides an optimization method and system for a tracking type photovoltaic bracket based on artificial intelligence.

[0005] In the first aspect of the present invention, an optimization method for a tracking type photovoltaic bracket based on artificial intelligence is provided, including: A photovoltaic sensor array real-time collects ambient light data, inputs the ambient light data into a light distribution prediction model to generate predicted light distribution information, inputs the bracket state parameters into a bracket dynamic response model to generate bracket response characteristics, and uses an optimization control module to execute two types of model operations at different times; The tracking type photovoltaic bracket is equipped with two types of components, an angle adjustment mechanism and a data collection unit. Among them, the data collection unit receives the predicted light distribution information, and the tracking type photovoltaic bracket analyzes the predicted light distribution information to obtain the optimal tracking angle; The tracking-type photovoltaic support calculates the support adjustment amount according to the current position parameter and the optimal tracking angle; The tracking-type photovoltaic support drives the angle adjustment mechanism to adjust the orientation of the photovoltaic panel according to the support adjustment amount; The tracking-type photovoltaic support monitors the actual power generation efficiency in real time through the data acquisition unit, performs feedback correction on the actual power generation efficiency, generates update parameters required for the tracking-type photovoltaic support to execute the next cycle of optimization, and completes the support optimization process.

[0006] Further, the predicted light distribution information is numerical data with high spatio-temporal resolution including environmental light intensity, incident angle, and time series characteristics.

[0007] Further, the support state parameter is structured data with dynamic change characteristics including the current angle of the photovoltaic panel, mechanical stress, and historical adjustment records, and does not include the predicted light distribution information.

[0008] Further, inputting the environmental light data into the light distribution prediction model to generate predicted light distribution information includes: Performing normalization processing on the environmental light data to obtain a standardized light sequence; Performing time series feature extraction on the standardized light sequence to obtain a characterized light sequence; performing noise suppression on the characterized light sequence to obtain a denoised light sequence; Adding a timestamp label and a spatial position label to the denoised light sequence to obtain an annotated light sequence; Inputting the annotated light sequence into the light distribution prediction model for prediction to obtain predicted light distribution information; Performing confidence evaluation on the predicted light distribution information to obtain an optimization weight coefficient; Generating the final predicted light distribution information according to the optimization weight coefficient.

[0009] Further, the normalization processing of the environmental light data is realized through the following process: Converting the environmental light data into an input sequence , calculating the normalization parameters and , where is the sequence mean, is the sequence standard deviation, and outputting the normalized sequence satisfies the relationship:

[0010]

[0011] where, is the length of the input sequence; represents the th ambient light data in the input sequence; represents the th data in the normalized sequence; is the average value of the input sequence , calculated by the formula

[0012] ; is the standard deviation of the input sequence , calculated by the formula

[0013] ;

[0014] Further, add timestamp tags and spatial position tags to the denoised light sequence to obtain the labeled light sequence. Among them, the timestamp tags provide temporal correlation information for the model, and the spatial position tags provide regional distribution information for the model. The denoised light sequence carries the original ambient light data.

[0015] Further, parsing the predicted light distribution information includes: Performing feature mapping on the predicted light distribution information to obtain a multi-dimensional feature vector; Performing gradient backpropagation on the multi-dimensional feature vector to update the parameters of the light distribution prediction model; Performing light intensity interpolation calculation on the updated light distribution prediction model to obtain the interpolated light distribution; Performing spatial clustering analysis on the interpolated light distribution to obtain regional light hotspots; Performing angular projection transformation on the regional light hotspots to obtain a set of candidate tracking angles; Performing power generation efficiency simulation on the set of candidate tracking angles to obtain the optimal tracking angle; Input the optimal tracking angle into the angle adjustment mechanism control instruction generation module.

[0016] Further, the gradient backpropagation is achieved in the following way: For the prediction error , with the learning rate as the update step size and the weight matrix as the model parameters, update the parameters of the th layer of the light distribution prediction model, satisfying the relationship:

[0017] where is the number of iterations; Indicates the weight matrix of the layer at the th iteration; Indicates the weight matrix of the layer after being updated at the th iteration; is the learning rate, which is used to control the step size of each parameter update; is the error with respect to the layer at the th iteration, and the update process continues until the prediction error is less than the set threshold.

[0018] Further, based on the model parameters after the gradient backpropagation is completed, within a subsequent time window with a length of , perform multi-step rolling optimization of the predicted illumination distribution information, that is, at each time step , predict the future p-step illumination distribution according to the historical data , where the rolling optimization objective function is:

[0019] where, is the time window length; is the length of the historical data used for prediction; is the current time step; is the prediction step size; is the actual illumination distribution data at the th future step; is the predicted illumination distribution data at the th future step, and the prediction effect of the illumination distribution prediction model is optimized by minimizing this objective function.

[0020] In the second aspect of the present invention, a tracking type photovoltaic bracket optimization system based on artificial intelligence is provided, including: a photovoltaic sensor array, an angle adjustment mechanism, and an optimization control module. Among them, the photovoltaic sensor array generates two types of data, namely, the ambient illumination data collected in real time and the photovoltaic panel state parameters, which are transmitted to the optimization control module through the same data interface at different times; The optimization control module runs two types of models, namely, an illumination distribution prediction model and a bracket dynamic response model. The optimal tracking angle is obtained by parsing through the illumination distribution prediction model, the bracket adjustment amount is calculated and generated, and the angle adjustment mechanism is controlled to adjust the orientation of the photovoltaic panel. The actual power generation efficiency feedback value is calculated through the bracket dynamic response model to complete the bracket optimization.

[0021] The above embodiments of the present invention have at least the following beneficial effects: The tracking photovoltaic support optimization method and system of the present invention can improve the power generation efficiency of the photovoltaic power generation system. By collecting ambient light data in real time through a photovoltaic sensor array and combining it with a light distribution prediction model, it is possible to generate prediction light distribution information with high spatio-temporal resolution, thereby providing a more accurate tracking angle for the photovoltaic support. At the same time, the support dynamic response model can consider dynamic parameters such as the current angle of the photovoltaic panel, mechanical stress, and historical adjustment records, making the adjustment of the support more scientific and reasonable, adapting to complex environmental changes, and further improving the power generation efficiency. In addition, the optimization control module executes two types of model operations at different times, which can efficiently coordinate light prediction and support response to ensure the stable operation of the entire system.

[0022] The present invention can also achieve intelligent management and optimization of the photovoltaic power generation system. The tracking photovoltaic support monitors the actual power generation efficiency in real time through a data acquisition unit, performs feedback correction on the actual power generation efficiency, and generates updated parameters required for the next cycle of optimization, thereby realizing a dynamic optimization process. This can not only improve the adaptive ability of the system, but also reduce manual intervention and lower the operation and maintenance costs. At the same time, the gradient backpropagation and multi-step rolling optimization of the light distribution prediction model can further improve the prediction accuracy of the model and the overall performance of the system, providing a strong guarantee for the efficient operation of the photovoltaic power generation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features, and advantages of the exemplary embodiments of the present invention will become readily understood. In the drawings, several embodiments of the present invention are shown by way of illustration and not limitation, wherein: Figure 1 It is a schematic flow chart of a method for optimizing a tracking photovoltaic support based on artificial intelligence provided by an embodiment of the present invention; Figure 2 It is a schematic structural diagram of a system for optimizing a tracking photovoltaic support based on artificial intelligence provided by an embodiment of the present invention; Figure 3 It schematically shows a schematic structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and then implement the present invention, and do not limit the scope of the present invention in any way. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to be able to convey the scope of the present invention fully to those skilled in the art.

[0025] Those skilled in the art know that the embodiments of the present invention can be implemented as a system, a device, equipment, a method, or a computer program product. Therefore, the present invention can be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0026] It should be noted that any number of elements in the drawings is for illustration rather than limitation, and any naming is only for distinction and does not have any limiting meaning.

[0027] The following refers to Figure 1 , Figure 1 which is a schematic flowchart of an optimization method for a tracking photovoltaic bracket based on artificial intelligence provided for an embodiment of the present invention. As Figure 1 shown, an optimization method for a tracking photovoltaic bracket based on artificial intelligence includes: S1 The photovoltaic sensor array collects environmental light data in real time, inputs the environmental light data into the light distribution prediction model to generate predicted light distribution information, inputs the bracket state parameters into the bracket dynamic response model to generate bracket response characteristics, and uses the optimization control module to execute two types of model operations at different times; S2 The tracking photovoltaic bracket is equipped with two types of components, an angle adjustment mechanism and a data collection unit. Among them, the data collection unit receives the predicted light distribution information, and the tracking photovoltaic bracket analyzes the predicted light distribution information to obtain the optimal tracking angle; S3 The tracking photovoltaic bracket calculates the bracket adjustment amount according to the current position parameters and the optimal tracking angle; S4 The tracking photovoltaic bracket drives the angle adjustment mechanism to adjust the orientation of the photovoltaic panel according to the bracket adjustment amount; S5 The tracking photovoltaic bracket monitors the actual power generation efficiency in real time through the data collection unit, and performs feedback correction on the actual power generation efficiency to generate update parameters required for the tracking photovoltaic bracket to execute the next cycle of optimization, completing the bracket optimization process.

[0028] It should be noted that the present invention proposes an optimization method for a tracking photovoltaic support based on artificial intelligence. The core lies in real-time collecting environmental light data through a photovoltaic sensor array and inputting this data into a light distribution prediction model to generate predicted light distribution information. Here, the photovoltaic sensor array refers to multiple sensors installed in a photovoltaic power generation system. They can detect parameters such as environmental light intensity and incident angle in real time and convert this data into electrical signals for output. The light distribution prediction model is a model constructed based on artificial intelligence algorithms. It can predict the future light distribution based on the input light data. At the same time, the support state parameters are input into the support dynamic response model to generate support response characteristics. Here, the support state parameters include the current angle of the photovoltaic panel, mechanical stress, and historical adjustment records, etc. These parameters can reflect the dynamic changes of the support. Through the optimization control module executing two types of model operations at different times, dynamic optimization control of the photovoltaic support is achieved, thereby improving the power generation efficiency of the photovoltaic power generation system.

[0029] Specifically, the photovoltaic sensor array can include multiple different types of sensors, such as light intensity sensors, angle sensors, etc. They are distributed at different positions of the photovoltaic panel and can collect environmental light data in real time. The data collected by these sensors, after preliminary processing, forms an environmental light data sequence and then is input into the light distribution prediction model. The light distribution prediction model can be constructed using deep learning algorithms, such as long short-term memory network LSTM or convolutional neural network CNN. These models can effectively extract the time series characteristics and spatial distribution characteristics of the light data. The current angle of the photovoltaic panel in the support state parameters refers to the inclination angle of the photovoltaic panel relative to the horizontal plane. The mechanical stress refers to the physical stress suffered by the support during long-term operation. The historical adjustment record refers to the adjustment action record of the support in the past period of time. These parameters together reflect the dynamic state of the support. The optimization control module can adopt a time-sharing scheduling algorithm. According to the output results of the light distribution prediction model and the support dynamic response model, it controls the actions of the angle adjustment mechanism and the data collection unit at different times, thereby achieving dynamic optimization control of the photovoltaic support.

[0030] Preferably, the construction process of the light distribution prediction model may include the following steps: First, normalize the collected environmental light data, that is, convert the original data into a standardized light sequence to eliminate the dimensional differences of the data collected by different sensors. Then, extract the temporal features of the standardized light sequence, for example, extract the time series features of the light data through a sliding window algorithm. Next, suppress the noise of the featureized light sequence, for example, use wavelet transform or Kalman filter algorithm to remove the noise in the light data. After that, add timestamp tags and spatial location tags to the denoised light sequence to provide temporal correlation information and regional distribution information for the model. Finally, input the labeled light sequence into the light distribution prediction model for training and prediction to obtain the predicted light distribution information. During the data processing, for example, during the normalization process, calculate the normalization parameters, that is, the sequence mean and standard deviation, and then convert the original data into a normalized sequence according to these parameters. For the confidence evaluation of the predicted light distribution information, the optimization weight coefficient can be calculated according to the prediction error and confidence interval of the model, so as to generate the final predicted light distribution information.

[0031] In some embodiments, the predicted light distribution information is numerical data with high spatio-temporal resolution that includes environmental light intensity, incident angle, and time series features.

[0032] It should be noted that the predicted light distribution information is numerical data with high spatio-temporal resolution that includes environmental light intensity, incident angle, and time series features. This means that this information not only covers the intensity of the light, but also includes the incident angle of the light and the features that change over time. Such high spatio-temporal resolution data can provide a more accurate reference for the optimization of the photovoltaic support, enabling the support to dynamically adjust the angle of the photovoltaic panel according to the change of light, thereby improving the power generation efficiency. High spatio-temporal resolution data means that the data has high precision both in time and space and can more precisely reflect the change of light.

[0033] Specifically, the environmental light intensity refers to the solar radiation power received per unit area, usually measured in watts per square meter (W / m²). The incident angle refers to the angle between the sun's rays and the normal of the photovoltaic panel plane, and this angle directly affects the light reception efficiency of the photovoltaic panel. The time series features reflect the change laws of the light intensity and incident angle over time, such as the change of light intensity at different time periods of a day. These parameters together constitute the predicted light distribution information and provide a comprehensive basis for the optimization of the photovoltaic support. For example, by analyzing the time series features, the change trend of the light intensity at different time periods can be predicted, so as to adjust the angle of the photovoltaic panel in advance to keep it in the best reception state all the time.

[0034] Preferably, the generation process of the predicted light distribution information can be further refined. First, the collected ambient light data is preprocessed, including operations such as removing noise and filling in missing values, to ensure the accuracy and integrity of the data. Then, machine learning algorithms, such as support vector machine (SVM) or neural network, are used to model and predict the preprocessed data. During the modeling process, the input parameters include historical light intensity data, geographical location information, seasonal factors, etc. These parameters jointly affect the prediction result of the light distribution. For example, geographical location information can provide information such as local sunshine duration and solar altitude angle, while seasonal factors affect the variation law of light intensity. Through these input parameters, the model can generate the predicted light distribution information including light intensity, incident angle, and time series characteristics, providing a scientific basis for the optimal control of the photovoltaic support.

[0035] In some embodiments, the support state parameter is structured data with dynamic change characteristics, including the current angle of the photovoltaic panel, mechanical stress, and historical adjustment records, and does not include the predicted light distribution information.

[0036] It should be noted that the support state parameter is structured data with dynamic change characteristics, including the current angle of the photovoltaic panel, mechanical stress, and historical adjustment records, and does not include the predicted light distribution information. This means that the support state parameter mainly focuses on the operating state and historical behavior of the photovoltaic panel itself, rather than the external environmental light conditions. Through these parameters, the system can better understand the current state of the photovoltaic panel and its past adjustment behavior, so as to perform dynamic adjustment more accurately and ensure that the photovoltaic panel can maintain the best working state under different environmental conditions.

[0037] Specifically, the current angle of the photovoltaic panel refers to the tilt angle of the photovoltaic panel relative to the horizontal plane, usually expressed in degrees. This angle directly affects the sunlight reception efficiency of the photovoltaic panel. Mechanical stress refers to the physical stress borne by the photovoltaic panel and its support during long-term operation, including stress generated by factors such as wind force and gravity, usually expressed in stress units such as Pascal. The historical adjustment record records the angle adjustment actions of the photovoltaic panel in the past period, including information such as the time of adjustment and the magnitude of the adjusted angle. These parameters jointly constitute the support state parameter, providing comprehensive input data for the dynamic response model of the photovoltaic support. For example, by analyzing the historical adjustment record, the system can understand the adjustment law of the photovoltaic panel under different environmental conditions, so as to optimize future adjustment strategies.

[0038] Preferably, the process of collecting and processing the support state parameters can be further refined. First, the current angle of the photovoltaic panel is collected in real time by an angle sensor installed on the photovoltaic panel, and the mechanical stress is measured by sensors such as strain gauges. These sensors convert the collected data into electrical signals and transmit them to the data acquisition unit. The data acquisition unit performs preliminary processing on this data, such as filtering and data format conversion, to ensure the accuracy and availability of the data. Then, the processed data is input into the support dynamic response model. This model can be constructed using a physical model or a data-driven model, such as a physical model based on finite element analysis or a data-driven model based on machine learning. In the model, the input parameters include the current angle of the photovoltaic panel, mechanical stress, and historical adjustment records, etc. The model calculates the response characteristics of the support based on these parameters, such as the adjustment speed, adjustment direction, etc. In this way, the system can more accurately control the angle adjustment of the photovoltaic panel and improve the overall performance of the photovoltaic power generation system.

[0039] In some embodiments, inputting the environmental light data into a light distribution prediction model to generate predicted light distribution information includes: Performing normalization processing on the environmental light data to obtain a normalized light sequence; Performing time series feature extraction on the normalized light sequence to obtain a featureized light sequence; performing noise suppression on the featureized light sequence to obtain a denoised light sequence; Adding timestamp tags and spatial position tags to the denoised light sequence to obtain an annotated light sequence; Inputting the annotated light sequence into a light distribution prediction model for prediction to obtain predicted light distribution information; Performing confidence evaluation on the predicted light distribution information to obtain an optimized weight coefficient; Generating final predicted light distribution information according to the optimized weight coefficient.

[0040] It should be noted that the process of inputting environmental light data into a light distribution prediction model to generate predicted light distribution information includes steps such as normalizing the environmental light data, extracting time series features, suppressing noise, adding timestamp and spatial position tags, etc. The purpose of these steps is to improve the quality and availability of the light data, making it more suitable for predicting light distribution. Normalization processing can eliminate the dimensional difference of the data, time series feature extraction can capture the time variation law of the light data, noise suppression can remove the interference in the data, and adding timestamp and spatial position tags provides richer context information for the model. Finally, an optimized weight coefficient is generated through confidence evaluation to adjust the reliability of the prediction result.

[0041] Specifically, the environmental light data refers to the original light intensity data collected by the photovoltaic sensor array, which usually contains random noise and dimensional differences. The normalization process is to convert these data into a dimensionless standardized sequence by calculating the mean and standard deviation of the data. The time series feature extraction is to analyze the standardized light sequence through a sliding window to extract the changing trend of the light intensity over time. The noise suppression uses filtering algorithms such as Kalman filtering or wavelet transform to remove the high-frequency noise in the light data. The timestamp label adds specific time information to each data point to reflect the time order of the light data; the spatial position label identifies the geographical location of the data collection point to help the model understand the spatial differences in the light distribution. These processed data are input into the light distribution prediction model to generate the predicted light distribution information including the light intensity, incident angle, and time series features.

[0042] Preferably, the construction of the light distribution prediction model can adopt deep learning algorithms such as Long Short-Term Memory Network (LSTM) or Convolutional Neural Network (CNN). In the process of model construction, first, the collected environmental light data is preprocessed, including normalization and noise suppression. The normalization process converts the original data into a standardized sequence by calculating the mean and standard deviation of the data, making the data collected by different sensors comparable. The noise suppression removes the random fluctuations in the data through a filtering algorithm to improve the accuracy of the data. Then, the time series features of the processed data are extracted, for example, the changing trend of the light intensity is extracted through the sliding window algorithm. Then, a timestamp and a spatial position label are added to each data point to enhance the model's ability to understand the time and space of the light data. Finally, these labeled data are input into the light distribution prediction model for training. The model generates the predicted light distribution information by learning the time series features and spatial distribution features of the light data. During the prediction process, the model also evaluates the confidence of the prediction results and generates optimized weight coefficients to adjust the reliability of the prediction results, thereby improving the prediction accuracy.

[0043] In some embodiments, the normalization process of the environmental light data is achieved through the following process: Convert the environmental light data into an input sequence , calculate the normalization parameters and , where is the sequence mean, is the sequence standard deviation, and output the normalized sequence satisfies the relationship:

[0044]

[0045] where, is the length of the input sequence; represents the th ambient light data in the input sequence; represents the th data in the normalized sequence; is the average value of the input sequence , calculated by the formula

[0046] ; is the standard deviation of the input sequence , calculated by the formula

[0047] ;

[0048] It should be noted that the normalization of ambient light data is a process of converting raw data into a standardized sequence for subsequent processing and analysis. The normalization process calculates the mean and standard deviation of the sequence to convert the data into a dimensionless form, thereby eliminating the dimensional differences between data collected by different sensors. This processing method can improve the comparability of data and the training effect of the model, providing higher-quality input data for the light distribution prediction model.

[0049] Specifically, the key parameters involved in the normalization process include the sequence mean and standard deviation. The sequence mean is calculated by adding up the values of all data points in the sequence and then dividing by the total number of data points, which reflects the average level of the data. The standard deviation is obtained by calculating the sum of the squares of the differences between each data point and the mean, dividing by the total number of data points, and then taking the square root, which reflects the degree of dispersion of the data. During the normalization process, each data point is transformed by subtracting the mean and dividing by the standard deviation to obtain the normalized data sequence. This processing method ensures that data collected by different sensors can be compared and analyzed on the same scale, providing a basis for the accuracy and reliability of the light distribution prediction model.

[0050] Preferably, the specific steps of the normalization process can be further refined. First, convert the ambient light data into an input sequence, and calculate the mean and standard deviation of the sequence. Then, normalize each data point according to the mean and standard deviation to obtain the normalized sequence. This process can be implemented through programming. For example, use the NumPy library in Python to efficiently calculate the mean and standard deviation and normalize the data. In practical applications, the normalization process is not only applicable to light intensity data but can also be extended to other light-related parameters such as temperature, humidity, etc. to improve the comprehensive performance of the model. In this way, the normalization process can provide high-quality input data for the light distribution prediction model, thereby improving the prediction accuracy and stability of the model.

[0051] In some embodiments, add a timestamp label and a spatial position label to the denoised light sequence to obtain an annotated light sequence. Among them, the timestamp label provides temporal correlation information for the model, and the spatial position label provides regional distribution information for the model. The denoised light sequence carries the original ambient light data.

[0052] It should be noted that adding a timestamp label and a spatial position label to the denoised light sequence is to provide richer context information for the light distribution prediction model. The timestamp label can reflect the time order of the light data and help the model understand the variation law of the light intensity over time; the spatial position label identifies the geographical location of the data collection point, enabling the model to capture the spatial differences in the light distribution. The addition of these labels helps to improve the accuracy and reliability of the light distribution prediction, enabling the model to more accurately predict the light conditions at different times and locations.

[0053] Specifically, the timestamp label refers to attaching a specific time mark to each light data point, usually represented in the form of a timestamp, such as 2025-04-25 12:00:00. The spatial position label refers to attaching a geographical location mark to each light data point, usually represented in the form of latitude and longitude coordinates, such as 30.5728°N, 114.2984°E. The addition process of these labels can be implemented through data processing software. For example, use the Pandas library in Python to label and organize the data. The timestamp label reflects the time attribute of the light data and helps the model identify the variation law of the light intensity during the day; the spatial position label reflects the geographical attribute of the light data and helps the model identify the light differences at different geographical locations. The combination of these labels provides more comprehensive input information for the light distribution prediction model, enabling it to more accurately predict the light distribution.

[0054] Preferably, the process of adding timestamp tags and spatial location tags can be further refined. First, obtain the timestamp and spatial location information of the light data from the data acquisition system. The timestamp can be obtained through the system clock, and the spatial location can be obtained through the GPS module. Then, store this information together with the light data in a data structure, such as using a Pandas Data Frame. In the data preprocessing stage, after denoising the light data, attach the timestamp and spatial location tags to each data point. For example, the assign method of Pandas can be used to add new columns to the Data Frame to represent the timestamp and spatial location respectively. The addition of these tags not only provides richer context information for the light distribution prediction model but also facilitates subsequent data analysis and model optimization. In this way, the model can better understand and predict the changes in light distribution, thereby improving the overall performance of the photovoltaic power generation system.

[0055] In some embodiments, parsing the predicted light distribution information includes: Perform feature mapping on the predicted light distribution information to obtain a multi-dimensional feature vector; Perform gradient backpropagation on the multi-dimensional feature vector to update the parameters of the light distribution prediction model; Perform light intensity interpolation calculation on the updated light distribution prediction model to obtain the interpolated light distribution; Perform spatial clustering analysis on the interpolated light distribution to obtain regional light hotspots; Perform angular projection transformation on the regional light hotspots to obtain a set of candidate tracking angles; Perform power generation efficiency simulation on the set of candidate tracking angles to obtain the optimal tracking angle; Input the optimal tracking angle into the angle adjustment mechanism control instruction generation module.

[0056] It should be noted that the process of parsing the predicted light distribution information includes a series of steps such as feature mapping, gradient backpropagation, light intensity interpolation calculation, spatial clustering analysis, angular projection transformation, and power generation efficiency simulation on the predicted light distribution information. The purpose of these steps is to extract the most valuable information from the predicted light distribution information to determine the optimal tracking angle of the photovoltaic panel. Through these steps, the system can adjust the angle of the photovoltaic panel more accurately, thereby maximizing the power generation efficiency.

[0057] Specifically, feature mapping is the process of converting predicted light distribution information into multi-dimensional feature vectors, which can more effectively represent the characteristics of light distribution. Gradient backpropagation is an optimization algorithm used to update the parameters of the light distribution prediction model to improve the prediction accuracy of the model. Light intensity interpolation calculation is to perform spatial interpolation on light intensity through mathematical methods to fill in the data blank areas. Spatial clustering analysis is the process of dividing light intensity data into several regional light hotspots, and these hotspot areas represent regions with higher light intensity. Angular projection transformation is the process of converting these hotspot areas into a set of candidate tracking angles. Power generation efficiency simulation is to simulate these candidate tracking angles to determine the optimal tracking angle. These steps together constitute a complete process for analyzing and predicting light distribution information, providing a scientific basis for the optimization of photovoltaic brackets.

[0058] Preferably, the process of analyzing and predicting light distribution information can be further refined. First, convert the predicted light distribution information into multi-dimensional feature vectors through feature mapping, and these feature vectors can include light intensity, incident angle, time series features, etc. Then, use the gradient backpropagation algorithm to update the parameters of the light distribution prediction model to reduce the prediction error. Specifically, for the prediction error, with the learning rate as the update step size, iterate and update the weight matrix of the model until the prediction error is less than the set threshold. Next, perform light intensity interpolation calculation on the updated light distribution prediction model, such as using bilinear interpolation or spline interpolation methods, to fill in the data blank areas. Then, perform spatial clustering analysis on the interpolated light distribution, such as using the K-means clustering algorithm, to divide the light intensity data into several regional light hotspots. After that, perform angular projection transformation on these hotspot areas to obtain a set of candidate tracking angles. Finally, perform power generation efficiency simulation on the set of candidate tracking angles, such as by simulating the power generation of photovoltaic panels at different angles, to determine the optimal tracking angle. Through these refined steps, the system can more accurately determine the optimal tracking angle of the photovoltaic panel, thereby improving the power generation efficiency of the photovoltaic power generation system.

[0059] In some embodiments, the gradient backpropagation is implemented as follows: for the prediction error , with the learning rate as the update step size, and with the weight matrix as the model parameter, update the parameters of the th layer of the light distribution prediction model, satisfying the relationship:

[0060] where is the number of iterations; represents the weight matrix of the th layer at the th iteration; Indicates the weight matrix of the layer after the th iteration update; is the learning rate, which is used to control the step size of each parameter update; is the error with respect to the layer at the th iteration, and the update process continues until the prediction error is less than the set threshold.

[0061] It should be noted that backpropagation of gradients is a technique for optimizing the parameters of the light distribution prediction model. In this way, the model can adjust its own parameters according to the prediction error, thereby improving the prediction accuracy. In this process, the learning rate is a key parameter, which determines the step size of each parameter update. By controlling the size of the learning rate, it can be ensured that the model converges stably during training, avoiding training instability caused by too large a step size or too slow a training speed caused by too small a step size. Through backpropagation of gradients, the model can continuously optimize its own parameters to better adapt to the changes in the input data, thereby improving the accuracy of light distribution prediction.

[0062] Specifically, the key parameters involved in backpropagation of gradients include prediction error, learning rate, and weight matrix. The prediction error refers to the difference between the model prediction value and the actual value, and it is an important indicator to measure the performance of the model. The learning rate is a hyperparameter used to control the step size of each parameter update. The weight matrix is the parameter of the model, which determines the output of the model. In the process of backpropagation of gradients, first calculate the partial derivative of the prediction error with respect to the weight matrix, and then update the weight matrix according to the learning rate and the partial derivative. This process will continue until the prediction error is less than the set threshold. In this way, the model can continuously optimize its own parameters to improve the prediction accuracy. In practical applications, the setting of the learning rate needs to be adjusted according to specific situations to ensure that the model can converge stably during training.

[0063] Preferably, the specific implementation process of gradient backpropagation can be further refined. First, select a suitable learning rate, for example, determine the optimal learning rate through methods such as cross-validation. Then, in each training iteration, calculate the partial derivative of the prediction error with respect to the weight matrix. Specifically, for the prediction error, use the learning rate as the update step size to iteratively update the weight matrix of the model. For example, for the parameters of the l-th layer of the model, at the k-th iteration, update the weight matrix according to the partial derivative of the prediction error with respect to the weight matrix. The update process continues until the prediction error is less than the set threshold. In practical applications, momentum methods or adaptive learning rate methods can be used to further improve the training efficiency and stability. Through these refined steps, the model can more effectively optimize its own parameters, thereby improving the accuracy and reliability of the light distribution prediction.

[0064] In some embodiments, based on the model parameters after the gradient backpropagation is completed, within a time window with a subsequent input length of , perform multi-step rolling optimization of the light distribution prediction information, that is, at each time step , according to the historical data predict the light distribution in the next p steps , where the rolling optimization objective function is:

[0065] where, is the time window length; is the length of the historical data used for prediction; is the current time step; is the prediction step size; is the actual light distribution data at the -th step in the future; is the predicted light distribution data at the -th step in the future. By minimizing this objective function, the prediction effect of the light distribution prediction model is optimized.

[0066] It should be noted that multi-step rolling optimization is an optimization method for the light distribution prediction model. It predicts the light distribution in the next multiple steps based on historical data at each time step, thereby improving the adaptability and prediction accuracy of the model to light changes. This method is particularly suitable for scenarios where the light conditions change rapidly, such as cloudy weather or environments with obstacles. By minimizing the objective function of the prediction error, the model can continuously adjust its own parameters to improve the accuracy and reliability of the prediction.

[0067] Specifically, the key parameters involved in multi-step rolling optimization include the time window length, historical data length, current time step, prediction step length, and objective function. The time window length refers to the length of consecutive time data used for prediction, which determines the amount of data that the model can use at each time step. The historical data length refers to the number of historical data points used to train the model, which affects the model's memory ability of past light changes. The current time step refers to the time point that the model is processing, and the prediction step length refers to the time span for the model to predict the future light distribution, which determines the prediction range of the model. The objective function is a mathematical expression used to evaluate the prediction accuracy. By minimizing the objective function, the model can optimize its own parameters. In practical applications, the time window length and prediction step length need to be adjusted according to specific scenarios. For example, in scenarios with rapid light changes, shorter time windows and longer prediction step lengths can be set.

[0068] Preferably, the specific implementation process of multi-step rolling optimization can be further refined. First, determine the time window length and historical data length, which determine the amount of data that the model can use at each time step. Then, at each time step, predict the future multi-step light distribution based on the historical data. Specifically, for the current time step, the model predicts the future light distribution based on the historical data, where the historical data length determines the number of past data points that the model can use, and the prediction step length determines the time range for the model to predict the future light distribution. The objective function is used to evaluate the accuracy of the prediction results. By minimizing the objective function, the model can continuously adjust its own parameters to improve the prediction accuracy. For example, the objective function can be defined as the sum of the squares of the errors between the predicted values and the actual values. In practical applications, optimization algorithms such as the gradient descent method can be used to minimize the objective function. In this way, the model can better adapt to the changes in light conditions, thereby improving the overall performance of the photovoltaic power generation system.

[0069] The above embodiments of the present invention have the following beneficial effects: The present invention can improve the energy capture efficiency of the photovoltaic power generation system. By using an artificial intelligence model to accurately predict the light distribution and combining the dynamic response characteristics of the bracket, the optimal tracking angle can be generated and the orientation of the photovoltaic panel can be adjusted in real time, thereby maximizing the light energy utilization rate. At the same time, data processing steps such as normalization processing, temporal feature extraction, and noise suppression can ensure the input quality of the prediction model, while gradient backpropagation and multi-step rolling optimization can continuously improve the model prediction accuracy, enabling the system to adapt to complex and variable environmental light conditions.

[0070] The present invention can reduce mechanical losses and extend the service life of equipment. By analyzing the support state parameters and historical adjustment records through a dynamic response model, the angle adjustment strategy can be optimized to avoid frequent ineffective actions. The confidence evaluation and feedback correction mechanism can correct prediction deviations in real time to ensure the reliability of adjustment instructions; spatial clustering and power generation efficiency simulation can accurately locate regional light hotspots and reduce energy losses caused by ineffective tracking. Finally, the closed-loop optimization process can continuously iterate and update parameters to keep the photovoltaic support in the best working state all the time.

[0071] As Figure 2 shown, an artificial intelligence-based optimized system for a tracking type photovoltaic support in some embodiments, the system includes: a photovoltaic sensor array 201, an angle adjustment mechanism 202, and an optimization control module 203. Among them, the photovoltaic sensor array generates two types of data, namely, ambient light data collected in real time and photovoltaic panel state parameters, which are transmitted to the optimization control module through the same data interface at different times; the optimization control module runs two types of models, namely, a light distribution prediction model and a support dynamic response model. The optimal tracking angle is obtained by analyzing through the light distribution prediction model, the support adjustment amount is calculated and generated to control the angle adjustment mechanism to adjust the orientation of the photovoltaic panel, and the actual power generation efficiency feedback value is calculated through the support dynamic response model to complete the support optimization.

[0072] It can be understood that the various modules described in the artificial intelligence-based optimized system for a tracking type photovoltaic support correspond to the respective steps in the artificial intelligence-based optimized method for a tracking type photovoltaic support described in the reference Figure 1 description. Therefore, the operations, features, and beneficial effects described above for the artificial intelligence-based optimized method for a tracking type photovoltaic support also apply to the artificial intelligence-based optimized system for a tracking type photovoltaic support and the modules included therein, and will not be elaborated herein.

[0073] Next, referring to Figure 3 , which shows a schematic structural diagram of an electronic device 300 suitable for implementing some embodiments of the present invention. The electronic device in some embodiments of the present invention may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Tablet Computers), PMPs (Portable Multimedia Players), in-vehicle terminals (such as in-vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 3 The terminal device shown is only an example and should not bring any limitation to the functions and usage scopes of the embodiments of the present invention.

[0074] As Figure 3As shown, the electronic device 300 may include a processing device (such as a central processing unit, a graphics processing unit, etc.) 301, which may perform various appropriate actions and processes according to a program stored in the read-only memory (ROM) 302 or a program loaded from the storage device 308 into the random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the electronic device 300 are also stored. The processing device 301, the ROM 302, and the RAM 303 are connected to each other through a bus 304. The input / output (I / O) interface 305 is also connected to the bus 304.

[0075] Generally, the following devices may be connected to the I / O interface 305: an input device 306 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 307 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 308 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 309. The communication device 309 may allow the electronic device 300 to communicate with other devices wirelessly or wiredly to exchange data. Although Figure 3 the electronic device 300 with various devices is shown, it should be understood that it is not required to implement or have all the shown devices. Instead, more or fewer devices may be implemented or had. Figure 3 Each block shown in may represent a device or, as needed, multiple devices.

[0076] Furthermore, the storage medium of the embodiments of the present application stores program instructions capable of implementing all the above methods. Among them, the program instructions may be stored in the above storage medium in the form of a software product, including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in the various embodiments of the present application. And the foregoing storage medium includes: various media capable of storing program codes such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk, or an optical disc, or a terminal device such as a computer, a server, a mobile phone, a tablet, etc.

[0077] The above description is only some preferred embodiments of the present invention and an explanation of the technical principles applied. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present invention is not limited to the technical solutions formed by the specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the technical solutions formed by mutually replacing the above features with the technical features (but not limited to) having similar functions disclosed in the embodiments of the present invention.

Claims

1. An optimization method for a tracking type photovoltaic support based on artificial intelligence, characterized in that, Including: The photovoltaic sensor array collects environmental light data in real time, inputs the environmental light data into the light distribution prediction model to generate predicted light distribution information, inputs the bracket state parameters into the bracket dynamic response model to generate bracket response characteristics, and uses the optimization control module to execute two types of model operations at different times; The tracking type photovoltaic bracket is equipped with two types of components, namely an angle adjustment mechanism and a data acquisition unit. Among them, the data acquisition unit receives the predicted light distribution information, and the tracking type photovoltaic bracket analyzes the predicted light distribution information to obtain the optimal tracking angle; The tracking type photovoltaic bracket calculates the bracket adjustment amount according to the current position parameters and the optimal tracking angle; The tracking type photovoltaic bracket drives the angle adjustment mechanism to adjust the orientation of the photovoltaic panel according to the bracket adjustment amount; The tracking type photovoltaic bracket monitors the actual power generation efficiency in real time through the data acquisition unit, performs feedback correction on the actual power generation efficiency, generates update parameters required for the tracking type photovoltaic bracket to execute the next cycle of optimization, and completes the bracket optimization process.

2. The optimized method for a tracking type photovoltaic bracket based on artificial intelligence according to claim 1, wherein, The predicted light distribution information is numerical data with high spatio-temporal resolution containing environmental light intensity, incident angle and time series characteristics.

3. The optimized method for a tracking type photovoltaic bracket based on artificial intelligence according to claim 1, wherein The bracket state parameters are structured data with dynamic change characteristics containing the current angle of the photovoltaic panel, mechanical stress and historical adjustment records, and do not contain the predicted light distribution information.

4. The optimized method for a tracking type photovoltaic bracket based on artificial intelligence according to claim 1, characterized in that, Inputting the environmental light data into the light distribution prediction model to generate predicted light distribution information includes: Performing normalization processing on the environmental light data to obtain a standardized light sequence; Performing time series feature extraction on the standardized light sequence to obtain a featureized light sequence; performing noise suppression on the featureized light sequence to obtain a denoised light sequence; Adding a timestamp label and a spatial position label to the denoised light sequence to obtain an annotated light sequence; Inputting the annotated light sequence into the light distribution prediction model for prediction to obtain predicted light distribution information; Performing confidence evaluation on the predicted light distribution information to obtain an optimized weight coefficient; Generating the final predicted light distribution information according to the optimized weight coefficient.

5. The optimized method for a tracking type photovoltaic bracket based on artificial intelligence according to claim 4, characterized in that, Performing normalization processing on the environmental light data is achieved through the following process: Convert the ambient light data into an input sequence , calculate the normalization parameters and , where is the sequence mean, is the sequence standard deviation, and output the normalized sequence satisfies the relationship: ； ； Among them, is the length of the input sequence; represents the th ambient light data in the input sequence; represents the th data in the normalized sequence; is the average value of the input sequence through the formula ； Calculating; is the standard deviation of the input sequence through the formula ； Calculating.

6. The optimized method for a tracking type photovoltaic support based on artificial intelligence according to claim 4, characterized in that, Adding a timestamp label and a spatial position label to the denoised light sequence to obtain an annotated light sequence, where the timestamp label provides time series correlation information for the model, the spatial position label provides regional distribution information for the model, and the denoised light sequence carries the original environmental light data.

7. A method for optimizing a tracking type photovoltaic support based on artificial intelligence according to claim 1, characterized in that Analyzing the predicted light distribution information includes: Performing feature mapping on the predicted light distribution information to obtain a multi-dimensional feature vector; Performing gradient backpropagation on the multi-dimensional feature vector to update the parameters of the light distribution prediction model; Performing light intensity interpolation calculation on the updated light distribution prediction model to obtain an interpolated light distribution; Performing spatial clustering analysis on the interpolated light distribution to obtain regional light hotspots; Performing angle projection transformation on the regional light hotspots to obtain a set of candidate tracking angles; Performing power generation efficiency simulation on the set of candidate tracking angles to obtain the optimal tracking angle; Input the optimal tracking angle into the angle adjustment mechanism control instruction generation module.

8. The optimized method for a tracking type photovoltaic support based on artificial intelligence according to claim 7, wherein The gradient backpropagation is achieved in the following way: for the prediction error , with the learning rate as the update step size and the weight matrix as the model parameter, update the parameters of the th layer of the light distribution prediction model to satisfy the relationship: ； in, is the number of iterations; Indicates Layer The weight matrix at the iteration; Indicates Layer The weight matrix after the iteration update; is the learning rate, which is used to control the step size of each parameter update; Error For Layer The partial derivative of the weight matrix at the iteration, the updating process continues until the prediction error Less than the set threshold.

9. The optimization method of a tracking type photovoltaic bracket based on artificial intelligence according to claim 8, characterized in that Based on the model parameters after the completion of the gradient backpropagation, within the subsequent time window with a length of , perform multi-step rolling optimization of the predicted information of the light distribution, that is, at each time step , predict the future light distribution for the next p steps according to the historical data , where the rolling optimization objective function is: ​ ； Among them, is the time window length; is the length of historical data for prediction; is the current time step; is the prediction step length; is the actual light distribution data at the step in the future; is the predicted light distribution data at the step in the future. The prediction effect of the light distribution prediction model is optimized by minimizing this objective function.

10. A tracking bracket optimization system for a photovoltaic power generation system, characterized in that, It includes a photovoltaic sensor array, an angle adjustment mechanism, and an optimization control module. Among them, the photovoltaic sensor array generates two types of data, namely, ambient light data collected in real time and photovoltaic panel state parameters, which are transmitted to the optimization control module through the same data interface at different times. The optimization control module runs two types of models, namely, a light distribution prediction model and a support dynamic response model. The optimal tracking angle is obtained by analyzing through the light distribution prediction model, the support adjustment amount is calculated and generated to control the angle adjustment mechanism to adjust the orientation of the photovoltaic panel. The actual power generation efficiency feedback value is calculated through the support dynamic response model to complete the support optimization.

Images