Three-dimensional visual management system of photovoltaic energy storage system

Abstract

The invention discloses a three-dimensional visual management system of a photovoltaic energy storage system, which relates to the technical field of data encryption, and comprises the following steps: collecting environmental data in real time, preprocessing the collected environmental data, transmitting the preprocessed environmental data to a central processing unit, and loading a three-dimensional visual model at the same time; a machine learning algorithm is adopted to analyze the environmental data in the central processing unit, establish a prediction model, predict the optimal angle of the photovoltaic panel and generate a performance evaluation report; according to the performance evaluation report, a photovoltaic panel adjustment instruction is generated, the photovoltaic panel is automatically adjusted to the optimal angle, and the latest photovoltaic panel state is reflected in the three-dimensional visualization model; detecting the state of the photovoltaic panel according to the newest state of the photovoltaic panel in the three-dimensional view, positioning a potential problem area, and generating an optimization suggestion; the system can automatically generate a photovoltaic panel adjustment instruction according to a performance evaluation result, realizes automatic angle adjustment, and reflects the latest photovoltaic panel state in real time in a three-dimensional visual interface.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic energy storage, and in particular to a three-dimensional visualization management system for a photovoltaic energy storage system. Background Art

[0002] In recent years, with the continuous growth of the global demand for clean energy and the progress of photovoltaic technology, photovoltaic energy storage systems have been widely used in the energy field. Traditional photovoltaic energy storage systems mainly rely on simple sensors and basic control algorithms to monitor and adjust the working state of photovoltaic panels to optimize energy output. However, with the development of smart grid technology and the Internet of Things (IoT), the management of photovoltaic energy storage systems is gradually moving towards intelligence and high efficiency. As a new technical means, the three-dimensional visualization management system has gradually become an important tool for improving the management efficiency of photovoltaic energy storage systems. Early three-dimensional visualization management systems were mainly used to display static system structures and layouts, while recent research has focused on expanding them into a comprehensive platform integrating data collection, analysis, prediction, and fault diagnosis. These systems can not only provide intuitive visual displays but also optimize the angle adjustment of photovoltaic panels through data analysis, thereby improving energy conversion efficiency.

[0003] Although existing three-dimensional visualization management systems have made significant progress in some aspects, there are still many deficiencies in practical applications. First, existing systems usually rely on fixed time intervals for data collection and update, resulting in the inability to achieve true real-time monitoring. This may lead to a lag in the angle adjustment of photovoltaic panels under rapidly changing weather conditions (such as sudden cloud cover or wind speed changes), thereby affecting power generation efficiency. Second, existing data analysis methods are mostly based on simple statistical models or empirical formulas, lacking comprehensive consideration of complex environmental variables (such as temperature, humidity, light intensity, etc.), resulting in limited prediction accuracy. In addition, existing systems often neglect the monitoring and evaluation of the health status of equipment, making it difficult to detect potential problems in a timely manner and propose effective maintenance suggestions. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a three-dimensional visualization management system for a photovoltaic energy storage system to solve the problems of lagging real-time monitoring, low prediction accuracy, and insufficient equipment health monitoring in the prior art.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a three-dimensional visualization management system for a photovoltaic energy storage system, which includes collecting environmental data in real time, preprocessing the collected environmental data, transmitting it into a central processing unit, and simultaneously loading a three-dimensional visualization model; using a machine learning algorithm to analyze the environmental data in the central processing unit, establishing a prediction model to predict the optimal angle of the photovoltaic panels, and generating a performance evaluation report; generating a photovoltaic panel adjustment instruction according to the performance evaluation report, automatically adjusting the photovoltaic panels to the optimal angle, and reflecting the latest state of the photovoltaic panels in a three-dimensional visualization interface; detecting the state of the photovoltaic panels according to the latest state of the photovoltaic panels in the three-dimensional visualization view, locating potential problem areas, and generating optimization suggestions.

[0008] As a preferred solution of the three-dimensional visualization management system for the photovoltaic energy storage system of the present invention, wherein: the environmental data includes the degradation index of the photovoltaic panel material obtained by a multispectral imaging sensor, the dangerous variable data captured by a millimeter-wave radar, the surface temperature, current, and voltage of the photovoltaic panel.

[0009] As a preferred solution of the three-dimensional visualization management system for the photovoltaic energy storage system of the present invention, wherein: the preprocessing of the collected environmental data is specifically as follows

[0010] Using a generative adversarial network to perform context-aware filling of the missing values in the environmental data, and simultaneously checking and deleting duplicate environmental data records;

[0011] Performing a moving average process on the filled environmental data, and performing wavelet transform decomposition on the data after the moving average process;

[0012] Scaling the decomposed environmental data and converting it into environmental data of the same scale;

[0013] Downsampling the high-frequency environmental data and introducing a dynamic time warping algorithm to align different frequency data streams;

[0014] Converting the environmental data that has been converted into the same scale and downsampled into JSON format and transmitting it to the central processing unit.

[0015] As a preferred solution of the three-dimensional visualization management system for the photovoltaic energy storage system of the present invention, wherein: the loading of the three-dimensional visualization model is specifically as follows

[0016] Install an augmented reality application on the collection device, and use the camera function of the AR application to scan the identification mark of the photovoltaic energy storage system;

[0017] After successful scanning, the application requests to obtain the three-dimensional model file of the photovoltaic energy storage system, and uses the simultaneous localization and mapping (SLAM) technology to correct the three-dimensional model spatial coordinates in real time through LiDAR point cloud matching, and prepares to load the three-dimensional visualization model;

[0018] Load the completed 3D visualization model by obtaining the 3D model file of the photovoltaic energy storage system and the WebSocket real-time communication protocol from the central processing unit.

[0019] Use the edge computing node to perform model lightweight processing and separate the rendering of texture mapping and skeletal animation.

[0020] As a preferred solution of the 3D visualization management system of the photovoltaic energy storage system described in the present invention, wherein: the machine learning algorithm is used to analyze the environmental data in the central processing unit and establish a prediction model. The specific steps are as follows.

[0021] Construct a bidirectional long short-term memory network as a time feature extractor and build a cascaded hybrid model with a support vector machine.

[0022] Calculate the statistical quantities of the average temperature, photovoltaic panel current, voltage value, and maximum light intensity of the environmental data in the central processing unit, and add time features.

[0023] Divide the analyzed environmental data set into a training set and a test set, use the training set data to train the prediction model, optimize the prediction model parameters, and adjust the regularization parameters through cross-validation to obtain a trained prediction model.

[0024] Adopt the knowledge distillation method to compress the model scale to the level that can be run on embedded devices.

[0025] As a preferred solution of the 3D visualization management system of the photovoltaic energy storage system described in the present invention, wherein: predict the best photovoltaic panel angle and generate a performance evaluation report. The specific steps are as follows.

[0026] Input the real-time collected environmental data into the trained prediction model, and calculate the predicted best photovoltaic panel angle through a linear kernel function.

[0027] Deploy a reinforcement learning agent to simulate different angle adjustment strategies in a virtual environment, and optimize the predicted best photovoltaic panel angle result by combining with the Q-learning algorithm.

[0028] Use the mean square error as an evaluation index, calculate the error of the prediction model on the test set, and generate a performance evaluation report according to the error.

[0029] As a preferred solution of the 3D visualization management system of the photovoltaic energy storage system described in the present invention, wherein: generate a photovoltaic panel adjustment instruction to automatically adjust the photovoltaic panel to the best angle. The specific steps are as follows.

[0030] Obtain the predicted best photovoltaic panel angle and the mean square error from the performance evaluation report.

[0031] Construct an inverse kinematic model of the robotic arm and use quintic polynomial interpolation to generate a smooth rotation trajectory;

[0032] Based on the central processing unit, read the actual angle of the current photovoltaic panel, calculate the angle difference with the predicted optimal angle, determine the adjustment direction, the magnitude of the rotation angle, and the smooth rotation trajectory, and generate a control signal;

[0033] Send the generated control signal to the motor and transmission mechanism through the central processing unit, and drive the photovoltaic panel to automatically adjust to the optimal angle;

[0034] Introduce a strain sensor feedback to establish a closed-loop control system, compensate for mechanical transmission errors in real time, and generate a torque compensation coefficient matrix.

[0035] As a preferred solution of the three-dimensional visualization management system of the photovoltaic energy storage system described in the present invention, wherein: reflecting the latest state of the photovoltaic panel in the three-dimensional visualization model, the specific steps are as follows.

[0036] Extract the current angle of the photovoltaic panel, the surface temperature distribution, and the light intensity from the central control unit;

[0037] Load a three-dimensional model containing the photovoltaic panel and the photovoltaic panel support structure in the three-dimensional visualization model, and initialize the three-dimensional visualization model according to the configuration file of the photovoltaic energy storage system;

[0038] According to the angle change amount between the current angle of the photovoltaic panel and the initial angle, perform a rotation operation on the three-dimensional visualization model through a rotation matrix to form the latest angle of the photovoltaic panel;

[0039] Add floating labels of the latest photovoltaic panel angle to the initialized three-dimensional visualization model to establish a new three-dimensional visualization model, and input the surface temperature, output voltage, and output current into the thermal analysis diagram, bar chart, and streamline diagram in the new three-dimensional visualization model.

[0040] As a preferred solution of the three-dimensional visualization management system of the photovoltaic energy storage system described in the present invention, wherein: detecting the state of the photovoltaic panel and locating potential problem areas, the specific steps are as follows.

[0041] According to the historical surface temperature of the photovoltaic panel, the historical output voltage of the photovoltaic panel, and the historical output current in the central processing unit, set the upper and lower limit thresholds of the surface temperature, output voltage, and output current and apply them to the thermal analysis diagram, bar chart, and streamline diagram in the new three-dimensional visualization model;

[0042] Deploy a graph neural network in the three-dimensional visualization model, construct a component fault propagation model, locate the root fault node through topological analysis, and generate a causal inference tree according to the fault node.

[0043] As a preferred solution of the three-dimensional visualization management system of the photovoltaic energy storage system described in the present invention, wherein: the generation of optimization suggestions is specifically carried out as follows.

[0044] Retrieve the temperature, voltage, and current data and historical fault logs before the photovoltaic panel alarm is triggered from the central processing unit, and start the trained prediction model. Input the temperature, voltage, current before the alarm is triggered, and the historical temperature, historical voltage, and historical current data in the historical fault logs to obtain a quantitative prediction of the temperature distribution, power generation, and current fluctuation of the photovoltaic panel and a risk level assessment.

[0045] Based on the quantitative prediction and risk level assessment, put forward targeted suggestions and generate optimization parameters.

[0046] Through the simulation function in the three-dimensional visualization model, simulate according to the simulation parameters and the collected environmental data of temperature, current, and voltage to obtain the optimized scenario simulation result.

[0047] Compare and analyze the optimized scenario simulation result with the temperature, voltage, and current data before the alarm is triggered, evaluate the actual effect of the optimization measures, and formulate a preventive maintenance plan according to the area of the actual potential problems.

[0048] The beneficial effects of the present invention are as follows: By collecting and preprocessing the environmental data of the photovoltaic energy storage system in real time, transmitting it to the central processing unit and loading the three-dimensional visualization model, the accurate monitoring of the system operating environment is realized, and the accuracy of the prediction model is improved; Machine learning algorithms are used to analyze the data, a prediction model is established to determine the optimal angle of the photovoltaic panel, and a performance evaluation report is generated to accurately predict the future performance of the system, improving the energy conversion efficiency; According to the evaluation report, adjustment instructions are automatically generated, the photovoltaic panel is automatically adjusted to the optimal angle and the status is updated in the three-dimensional model, reducing the need for manual intervention, realizing automation and real-time status update; Based on the latest status in the three-dimensional visualization model, the health status of the photovoltaic panel is detected, potential problem areas are located and optimization suggestions are generated, continuously monitoring the system health and providing optimization guidance, preventing major failures from occurring, and reducing the maintenance cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0050] Figure 1 It is a flowchart of the three-dimensional visualization management system of the photovoltaic energy storage system in Embodiment 1.

[0051] Figure 2 It is a system diagram of the 3D visualization management system for the photovoltaic energy storage system in Embodiment 1. Specific Embodiments

[0052] To make the above objects, features, and advantages of the present invention more obvious and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings of the specification.

[0053] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0054] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation manner of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or selectively exclusive embodiments from other embodiments.

[0055] Embodiment 1, referring to Figure 1 and Figure 2 is the first embodiment of the present invention. This embodiment provides a 3D visualization management system for a photovoltaic energy storage system, including the following steps:

[0056] Collect environmental data in real time, preprocess the collected environmental data, transmit it into the central processing unit, and load the 3D visualization model at the same time.

[0057] The environmental data includes the degradation index of the photovoltaic panel material obtained by a multispectral imaging sensor, the dangerous variable data captured by a millimeter-wave radar, the surface temperature, current, and voltage of the photovoltaic panel; the missing values of the environmental data are filled by dynamic interpolation using the quantum genetic algorithm, and at the same time, duplicate environmental data records are checked and deleted.

[0058] Dynamic interpolation using the quantum genetic algorithm refers to a data processing method that combines the parallel search ability of quantum computing and the optimization mechanism of the genetic algorithm, and adaptively adjusts the interpolation parameters in real time according to the data characteristics to fill the missing values in the optimal way.

[0059] Based on all received data packets, check whether each data packet has a corresponding measurement value. If the environmental data at a certain timestamp is empty, mark it as a missing value, and use the adjacent non-empty values for linear interpolation calculation to obtain the filled data.

[0060] Perform a moving average process on the filled environmental data, and perform wavelet transform decomposition on the data after the moving average process.

[0061] Calculate the Z-Score for each data record after data cleaning. If the calculated value |Z| > 3, then the data point is considered an outlier. For each identified outlier, replace it with the median within its sliding window. The size of the sliding window is set according to the periodicity and trend of the data, application scenario requirements, and specific conditions such as computing resources and efficiency. For example, select ±5 time steps as the window size.

[0062] Wavelet transform decomposition is a signal processing technique used to decompose data into components of different frequencies.

[0063] Based on the characteristics of the photovoltaic data (such as noise type, signal smoothness), select a suitable wavelet basis (such as Daubechies wavelet) to match the time-frequency characteristics of the data; perform multi-level wavelet decomposition on the environmental data after moving average processing, and decompose the signal layer by layer into low-frequency approximation components (Approximation Coefficients, reflecting the overall trend) and high-frequency detail components (Detail Coefficients, containing noise and short-term fluctuations); perform inverse wavelet transform on the processed low-frequency approximation components and the denoised high-frequency detail components to synthesize the denoised environmental data, removing high-frequency interference while retaining valid information.

[0064] Scale the decomposed environmental data and convert it into environmental data of the same scale.

[0065] Based on the application requirements and the requirements of subsequent analysis, determine the target scale for all data conversions. For example, if the target is daily light intensity analysis, then all light intensity data is adjusted to the scale of the daily average value.

[0066] Adopt a standardization method to adjust environmental data of different scales, scale it, and convert it into the target size.

[0067] Downsample the high-frequency environmental data and introduce the dynamic time warping algorithm to align different frequency data streams.

[0068] Select an appropriate downsampling method according to the characteristics of the collected data and the application scenario. For example, use other statistics such as the mean, maximum value, or minimum value as the result of downsampling, and calculate the corresponding statistical indicators for the downsampled data.

[0069] Convert the environmental data that has been converted to the same scale and downsampled into JSON format and transmit it to the central processing unit.

[0070] Integrate the data converted to the same size, the downsampled environmental data, and the calculated statistical metrics into a JSON structure, and determine the key data such as timestamps, sensor IDs, eigenvalue, and their statistical metrics in each record. Then, use the HTTP / HTTPS protocol to send the data packet to the central processing unit through the RESTful API (Representational State Transfer Application Programming Interface) interface.

[0071] Install an augmented reality application on the collection device. Use the camera function of the AR application to scan the identification mark of the photovoltaic energy storage system. After successful scanning, the application requests to obtain the 3D model file of the photovoltaic energy storage system. Apply the Simultaneous Localization and Mapping (SLAM) technology to correct the 3D model spatial coordinates in real time through LiDAR point cloud matching, prepare to load the 3D visualization model, and load the 3D visualization model by obtaining the 3D model file of the photovoltaic energy storage system from the central processing unit and the WebSocket real-time communication protocol.

[0072] The application initializes a WebSocket connection to the data service endpoint of the central processing unit. After successful connection, it can subscribe to the data stream of interest and load the obtained 3D model file into the AR scene to complete the loading of the 3D visualization model.

[0073] Use an edge computing node to perform model lightweight processing and separate the rendering of texture mapping and skeletal animation.

[0074] Texture mapping is the process of applying a 2D image (texture) to the surface of a 3D model to increase details; skeletal animation is to drive the actions of the model through a skeletal system, such as the adjustment actions of photovoltaic panels.

[0075] Use a mesh simplification algorithm to reduce the number of polygons of the photovoltaic panel model and convert the texture of the resolution to an adaptive scalable texture format; split the 3D model into three independent data packets: geometric skeleton, texture atlas, and animation data, and preload the geometric skeleton and animation data at the edge computing node, and asynchronously load the texture atlas on demand; use deferred rendering technology to dynamically map the texture and separate the animation update through a double-buffer mechanism; train a neural network in the cloud to evaluate it, and automatically switch to a hierarchical rendering mode when the computing power of the edge node is insufficient.

[0076] Adopt machine learning algorithms to analyze the environmental data in the central processing unit, establish a prediction model, predict the optimal angle of the photovoltaic panel, and generate a performance evaluation report.

[0077] Construct a bidirectional long short-term memory network as a time feature extractor and build a cascaded hybrid model with a support vector machine.

[0078] The steps to construct a cascaded hybrid model of Bi-LSTM (bidirectional long short-term memory network) and SVM are as follows:

[0079] Preprocess time series data (such as temperature and current); design a bidirectional LSTM network to extract time series context features through forward and backward layers, and add a Dropout layer to prevent overfitting; extract the concatenated hidden states at the end of the LSTM as high-dimensional features, and after dimensionality reduction by normalization or principal component analysis, input them into the SVM. The SVM selects a linear or RBF kernel function according to the task type, optimizes the hyperparameters through grid search, and completes classification or regression prediction. The model adopts a phased training strategy, first freezes the Bi-LSTM to train the SVM, then unfreezes and fine-tunes the overall network, and combines L2 regularization and early stopping to control overfitting. This cascaded model combines the time series modeling ability of Bi-LSTM with the discriminative generalization advantage of SVM, and is suitable for scenarios such as photovoltaic angle prediction that require both long-term dependence and non-linear decision-making. The measured results can reduce the prediction error by 15%.

[0080] Calculate the average temperature, average photovoltaic panel current, average voltage value, and the statistic of the maximum light intensity of the environmental data in the central processing unit, and add time features.

[0081] According to the summation formula, calculate the average temperature, average photovoltaic panel current, average voltage value, and the statistic of the maximum light intensity of the environmental data in the central processing unit, and add them to the statistic according to the time when the environmental data is collected.

[0082] Divide the analyzed environmental data set into a training set and a test set, use the training set data to train the prediction model, optimize the prediction model parameters, and adjust the regularization parameters through cross-validation to obtain a trained prediction model.

[0083] Add the number of hours in a day as a feature to the data set, and randomly allocate the data according to a certain ratio (such as 7:3 or 8:2) to the data set for model training (training set) and the data set for evaluating the generalization ability of the model (test set);

[0084] Select the LinearSVR class (linear support vector regression class) in the scikit-learn library (machine learning library), and input the training set and the best angle of the trained photovoltaic panel into the model for training.

[0085] Set a reasonable search range, such as 0.01, 0.1, 1, 10, 100. According to GridSearchCV (grid search cross-validation), automatically perform cross-validation within the set search range, evaluate the performance of each test set and training set in the data set, and select the combination of the test set and training set with the highest average validation score.

[0086] Adopt the knowledge distillation method to compress the model size to the level that can be run on embedded devices.

[0087] Knowledge distillation is a technique for model compression by transferring the knowledge of a large model to a small model.

[0088] Input the environmental data collected in real time into the trained prediction model, and calculate the optimal photovoltaic panel angle through a linear kernel function.

[0089] Simultaneously learn the optimal photovoltaic panel angles of the prediction set and the training set based on the input features (photovoltaic panel temperature, current, voltage, light intensity, and time features), the feature matrix, and the target vector, using the prediction value with the highest average validation score and the training set.

[0090] Deploy a reinforcement learning agent to simulate different angle adjustment strategies in a virtual environment, and optimize the result of the optimal photovoltaic panel angle prediction by combining the Q-learning algorithm;

[0091] The Q-learning algorithm explores actions (random adjustment) and exploits known optimal actions through the ε-greedy strategy.

[0092] By constructing a virtual simulation environment for the photovoltaic system, deploy a reinforcement learning agent to dynamically explore angle adjustment actions based on the Q-learning algorithm, optimize the Q-value table with the goal of maximizing power generation efficiency, and finally output a lightweight optimal strategy and deploy it to an embedded controller to achieve autonomous and efficient adjustment of the photovoltaic panel angle.

[0093] Use the mean square error as an evaluation metric, calculate the error of the prediction model on the test set, and generate a performance evaluation report based on the error.

[0094] Evaluate the performance of the 3D visualization model by calculating the mean absolute error (MAE) and the mean square error (MSE). The relevant formulas are as follows:

[0095]

[0096] where, y i is the actual value of the i-th sample, is the predicted value of the i-th sample, and n is the total number of samples.

[0097] Based on the difference between the calculated MSE and MAE results, draw a difference graph and analyze the periodic changes and distribution in the graph: if the periodic changes are randomly distributed, it indicates that the model performs well; if there is an obvious trend, there are deficiencies; if the difference significantly deviates from the normal distribution, it means there is a problem with the model assumption;

[0098] Based on the conclusions drawn from the MSE and MAE, prepare a detailed performance evaluation report covering data analysis, an overview of model performance, and improvement suggestions, etc., and save it in the central processing unit in JSON format;

[0099] Generate a photovoltaic panel adjustment instruction according to the performance evaluation report, automatically adjust the photovoltaic panel to the optimal angle, and reflect the latest state of the photovoltaic panel in the 3D visualization model;

[0100] Obtain the predicted optimal photovoltaic panel angle and mean square error from the performance evaluation report; construct an inverse kinematic model of the robotic arm, and use a fifth-order polynomial interpolation to generate a smooth rotation trajectory; based on the central processing unit, read the actual angle of the current photovoltaic panel, calculate the angle difference with the predicted optimal angle, determine the adjustment direction and the magnitude of the rotation angle, and generate a control signal; send the generated control signal to the motor and transmission mechanism through the central processing unit, and drive the photovoltaic panel to automatically adjust to the optimal angle.

[0101] The steps to generate the control signal are as follows:

[0102] Obtain the current timestamp from the central processing unit, find the corresponding optimal photovoltaic panel angle θ in the optimal photovoltaic panel angle recommendation, and construct a standard instruction format that includes the photovoltaic panel ID and the optimal photovoltaic panel angle. The format is as follows:

[0103] {"panel_id":"001","target_angle":35}

[0104] Among them, "panel_id":"001" represents the photovoltaic panel ID, and "target_angle":35 represents the optimal photovoltaic panel angle.

[0105] The steps to drive the photovoltaic panel to automatically adjust to the optimal angle are as follows:

[0106] Transmit the latest photovoltaic panel angle to the 3D visualization interface through the API interface. In the 3D visualization scene, use the rotation matrix to update the position of the photovoltaic panel angle to accurately reflect the actual angle change. The rotation matrix formula is as follows:

[0107]

[0108] Among them, R(θ) represents a rotation matrix that depends on the angle θ; θ represents the rotation angle; cosθ and sinθ represent the cosine and sine values of the angle θ respectively.

[0109] Introduce strain sensor feedback to establish a closed-loop control system, and compensate for mechanical transmission errors in real time and generate a torque compensation coefficient matrix.

[0110] Real-time monitor the deformation data of the mechanical transmission chain through the strain sensor, feedback it to the closed-loop control system, dynamically calculate the torque compensation coefficient matrix of each joint, offset the transmission errors caused by gear clearance, shaft bending and friction, and achieve sub-millimeter positioning accuracy for photovoltaic panel angle adjustment.

[0111] Extract the current PV panel angle, surface temperature distribution, and light intensity from the central control unit; load a 3D model containing the PV panel and its support structure in the 3D visualization model, and initialize the 3D visualization model according to the PV energy storage system configuration file; perform a rotation operation on the 3D visualization model through a rotation matrix based on the angle change between the current PV panel angle and the initial angle to form the latest PV panel angle; add floating labels of the latest PV panel angle in the initialized 3D visualization model to establish a new 3D visualization model, and input the surface temperature, output voltage, and output current into the thermal analysis chart, bar chart, and streamline chart in the new 3D visualization model.

[0112] The specific steps to initialize the 3D visualization model are as follows:

[0113] Create or read the configuration file of the energy storage system, including the basic information of the PV panel, the address of the sensor data interface, etc.;

[0114] Use 3D modeling software (such as Blender) or directly create the basic model of the PV panel in the 3D visualization engine, adjust the model size according to the dimension information in the configuration file, and set the position and initial angle of each PV panel object to the values defined in the configuration file to ensure that all model elements are correctly loaded and displayed in the scene.

[0115] Detect the PV panel status according to the latest status of the PV panel in the 3D view, locate potential problem areas, and generate optimization suggestions.

[0116] Set the upper and lower limit thresholds of the surface temperature, output voltage, and output current according to the historical PV panel surface temperature, historical PV panel output voltage, and historical PV panel output current in the central processing unit and apply them to the thermal analysis chart, bar chart, and streamline chart in the new 3D visualization model.

[0117] Extract the historical surface temperature, output voltage, and output current data of the PV panel from the central processing unit, set the upper and lower limit thresholds within the normal operation range using the 3σ principle. For example, for the surface temperature, calculate its average value μ T and standard deviation σ T , set the upper limit as μ T +3σ T , set the lower limit as μ T -3σ T , and similarly, process the output voltage V and current I in the same way to obtain the upper and lower limit thresholds of the grid, and add threshold lines to the thermal analysis chart, bar chart, and streamline chart in the new 3D visualization model.

[0118] Based on the real-time data points of surface temperature, output voltage, and output current shown in the thermal analysis diagram, bar chart, and streamline diagram, when the real-time data points exceed the upper and lower limit thresholds, an alarm is immediately triggered and the detailed time is recorded in the historical fault log.

[0119] Deploy a graph neural network in the 3D visualization model, construct a component fault propagation model, locate the root fault node through topological analysis, and generate a causal inference tree based on the fault node.

[0120] A graph neural network (GNN) refers to a deep learning model that takes graph-structured data (e.g., nodes = components, edges = connection relationships) as input and learns the fault propagation pattern between nodes through a message passing mechanism.

[0121] The steps to construct a component fault propagation model are as follows.

[0122] Take each component in the photovoltaic system as a node in the graph and assign attributes such as model number and location coordinates; use a graph attention network to capture the shutdown press propagation path through the multi-head attention mechanism, simulate random node failures, and generate propagation path labels; jointly optimize node classification and edge propagation prediction, and use curriculum learning to gradually increase the complexity of the fault propagation path.

[0123] A causal inference tree takes the root node as the root and constructs a tree structure according to the fault propagation time sequence and probability weight to visually display the fault causal relationship chain.

[0124] Retrieve the temperature, voltage, and current data and historical fault log before the photovoltaic panel alarm is triggered from the central processing unit, and start the trained prediction model. Input the temperature, voltage, current before the alarm is triggered, and the historical temperature, historical voltage, and historical current data in the historical fault log to obtain a quantitative prediction of the temperature distribution, power generation, and current fluctuation of the photovoltaic panel and a risk level assessment.

[0125] Quantitative prediction refers to using the trained model to predict the future performance parameters of the photovoltaic panel, such as the specific values of temperature distribution, power generation, and current fluctuation, based on the temperature, voltage, current before the alarm is triggered, and the historical temperature, historical voltage, and historical current data in the historical fault log.

[0126] Risk level assessment is to classify the risks that the photovoltaic panel may encounter based on the prediction results to identify potential problems of different severity levels.

[0127] Based on quantitative prediction and risk level assessment, targeted suggestions are put forward and optimization parameters are generated. Through the simulation function in the 3D visualization model, simulations are carried out according to the simulation parameters and the collected environmental data of temperature, current, and voltage, and the optimized scenario simulation results are obtained. The optimized scenario simulation results are compared and analyzed with the temperature, voltage, and current data before the alarm is triggered to evaluate the actual effect of the optimization measures, and a preventive maintenance plan is formulated according to the actual potential problem areas.

[0128] Extract the temperature, voltage, and current data at the same time before and after the simulation, calculate the differences of each index, and use charts to show the change trends before and after the optimization.

[0129] According to the results of the comparative analysis, identify the areas with high risks after the simulation, and formulate maintenance strategies, such as regular inspections, enhanced cooling measures, adjusting the angle of the photovoltaic panels, etc.

[0130] In summary, the present invention realizes precise monitoring of the system operating environment and improves the accuracy of the prediction model by: collecting and preprocessing the environmental data of the photovoltaic energy storage system in real time, transmitting it to the central processing unit and loading the 3D visualization model; using machine learning algorithms to analyze the data, establishing a prediction model to determine the optimal angle of the photovoltaic panels, and generating a performance evaluation report to accurately predict the future performance of the system, thereby improving the energy conversion efficiency; automatically generating adjustment instructions according to the evaluation report, automatically adjusting the photovoltaic panels to the optimal angle and updating the status in the 3D model, reducing the need for manual intervention, and realizing automation and real-time status update; based on the latest status in the 3D visualization model, detecting the health status of the photovoltaic panels, locating potential problem areas and generating optimization suggestions, continuously monitoring the system health and providing optimization guidance, preventing major failures from occurring, and reducing the maintenance cost.

[0131] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A three-dimensional visualization management system for a photovoltaic energy storage system, characterized in that: including, a data acquisition module that collects environmental data in real time, pre - processes the collected environmental data, transmits it into the central processing unit, and simultaneously loads a 3D visualization model; a prediction and evaluation module that uses machine learning algorithms to analyze the environmental data in the central processing unit, establish a prediction model, predict the optimal photovoltaic panel angle, and generate a performance evaluation report; an adjustment and feedback module that generates a photovoltaic panel adjustment instruction according to the performance evaluation report, automatically adjusts the photovoltaic panel to the optimal angle, and reflects the latest photovoltaic panel state in the 3D visualization model; a monitoring and optimization module that detects the photovoltaic panel state, locates potential problem areas, and generates optimization suggestions according to the latest state of the photovoltaic panel in the 3D view.

2. The three-dimensional visualization management system of the photovoltaic energy storage system according to claim 1, wherein: The environmental data includes the surface temperature, current, voltage, and light intensity of the photovoltaic panel.

3. The three-dimensional visualization management system of the photovoltaic energy storage system according to claim 1, characterized in that: The pre - processing of the collected environmental data is as follows: Use a generative adversarial network to perform context - aware filling of missing values in the environmental data, and at the same time check and delete duplicate environmental data records; Perform a moving average process on the filled environmental data, and perform wavelet transform decomposition on the data after the moving average process; Scale the decomposed environmental data and convert it into environmental data of the same scale; Down - sample the high - frequency environmental data and introduce a dynamic time warping algorithm to align different - frequency data streams; Convert the environmental data that has been converted to the same scale and down - sampled into JSON format and transmit it to the central processing unit.

4. The three-dimensional visualization management system of the photovoltaic energy storage system according to claim 1, characterized in that: The loading of the 3D visualization model is as follows: Install an augmented reality application on the collection device, and use the camera function of the AR application to scan the identification mark of the photovoltaic energy storage system; After successful scanning, the application requests to obtain the 3D model file of the photovoltaic energy storage system, applies simultaneous localization and mapping technology, and real - time corrects the 3D model spatial coordinates through LiDAR point cloud matching to prepare for loading the 3D visualization model; Load the 3D visualization model by obtaining the 3D model file of the photovoltaic energy storage system from the central processing unit and the WebSocket real - time communication protocol; Use an edge computing node to perform model lightweight processing and separate rendering of texture mapping and skeletal animation.

5. The three-dimensional visualization management system of the photovoltaic energy storage system according to claim 1, characterized in that: The use of machine learning algorithms to analyze the environmental data in the central processing unit and establish a prediction model is as follows: Construct a bidirectional long short - term memory network as a time feature extractor and build a cascaded hybrid model with a support vector machine; Calculate the statistical quantities of the average temperature, photovoltaic panel current, voltage value, and maximum light intensity of the environmental data in the central processing unit, and add time features; Divide the analyzed environmental data set into a training set and a test set, use the training set data to train the prediction model, optimize the prediction model parameters, and adjust the regularization parameters through cross - validation to obtain a trained prediction model; Use the knowledge distillation method to compress the model scale to the level that can be run on an embedded device.

6. The three-dimensional visualization management system of the photovoltaic energy storage system according to claim 1, wherein: The prediction of the optimal photovoltaic panel angle and the generation of a performance evaluation report are as follows: Input the environmental data collected in real time into the trained prediction model, and calculate the predicted optimal photovoltaic panel angle through a linear kernel function; Deploy a reinforcement learning agent to simulate policy adjustments at different angles in a virtual environment, and combine the Q-learning algorithm to optimize the predicted optimal angle result of the photovoltaic panel; Use the mean square error as an evaluation index to calculate the error of the prediction model on the test set, and generate a performance evaluation report based on the error.

7. The three-dimensional visualization management system of the photovoltaic energy storage system according to claim 1, characterized in that: The steps to generate the photovoltaic panel adjustment instruction and automatically adjust the photovoltaic panel to the optimal angle are as follows: Obtain the predicted optimal angle of the photovoltaic panel and the mean square error from the performance evaluation report; Construct an inverse kinematic model of the robotic arm and use a fifth-degree polynomial interpolation to generate a smooth rotation trajectory; Based on the central processing unit, read the actual angle of the current photovoltaic panel, calculate the angle difference with the predicted optimal angle, determine the adjustment direction, the magnitude of the rotation angle, and the smooth rotation trajectory, and generate a control signal; Send the generated control signal to the motor and transmission mechanism through the central processing unit, and drive the photovoltaic panel to automatically adjust to the optimal angle; Introduce a strain sensor feedback to establish a closed-loop control system, compensate for mechanical transmission errors in real time, and generate a torque compensation coefficient matrix.

8. The three-dimensional visualization management system of the photovoltaic energy storage system according to claim 1, characterized in that: The steps to reflect the latest state of the photovoltaic panel in the 3D visualization model are as follows: Extract the current angle of the photovoltaic panel, the surface temperature distribution, and the light intensity from the central control unit; Load a 3D model containing the photovoltaic panel and the photovoltaic panel support structure in the 3D visualization model, and initialize the 3D visualization model according to the photovoltaic energy storage system configuration file; According to the angle change amount between the current angle of the photovoltaic panel and the initial angle, perform a rotation operation on the 3D visualization model through a rotation matrix to form the latest angle of the photovoltaic panel; Add floating labels of the latest angle of the photovoltaic panel to the initialized 3D visualization model to establish a new 3D visualization model, and input the surface temperature, output voltage, and output current into the thermal analysis chart, bar chart, and streamline chart in the new 3D visualization model.

9. The three-dimensional visualization management system of the photovoltaic energy storage system according to claim 1, wherein: The steps to detect the state of the photovoltaic panel and locate potential problem areas are as follows: Set the upper and lower limit thresholds of the surface temperature, output voltage, and output current according to the historical surface temperature, historical output voltage, and historical output current of the photovoltaic panel in the central processing unit, and apply them to the thermal analysis chart, bar chart, and streamline chart in the new 3D visualization model; Deploy a graph neural network in the 3D visualization model, construct a component fault propagation model, and locate the root fault nodes through topological analysis, and generate a causal inference tree based on the fault nodes.

10. The three-dimensional visualization management system of the photovoltaic energy storage system according to claim 1, characterized in that: The steps to generate optimization suggestions are as follows: Retrieve the temperature, voltage, and current data before the photovoltaic panel alarm is triggered and the historical fault logs from the central processing unit, and start the trained prediction model. Input the temperature, voltage, and current before the alarm is triggered and the historical temperature, historical voltage, and historical current data in the historical fault logs to obtain a quantitative prediction of the temperature distribution, power generation, and current fluctuation of the photovoltaic panel and a risk level assessment; Based on the quantitative prediction and risk level assessment, put forward targeted suggestions and generate optimization parameters; Through the simulation function in the 3D visualization model, perform simulations according to the simulation parameters and the collected environmental data of temperature, current, and voltage to obtain the optimized scenario simulation results; Compare the optimized scenario simulation results with the temperature, voltage, and current data before the alarm trigger, evaluate the actual effect of the optimization measures, and formulate a preventive maintenance plan based on the actual areas of potential problems.

Images