MPPT photovoltaic power optimization method based on photovoltaic modules, and system

Abstract

An MPPT photovoltaic power optimization method based on photovoltaic modules, and a system, which relate to the technical field of photovoltaic power optimization. The method comprises: collecting photovoltaic signals, and analyzing the collected signals; on the basis of a deep learning algorithm, learning historical data and real-time environmental data to predict future illumination and temperature change trends, and generating a corresponding power point prediction model; using an edge computing device to locally process the data and implement control, and adjusting the output power of a photovoltaic module in real time on the basis of a generated MPPT control strategy; and by means of a reinforcement learning algorithm, optimizing an installation layout of photovoltaic modules. By means of introducing a deep learning model combined with dynamic step size adjustment and perturb-and-observe strategies, accurate and efficient tracking of a maximum power point for a photovoltaic system in a complex environment is realized, thus effectively improving the stability and response speed of power output. Power fluctuations and losses are reduced, thereby significantly improving the overall energy efficiency of a photovoltaic power generation system.

Description

MPPT Photovoltaic Power Optimization Method and System Based on Photovoltaic Modules Technical Field

[0001] This application relates to the field of photovoltaic power optimization technology, and in particular to a method and system for MPPT photovoltaic power optimization based on photovoltaic modules. Background Technology

[0002] With the increasing global demand for renewable energy, photovoltaic (PV) power generation systems have become an important component of modern power systems due to their clean, efficient, and sustainable characteristics. However, the output power of PV systems is affected by various environmental factors, including changes in sunlight intensity, temperature, module orientation, and shading, causing their maximum power point (MPP) to fluctuate dynamically. To maximize the power output of PV systems under various complex conditions, maximum power point tracking (MPPT) technology has emerged. Existing MPPT methods, such as perturbation and observation (P&O) and incremental conductance methods, while capable of tracking the MPP well in certain situations, still suffer from insufficient response speed and tracking accuracy in complex environments, especially under conditions of rapid changes in sunlight or partial shading of modules, which can easily lead to decreased tracking efficiency or even failure. Furthermore, the relatively fixed step size adjustment of traditional methods results in significant oscillations or efficiency losses during the power tracking process. To address these issues, combining deep learning and intelligent control technologies to further improve the MPPT efficiency of PV systems has become a current research focus. Summary of the Invention

[0003] In view of the aforementioned existing problems, this application is hereby filed.

[0004] Therefore, this application provides a photovoltaic power optimization method based on photovoltaic modules (MPPT), which mainly solves the problems of slow response speed, low tracking accuracy, and power fluctuations of traditional MPPT algorithms in complex and variable environmental conditions. Existing MPPT methods, when encountering rapidly changing illumination conditions or partial shading of modules, may suffer from slow tracking or significant power loss due to their fixed step size adjustment method, failing to achieve accurate and real-time tracking of the maximum power point. Furthermore, traditional methods lack effective prediction mechanisms when processing multi-dimensional environmental data, relying solely on currently collected data for adjustment, making it difficult to cope with power fluctuations in complex environments. This application introduces a deep learning model (such as LSTM or Transformer) to predict future power change trends by learning from historical operating data and real-time environmental data of the photovoltaic system. Combined with dynamic step size adjustment and perturbation observation, various control strategies are employed to achieve accurate and efficient tracking of the maximum power point, effectively solving the problem of poor performance of traditional MPPT algorithms in complex environments.

[0005] To address the aforementioned technical problems, this application provides the following technical solution: a photovoltaic power optimization method based on photovoltaic modules using MPPT (Multi-Level Photovoltaic Power Optimization), comprising:

[0006] The photovoltaic (PV) module's input and output signals are acquired in real time through various sensors. These signals are then transmitted to a data processing module, which uses a neural network algorithm to analyze the data, assess the PV module's current state, and determine its output power curve and maximum power point (MPPT). Based on deep learning algorithms, historical and real-time environmental data are used to predict future trends in sunlight and temperature, generating a corresponding MPPT prediction model for dynamically adjusting the MPPT control strategy. Edge computing devices are used to process the data locally and implement control, adjusting the PV module's output power in real time according to the generated MPPT control strategy to ensure the power output always approaches the maximum power point. Finally, reinforcement learning algorithms are used to optimize the PV module's installation layout, including installation density, orientation, and tilt angle, to achieve optimal power output for the PV system.

[0007] As a preferred embodiment of the MPPT photovoltaic power optimization method based on photovoltaic modules described in this application, the input and output signals include current, voltage, light intensity, temperature, and wind speed signals.

[0008] As a preferred embodiment of the MPPT photovoltaic power optimization method based on photovoltaic modules described in this application, the real-time acquisition of input and output signals includes multiple current sensors, voltage sensors, light intensity sensors, temperature sensors, and wind speed sensors on the photovoltaic modules. Each sensor collects corresponding data of the environment and the modules, and the data is transmitted to the central processing unit via a wireless or wired network.

[0009] As a preferred embodiment of the MPPT photovoltaic power optimization method based on photovoltaic modules described in this application, the analysis of the collected data using a neural network algorithm includes learning from historical operating data and real-time environmental data of the photovoltaic system to capture the variation law of photovoltaic module power output under different environmental conditions, and predicting future power fluctuations through time series analysis; the prediction model is trained based on light intensity, temperature, and the current output voltage and current parameters of the module.

[0010] As a preferred embodiment of the MPPT photovoltaic power optimization method based on photovoltaic modules described in this application, the step of analyzing the collected data using a neural network algorithm further includes assuming that the input multi-sensor signal is... Define the input signal matrix ,in For the number of sensors, The number of sampling points within the time window, for each sensor The signal is represented as .

[0011] CNNs are used to extract features from input signals, defining the convolution kernel. As the first Layer convolution kernel, where The width of the convolution kernel. Let the height of the convolution kernel be denoted as: The convolution operation can be represented as:

[0012]

[0013] in, For the first Feature map of the layer For activation function, For bias terms, This represents the convolution operation. Through multiple layers of convolution and pooling operations in a CNN, features are extracted from the signal, generating a high-dimensional feature vector. ,in For feature dimensions;

[0014] Extracting feature vectors Then, the correlation between signals from different sensors was analyzed using Bayesian estimation, assuming that the feature distribution of each sensor is a conditional probability distribution. Bayes' theorem is expressed as:

[0015]

[0016] in, and They are from the first and the Feature vectors of each sensor Let be the conditional probability, representing the probability under given conditions. In the case of The probability of occurrence for The probability, for The probability of.

[0017] To enhance the performance of Bayesian estimation, prior information is introduced. The correlation of signals from different sensors is integrated into a weighted model. :

[0018]

[0019] in, These are weighting coefficients used to adjust the balance between prior and observed data; Let represent the prior distribution, and let represent the prior hypothesis about the characteristic distribution of sensor i.

[0020] To filter out noise, the system processes the feature vectors. The state equation of the Kalman filter is as follows: and observation equations They are respectively:

[0021]

[0022]

[0023] in, Here is the state transition matrix. For the observation matrix, For process noise, To observe the noise, the update equation for the Kalman filter is:

[0024] Prediction steps:

[0025]

[0026] Update steps:

[0027]

[0028] in, For Kalman gain, The process noise covariance matrix is... To observe the noise covariance matrix; For Kalman filtering The covariance matrix of the eigenvectors at each time step; The updated feature vector incorporates corrections from both predicted and observed data. For Kalman filtering The predicted value of the feature vector at time step; This is a matrix with diagonal elements of 1 and off-diagonal elements of 0; the current filter removes random noise from the signal, resulting in a smooth eigenvector. This provides input data for subsequent power prediction and MPPT control.

[0029] As a preferred embodiment of the MPPT photovoltaic power optimization method based on photovoltaic modules described in this application, the generation of the corresponding power point prediction model includes an MPPT control strategy based on the prediction results of the deep learning model. The system can adjust the operating voltage and current of the photovoltaic modules in real time to ensure that the photovoltaic modules always approach or reach the maximum power point under the current environmental conditions. The control strategy includes step adjustment, disturbance observation, and dynamic step size adjustment. The deep learning model generates tracking and adjustment control commands for the maximum power point based on light intensity, temperature, and the output characteristics of the photovoltaic modules.

[0030] The reinforcement learning algorithm includes optimizing and adjusting the installation layout of photovoltaic modules through simulation experiments and field tests. The optimization and adjustment includes dynamic adjustment of module spacing, orientation and tilt angle. The reinforcement learning model continuously experiments and updates the layout strategy to select the arrangement with the optimal power output, and automatically adjusts the module installation scheme for different geographical environments and lighting conditions.

[0031] As a preferred embodiment of the MPPT photovoltaic power optimization method based on photovoltaic modules described in this application, the reinforcement learning algorithm further includes an optimization process based on real-time power output data of photovoltaic modules and environmental parameters, including light intensity, geographical location, and shadow distribution factors.

[0032] As a preferred embodiment of the MPPT photovoltaic power optimization system based on photovoltaic modules described in this application, the system includes: a data acquisition module, a data processing module, a central processing unit, and an edge computing module; the data acquisition module is used to acquire input and output signals of the photovoltaic modules in real time, acquiring the signals through different sensors, and acquiring module characteristic information; the data processing module includes analyzing the acquired data using a neural network algorithm and implementing the algorithm; the central processing unit is used to synchronously calibrate the data from different sensors and ensure the timeliness and accuracy of the data through time series analysis; the edge computing module is equipped with an edge computing device including an embedded processor and local memory, the processor is used to calculate and execute the MPPT control algorithm in real time, and the local memory is used to cache input data from sensors and deep learning models.

[0033] A computer device includes a memory and a processor, the memory storing a computer program, characterized in that the processor executes the computer program to implement the steps of an MPPT photovoltaic power optimization method based on photovoltaic modules.

[0034] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of an MPPT photovoltaic power optimization method based on photovoltaic modules.

[0035] The beneficial effects of this application are as follows: This application proposes a photovoltaic power optimization method based on photovoltaic modules (MPPT), which has the following beneficial effects. First, by introducing a deep learning model, the system can learn and predict the power output variation of the photovoltaic system under different illumination, temperature, and shading conditions in real time, achieving more accurate power point prediction. Compared with the traditional MPPT algorithm that only relies on current data for adjustment, deep learning can capture long-term dependencies in complex environments, identify power fluctuation trends in advance, and significantly improve the tracking efficiency of the maximum power point. Second, this application combines step adjustment, perturbation observation, and dynamic step size adjustment strategies. The system can flexibly respond to rapid environmental changes, especially under conditions of rapid changes in illumination and partial shading, and can adjust the operating voltage and current more quickly, keeping the photovoltaic system close to the maximum power output and reducing power loss. In addition, the dynamic step size adjustment mechanism automatically adjusts the step size according to the power change amplitude, avoiding the power oscillation problem caused by a fixed step size, which improves the tracking response speed and reduces energy loss caused by improper adjustment. Finally, the entire MPPT optimization strategy makes full use of multi-dimensional real-time sensor data and historical data, and can maintain efficient operation in complex and ever-changing photovoltaic environments, greatly improving the overall energy efficiency of photovoltaic systems and having broad application prospects. Attached Figure Description

[0036] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 is a schematic flowchart of an MPPT photovoltaic power optimization method based on photovoltaic modules provided in an embodiment of this application.

[0038] Figure 2 is a schematic diagram of the photovoltaic module IV curve and MPP definition annotation of the MPPT photovoltaic power optimization method based on photovoltaic modules provided in an embodiment of this application. Embodiments of the present invention

[0039] To make the above-mentioned objectives, features, and advantages of this application more apparent and understandable, the specific embodiments of this application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without inventive effort should fall within the scope of protection of this application.

[0040] Many specific details are set forth in the following description in order to provide a full understanding of this application. However, this application may also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0041] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of this application. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that excludes other embodiments.

[0042] This application is described in detail with reference to the schematic diagrams. When detailing the embodiments of this application, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not according to the usual scale. Furthermore, the schematic diagrams are merely examples and should not limit the scope of protection of this application. In actual fabrication, the three-dimensional spatial dimensions of length, width, and depth should be included.

[0043] Furthermore, it should be noted in the description of this application that the terms "upper," "lower," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are used solely for the convenience of describing this application and for simplification, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application. In addition, the terms "first," "second," or "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0044] Unless otherwise expressly specified and limited, the terms "installation," "connection," and "joining" in this application should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; similarly, they can refer to mechanical connections, electrical connections, or direct connections, or indirect connections through an intermediate medium, or internal connections between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0045] Example 1, referring to Figure 1, is the first embodiment of this application. This embodiment provides an MPPT photovoltaic power optimization method based on photovoltaic modules, including:

[0046] S1: Real-time acquisition of input and output signals of photovoltaic modules, by acquiring the signals through different sensors.

[0047] S2: The collected signals are transmitted to the data processing module. The data processing module uses a neural network algorithm to analyze the collected data, evaluate the current state of the photovoltaic module, and determine the output power curve and maximum power point of the photovoltaic module.

[0048] S3: Based on deep learning algorithms, it predicts future trends in illumination and temperature by learning from historical data and real-time environmental data, and generates corresponding power point prediction models for dynamically adjusting the maximum power point tracking (MPPT) control strategy.

[0049] S4: Using edge computing devices, the data is processed locally and control is implemented. The output power of the photovoltaic modules is adjusted in real time according to the generated MPPT control strategy to ensure that the power output is always close to the maximum power point.

[0050] S5: Optimize the installation layout of photovoltaic modules through reinforcement learning algorithms, including module installation density, orientation, and tilt angle, to achieve optimal power output of the photovoltaic system.

[0051] The input and output signals include current, voltage, light intensity, temperature, and wind speed signals.

[0052] The real-time acquisition of input and output signals includes multiple current sensors, voltage sensors, light intensity sensors, temperature sensors, and wind speed sensors on the photovoltaic module. Each sensor collects corresponding data of the environment and the module, and the data is transmitted to the central processing unit via wireless or wired networks.

[0053] The analysis of the collected data using neural network algorithms includes learning from the historical operating data and real-time environmental data of the photovoltaic system to capture the variation pattern of photovoltaic module power output under different environmental conditions, and predicting future power fluctuations through time series analysis.

[0054] The prediction model is trained based on light intensity, temperature, and the current output voltage and current parameters of the component.

[0055] Assume the input multi-sensor signal is Define the input signal matrix ,in For the number of sensors, The number of sampling points within the time window, for each sensor The signal is represented as .

[0056] CNNs are used to extract features from input signals, defining the convolution kernel. As the first Layer convolution kernel, where The width of the convolution kernel. Let the height of the convolution kernel be denoted as: The convolution operation can be represented as:

[0057]

[0058] in, For the first Feature map of the layer For activation function, For bias terms, This represents the convolution operation. Through multiple layers of convolution and pooling operations in a CNN, features are extracted from the signal, generating a high-dimensional feature vector. ,in For feature dimensions.

[0059] Extracting feature vectors Then, the correlation between signals from different sensors was analyzed using Bayesian estimation, assuming that the feature distribution of each sensor is a conditional probability distribution. Bayes' theorem is expressed as:

[0060]

[0061] in, and They are from the first and the Feature vectors of each sensor Let be the conditional probability, representing the probability under given conditions. In the case of The probability of occurrence for The probability, for The probability of.

[0062] To enhance the performance of Bayesian estimation, prior information is introduced. The correlation of signals from different sensors is integrated into a weighted model. :

[0063]

[0064] in, These are weighting coefficients used to adjust the balance between prior and observed data; Let represent the prior distribution, and let represent the prior hypothesis about the characteristic distribution of sensor i.

[0065] To filter out noise, the system processes the feature vectors. The state equation of the Kalman filter is as follows: and observation equations They are respectively:

[0066]

[0067]

[0068] in, Here is the state transition matrix. For the observation matrix, For process noise, To observe the noise, the update equation for the Kalman filter is:

[0069] Prediction steps:

[0070]

[0071] Update steps:

[0072]

[0073] in, For Kalman gain, The process noise covariance matrix is... To observe the noise covariance matrix; For Kalman filtering The covariance matrix of the eigenvectors at each time step; The updated feature vector incorporates corrections from both predicted and observed data. For Kalman filtering The predicted value of the feature vector at time step; This is a matrix with diagonal elements of 1 and off-diagonal elements of 0; the current filter removes random noise from the signal, resulting in a smooth eigenvector. This provides input data for subsequent power prediction and MPPT control.

[0074] By integrating high-quality data from CNN feature extraction, Bayesian correlation analysis, and Kalman filtering for denoising. Power prediction can be performed. Let the power output prediction model be... It can be represented by a simple linear regression model or LSTM neural network as follows:

[0075]

[0076] in, For the prediction function, This represents the prediction error.

[0077] Based on the predicted power value, the MPPT algorithm adjusts the operating point of the photovoltaic system in real time to bring it close to the maximum power point. The adjustment rule of MPPT can be expressed as:

[0078]

[0079] in, This represents the optimal operating voltage for the photovoltaic module at time t. This represents the optimal operating voltage for the photovoltaic module at time t. This represents the change in voltage during MPPT adjustment.

[0080] The generation of the corresponding power point prediction model includes an MPPT control strategy based on the prediction results of the deep learning model. The system can adjust the operating voltage and current of the photovoltaic module in real time so that the photovoltaic module always approaches or reaches the maximum power point under the current environmental conditions. The control strategy includes step adjustment, disturbance observation, and dynamic step size adjustment.

[0081] The deep learning model generates tracking and adjustment control commands for the maximum power point based on light intensity, temperature, and the output characteristics of the photovoltaic module.

[0082] Predicting future moments using deep learning models power output Assume at time... The actual power output of the system is Calculate the current power error:

[0083]

[0084] in, The maximum power point predicted by the deep learning model The corresponding power, This indicates the deviation between the current power and the maximum power point. This error feedback guides the system to make adjustments.

[0085] Step-by-step regulation method adjusts the voltage or current To approach the maximum power point Let the voltage at the current moment be... The voltage at the next moment is adjusted according to the step-by-step adjustment method as follows:

[0086]

[0087] in, This is the step adjustment amount. It can be adjusted based on the power error. To dynamically adjust the step size:

[0088]

[0089] in, It is the proportional coefficient for step adjustment. In this way, the system gradually adjusts the voltage to approach the maximum power point.

[0090] Traditional The algorithm uses perturbation voltage The algorithm observes power changes to determine the adjustment direction. If the power increases, the perturbation continues in that direction; otherwise, the perturbation direction is changed. The improved algorithm incorporates prediction information from deep learning, using the predicted... Adjust step direction:

[0091]

[0092] in, It is the perturbation step size. The sign function determines the direction of voltage adjustment based on the difference between the predicted power and the current power. Unlike traditional P&O algorithms, this method uses the predicted power point rather than the actual measured power as a reference.

[0093] The dynamic step size adjustment method dynamically adjusts the step size based on the changing trends of power error and voltage / current. Its goal is to adjust the step size as needed when the system approaches a certain level. The step size adjustment formula is:

[0094]

[0095] in, The adjustment coefficient is the step size. The adjustment is dynamically adjusted based on the magnitude of the power error. When the error is large, the step size increases; when the error is small, the step size decreases, ensuring fine-grained adjustment as the system approaches the maximum power point.

[0096] By combining deep learning models, we can utilize power gradient information to more accurately adjust system parameters. Let the gradient of power relative to voltage predicted by the deep learning model be... The system can use gradient information for dynamic adjustment:

[0097]

[0098] in, This is the learning rate, used to control the magnitude of the adjustment. Using this gradient-based method, the system can directly adjust in the direction of increasing power, improving adjustment efficiency.

[0099] Based on real-time power error Choose different adjustment strategies:

[0100] When the power error is large, step adjustment and dynamic step size adjustment methods are used to quickly adjust the voltage to approach the maximum power point.

[0101] When the power error is small, a gradient-based tracking method is used for fine-tuning.

[0102] When the trend of power change is unclear, the perturbation observation method combined with the prediction results of deep learning is used to determine the direction of adjustment.

[0103] Therefore, the final voltage adjustment formula can be expressed as:

[0104]

[0105] in, The value is determined based on the current power error, gradient, and disturbance observations:

[0106]

[0107] in, and It is the error threshold, used to distinguish different adjustment strategies.

[0108] The reinforcement learning algorithm includes optimizing and adjusting the installation layout of photovoltaic modules through simulation experiments and field tests. The optimization and adjustment includes dynamic adjustment of module spacing, orientation and tilt angle. The reinforcement learning model continuously experiments and updates the layout strategy to select the arrangement with the optimal power output, and automatically adjusts the module installation scheme for different geographical environments and lighting conditions.

[0109] Through multiple iterations and feedback, the system gradually optimizes the installation density and arrangement of the components to ensure maximum power output under different environments, while also taking into account the shading problem between photovoltaic modules and land utilization.

[0110] The reinforcement learning algorithm also includes an optimization process based on real-time power output data of the photovoltaic module and environmental parameters, including light intensity, geographical location, and shadow distribution factors.

[0111] The optimized installation layout is determined through multiple iterations of reinforcement learning algorithms. The system will adjust the layout strategy in real time based on the geographical environment, component characteristics and meteorological data, and verify its effectiveness through simulation and field testing to ensure the best power output efficiency of photovoltaic modules in complex geographical environments.

[0112] Example 2, referring to Figure 2, is an embodiment of this application, providing an MPPT photovoltaic power optimization method based on photovoltaic modules. To verify the beneficial effects of this application, scientific demonstration is carried out through experiments.

[0113] The neural network algorithm in this application is a Long Short-Term Memory (LSTM) network or a Transformer structure, which predicts future power fluctuations by learning historical data of the photovoltaic system and real-time environmental data. In order to achieve accurate tracking of the maximum power point, the model processes multi-dimensional time-series data such as input irradiance, temperature, output voltage and current, and captures the dynamic changes of photovoltaic system power under complex environmental conditions through time series analysis methods.

[0114] Suppose the input multidimensional time series data is ,in Indicates the number of sensors (e.g., light intensity, temperature, voltage, current). This represents the number of time steps. The data at each time step can be represented as a vector:

[0115]

[0116] in, Indicates from the Each sensor in time The measured value at that time.

[0117] The basic state equations of LSTM are as follows:

[0118] Forgotten Gate:

[0119]

[0120] in, For the output of the forget gate, Here is the weight matrix for the forget gate. For bias terms, This is the hidden state from the previous time step. Input for the current time step. This is the activation function (usually the Sigmoid function).

[0121] Input Gate:

[0122]

[0123] in, For input gate, Candidate memory cell state, and This is the weight matrix. and This is a bias term.

[0124] Memory unit update:

[0125]

[0126] in, This is the updated memory cell state. This represents the state of the memory unit at the previous time step.

[0127] Output gate:

[0128]

[0129] in, For output gate, This represents the hidden state at the current time step. The final output of the LSTM. This indicates a prediction of power output for the next time step.

[0130]

[0131] in, The weight matrix of the output layer. This is a bias term.

[0132] The Transformer is a neural network architecture based on the attention mechanism, suitable for handling time series problems with long-term dependencies. To capture the dynamic characteristics of power fluctuations under complex environmental conditions, we can apply the Transformer structure to power prediction.

[0133] The core of the Transformer lies in its self-attention mechanism, which learns important dependencies by comparing the similarity of each element in the input sequence with other elements. The self-attention mechanism can be represented as:

[0134]

[0135] in, These are the query matrix, key matrix, and value matrix, respectively. is the dimension of the key vector.

[0136] In our application scenario, the input signal After linear transformation, we obtain :

[0137]

[0138] Transformer's multi-head self-attention mechanism computes multiple attention heads in parallel to capture different dependencies:

[0139]

[0140] Finally, the feedforward network in the Transformer is applied to the output at each time step:

[0141]

[0142] The prediction results output by the Transformer This indicates a power prediction for the next time step:

[0143]

[0144] Whether it's LSTM or Transformer, the predicted power value Based on input multidimensional time series data The loss function uses mean squared error (MSE) to optimize the model parameters.

[0145]

[0146] in, For true power output, This represents the power value predicted by the model.

[0147] Future power prediction based on LSTM or Transformer The system can dynamically adjust the operating point (voltage, current) of the photovoltaic modules to achieve maximum power point tracking. The MPPT adjustment strategy is as follows:

[0148]

[0149] in, To adjust the step size, It represents the power gradient relative to voltage, and is used to guide voltage adjustments.

[0150] Furthermore, by tracking the changes in output voltage and current of the photovoltaic array in real time due to variations in the operating environment, relevant control methods or technologies can be used to ensure that the photovoltaic array operates at its maximum power point (MPP) on its I or PV curve. This automatic adjustment to achieve maximum power output is known in the industry as Maximum Power Point Tracking (MPPT) control technology. Figure 2 shows the typical volt-ampere characteristic and power output characteristic curves of a photovoltaic module. It can be seen that the power curve has a peak point, which is the maximum power point.

[0151] It should be noted that the above embodiments are only used to illustrate the technical solutions of this application and are not intended to limit it. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of this application without departing from the spirit and scope of the technical solutions of this application, and all such modifications or substitutions should be covered within the scope of the claims of this application.

[0152] Example 3

[0153] The third embodiment of this application differs from the first two embodiments in that:

[0154] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0155] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or one or more blocks of the block diagrams.

[0156] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0157] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0158] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0159] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

[0160] Example 4 is the fourth embodiment of this application. This embodiment provides an MPPT photovoltaic power optimization system based on photovoltaic modules, including: the system includes a data acquisition module, a data processing module, a central processing unit, and an edge computing module;

[0161] The data acquisition module is used to acquire input and output signals of the photovoltaic module in real time. It acquires the signals through different sensors and also acquires module characteristic information.

[0162] The data processing module includes analyzing the collected data using neural network algorithms and implementing the algorithms. The data processing module also receives system control commands, which include voltage and current adjustment values. The signals are transmitted through a digital signal processor (DSP) or microcontroller. The edge computing device parses and executes these control commands locally to ensure the system's real-time response capability and power optimization effect.

[0163] The central processing unit is used to synchronously calibrate data from different sensors and ensure the timeliness and accuracy of the data through time series analysis.

[0164] The edge computing module, which houses the edge computing device, includes an embedded processor and local memory. The processor performs real-time calculations and executes the MPPT control algorithm, while the local memory caches input data from sensors and deep learning models. The edge computing device exchanges data with the central control system via a low-latency communication network. However, in the event of network latency or outages, the edge device can operate independently, maintaining real-time control over the power output of the photovoltaic modules.

Claims

1. A photovoltaic power optimization method based on photovoltaic modules using MPPT, characterized in that: include, The photovoltaic module's input and output signals are acquired in real time and transmitted to the data processing module. The data processing module uses neural network algorithms to analyze the acquired data and evaluate the current status of the photovoltaic module. The analysis of the collected data using neural network algorithms includes learning from the historical operating data and real-time environmental data of the photovoltaic system to capture the variation pattern of photovoltaic module power output under different environmental conditions, and predicting future power fluctuations through time series analysis. The prediction model is trained based on light intensity, temperature, and the current output voltage and current parameters of the component. Assume the input multi-sensor signal is Define the input signal matrix ,in For the number of sensors, The number of sampling points within the time window, for each sensor The signal is represented as ； CNNs are used to extract features from input signals, defining the convolution kernel. As the first Layer convolution kernel, where The width of the convolution kernel. Let the height of the convolution kernel be denoted as: The convolution operation can be represented as: in, For the first Feature map of the layer For activation function, For bias terms, This represents a convolution operation; through multiple layers of convolution and pooling operations in a CNN, features are extracted from the signal to generate a high-dimensional feature vector. ,in For feature dimensions; Extracting feature vectors Then, the correlation between signals from different sensors was analyzed using Bayesian estimation, assuming that the feature distribution of each sensor is a conditional probability distribution. Bayes' theorem is expressed as: in, and They are from the first and the Feature vectors of each sensor Let be the conditional probability, representing the probability under given conditions. In the case of The probability of occurrence for The probability, for The probability of; To enhance the performance of Bayesian estimation, prior information is introduced. The correlation of signals from different sensors is integrated into a weighted model. ： in, These are weighting coefficients used to adjust the balance between prior and observed data; Let represent the prior distribution, and let represent the prior hypothesis about the characteristic distribution of sensor i. To filter out noise, the system processes the feature vectors. The state equation of the Kalman filter is as follows: and observation equations They are respectively: in, Here is the state transition matrix. For the observation matrix, For process noise, To observe the noise, the update equation for the Kalman filter is: Prediction steps: Update steps: in, For Kalman gain, The process noise covariance matrix is... To observe the noise covariance matrix; For Kalman filtering The covariance matrix of the eigenvectors at each time step; The updated feature vector incorporates corrections from both predicted and observed data. For Kalman filtering The predicted value of the feature vector at time step; This is a matrix with diagonal elements of 1 and off-diagonal elements of 0; the current filter removes random noise from the signal, resulting in a smooth eigenvector. This provides input data for subsequent power prediction and MPPT control; Based on deep learning algorithms, by learning from historical data and real-time environmental data, the system predicts future trends in light and temperature changes and generates corresponding power point prediction models. Using edge computing devices, the data is processed locally and control is implemented, adjusting the output power of the photovoltaic modules in real time according to the generated MPPT control strategy; By using reinforcement learning algorithms, the installation layout of photovoltaic modules is optimized to achieve the optimal power output of the photovoltaic system.

2. The MPPT photovoltaic power optimization method based on photovoltaic modules as described in claim 1, characterized in that: The input and output signals include current, voltage, light intensity, temperature, and wind speed signals.

3. The MPPT photovoltaic power optimization method based on photovoltaic modules as described in claim 2, characterized in that: The real-time acquisition of input and output signals includes multiple current sensors, voltage sensors, light intensity sensors, temperature sensors, and wind speed sensors on the photovoltaic module. Each sensor collects corresponding data of the environment and the module, and the data is transmitted to the central processing unit via wireless or wired networks.

4. The MPPT photovoltaic power optimization method based on photovoltaic modules as described in claim 3, characterized in that: The generation of the corresponding power point prediction model includes an MPPT control strategy based on the prediction results of the deep learning model. The system can adjust the operating voltage and current of the photovoltaic module in real time to keep the photovoltaic module close to or reach the maximum power point under the current environmental conditions. The control strategy includes step adjustment, disturbance observation, and dynamic step size adjustment. The deep learning model generates tracking and adjustment control commands for the maximum power point based on light intensity, temperature, and the output characteristics of the photovoltaic module. The reinforcement learning algorithm includes optimizing and adjusting the installation layout of photovoltaic modules through simulation experiments and field tests. The optimization and adjustment includes dynamic adjustment of module spacing, orientation and tilt angle. The reinforcement learning model continuously experiments and updates the layout strategy to select the arrangement with the optimal power output, and automatically adjusts the module installation scheme for different geographical environments and lighting conditions.

5. The MPPT photovoltaic power optimization method based on photovoltaic modules as described in claim 4, characterized in that: The reinforcement learning algorithm also includes an optimization process based on real-time power output data of the photovoltaic module and environmental parameters, including light intensity, geographical location, and shadow distribution factors.

6. A system employing the MPPT photovoltaic power optimization method based on photovoltaic modules as described in any one of claims 1 to 5, characterized in that: The system includes a data acquisition module, a data processing module, a central processing unit, and an edge computing module; The data acquisition module is used to acquire input and output signals of the photovoltaic module in real time. It acquires the signals through different sensors and also acquires module characteristic information. The data processing module includes analyzing the collected data using neural network algorithms and implementing the algorithms. The central processing unit is used to synchronously calibrate data from different sensors and ensure the timeliness and accuracy of the data through time series analysis. The edge computing module is used to carry edge computing devices, including an embedded processor and local memory. The processor is used to calculate and execute the MPPT control algorithm in real time, and the local memory is used to cache input data from sensors and deep learning models.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images