A photovoltaic module adaptive MPPT optimization method, system, device and storage medium

Abstract

The present application relates to the technical field of photovoltaic power generation, and discloses a photovoltaic module adaptive MPPT optimization method, system, device and storage medium, multi-dimensional parameter information of the photovoltaic module is collected through adaptive weight distribution, a weighted original signal is generated, and a filter signal is generated through dynamic filtering processing; based on the filter signal, a principal component feature vector is generated through principal component analysis, and the intensity of illumination is solved, and a multi-parameter coupling prediction model is used to calculate the predicted power; based on the predicted power, the control action optimization reference voltage is optimized through model predictive control, and the multi-parameter abnormality detection weight is adjusted based on adaptive fuzzy logic, and predictive fault correction is performed, so that maximum power point tracking is realized. Through adaptive weight distribution and dynamic Gaussian filtering, the present application effectively reduces noise interference; combined with principal component analysis and a multi-parameter coupling prediction model, the power output is accurately predicted; through model predictive control, the reference voltage is optimized, so that rapid and accurate maximum power point tracking is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic power generation, and particularly relates to a photovoltaic module adaptive MPPT optimization method, system, device and storage medium. BACKGROUND

[0002] As an important form of renewable energy, photovoltaic power generation is widely used in distributed power generation and grid-connected systems. Maximum power point tracking (MPPT) technology is the core of improving the efficiency of photovoltaic systems, aiming to dynamically adjust the operating voltage to track the maximum power point. Traditional MPPT methods, such as perturbation and observation method and incremental conductance method, perform well in stable environments, but in complex conditions such as light fluctuation, temperature change or partial shading, they are easily trapped in local maximum power points, leading to efficiency decline. In recent years, MPPT methods based on fuzzy logic, neural networks and other intelligent algorithms have gradually emerged, but these methods rely on single parameters such as voltage and current, and are difficult to fully capture the coupling effects of multi-dimensional parameters such as temperature and impedance. In addition, existing technologies lack dynamic filtering mechanisms in signal processing, making it difficult to effectively deal with noise interference, and have limited fault detection and correction capabilities, affecting system robustness.

[0003] Therefore, there is an urgent need for an MPPT optimization method that integrates multi-parameter coupling, dynamic signal processing and intelligent control to improve the power tracking accuracy and stability of photovoltaic systems in complex environments. The present application proposes an MPPT optimization scheme based on adaptive weight distribution, dynamic Gaussian filtering, principal component analysis, model predictive control and fuzzy logic, aiming to solve the above problems. SUMMARY

[0004] The present application aims to provide a photovoltaic module adaptive MPPT optimization method, system, device and storage medium to solve the problem of low power tracking accuracy and poor efficiency in complex environments in the prior art.

[0005] To achieve one of the above application purposes, an embodiment of the present application provides a photovoltaic module adaptive MPPT optimization method, which comprises: collecting multi-dimensional parameter information of the photovoltaic module through adaptive weight distribution, generating a weighted original signal, and generating a filtered signal through dynamic filtering processing; based on the filtered signal, generating a principal component feature vector through principal component analysis, calculating the intensity of light, and calculating the predicted power using a multi-parameter coupling prediction model; based on the predicted power, optimizing the reference voltage of the control action through model predictive control, and adjusting the multi-parameter abnormality detection weight based on adaptive fuzzy logic, and performing predictive fault correction to realize maximum power point tracking.

[0006] As a further improvement of an embodiment of the present application, the method further comprises that the acquiring the multi-dimensional parameter information of the photovoltaic module by adaptive weight distribution comprises acquiring voltage, current, temperature and impedance of the photovoltaic module, and generating a weighted original signal by adaptive weight distribution, with a formula as follows: ;

[0007] wherein, is the weighted original signal, is the voltage, is the current, is the temperature, is the impedance, and t is a time variable, used to represent the value of the multi-dimensional parameter of the photovoltaic module at a certain time, is the corresponding weight, dynamically adjusted based on the parameter change rate, and the adjustment formula of the weight is as follows:

[0008] ;

[0009] wherein, is the corresponding weight; represents , , , , is the value of the i-th parameter at the previous time; is the weight adjustment sensitivity coefficient, used to adjust the response degree of the weight to the parameter change; j is an index, traversing the four parameters of voltage, current, temperature and impedance, ranging from 1 to 4.

[0010] As a further improvement of an embodiment of the present application, the method further comprises that the generating a filter signal by dynamic filtering processing is used to reduce noise and enhance the stability of maximum power point tracking; and the dynamic filtering processing adopts dynamic Gaussian filtering, with a formula as follows: ;

[0011] wherein, is the filter signal generated after the dynamic Gaussian filtering processing at t; n is an index of a certain time within a sampling window, and N is the size of the sampling window; is the sampling time interval; is the dynamic variance, adjusted based on the change rate of the light intensity and the temperature, with a formula as follows: ;

[0012] wherein, is the reference variance; is the light intensity, with an initial value being a calibration value, is the light intensity at the previous time, is the standard light intensity; is a temperature variation weight coefficient, used to balance the influence of temperature on filtering; is a temperature of a previous time.

[0013] As a further improvement of an embodiment of the present application, the method further comprises that the "generating a principal component feature vector by principal component analysis, calculating the light intensity, and calculating the predicted power by using a multi-parameter coupling prediction model" comprises that the principal component feature vector is expressed as: ;

[0014] wherein, is a principal component feature vector of a current time, is a principal component matrix trained by offline analysis of historical photovoltaic system operation data, is a voltage and a current obtained based on a filtered signal; and a calculation formula of the predicted power is: ;

[0015] wherein, is a predicted power of a next time, is a current control action, is a multi-parameter coupling prediction function trained by learning historical photovoltaic system operation data and corresponding control actions, and capable of capturing a nonlinear relationship among multiple variables to predict a future power output of the system.

[0016] As a further improvement of an embodiment of the present application, the method further comprises that a calculation formula of the light intensity is:

[0017] ;

[0018] wherein, is an instantaneous power, is an impedance correction factor, is a power value under standard test conditions, is a temperature value under standard test conditions, is a temperature compensation factor, used to correct a nonlinear influence of temperature on calculation of the light intensity; and a calculation formula of the impedance correction factor is:

[0019] ;

[0020] wherein, is a dynamic impedance influence coefficient, is a reference impedance influence coefficient, is an impedance estimation sensitivity coefficient, is a photovoltaic module impedance value under standard test conditions, is a photovoltaic module impedance of a previous time; is an interaction influence coefficient; and a calculation formula of the temperature compensation factor is:

[0021] ;

[0022] wherein, a represents a temperature attenuation index, used to represent the nonlinear influence of temperature on the solution of the light intensity; and The solution is realized by iterative solution, and the specific steps include: initializing to the light intensity of the previous moment or a preset initial value; based on the current , the is calculated; based on the current and , the is calculated; through iterative calculation, until is less than a preset convergence threshold or reaches a maximum number of iterations; wherein, is the initial value of the light intensity in the iterative solution process, is the light intensity value obtained by the n-th iteration in the iterative solution process, is the impedance correction factor obtained by the n-th iteration in the iterative solution process, is the light intensity value obtained by the n+1-th iteration in the iterative solution process, is the absolute value of the difference between the light intensity values obtained by adjacent two iterations, used to judge whether the iteration converges.

[0023] As a further improvement of an embodiment of the present application, the method further comprises that the optimization of the reference voltage of the control action by the model predictive control optimization includes maximizing the predicted power by optimizing the objective function, and the formula is:

[0024] ;

[0025] wherein, is the optimal control action, is the length of the prediction time domain, is the predicted power at the future k moment; and the update formula of the reference voltage is:

[0026] ;

[0027] wherein, is the reference voltage at the next moment, is the voltage adjustment optimized based on the model predictive control, is the optimal working voltage at the next moment based on the model prediction, is the reference voltage value at the current moment.

[0028] As a further improvement of the embodiment of the present application, the method further comprises that the adjusting the multi-parameter anomaly detection weight based on adaptive fuzzy logic, the executing the predictive fault correction comprises calculating an anomaly index to detect the system operation state, the formula is: ;

[0029] wherein, is a temperature change value, is an impedance change value, is a weight coefficient of anomaly detection, used to balance the contribution of power, temperature and impedance change to the anomaly index; the adaptive adjustment is realized by a fuzzy logic controller, the fuzzy logic controller takes the illumination intensity change rate, the temperature change rate and the current anomaly index trend as input, and outputs the adjustment amount of the corresponding weight; the fuzzy logic controller realizes adaptive adjustment by the following steps: defining the fuzzy set and membership function of the input variable and output variable, and establishing a fuzzy rule base; according to the fuzzification result of the current system state and the fuzzy rule base, fuzzy reasoning is carried out; the fuzzy reasoning result is defuzzified into a clear weight adjustment amount ; the weight is updated as:

[0030] ;

[0031] wherein, is the anomaly detection weight value of the previous moment, is the weight adjustment amount calculated by the adaptive fuzzy logic controller at the current moment, X is an index, corresponding to P, T, Z; the correction is triggered when the anomaly index exceeds the preset threshold, and the correction formula is: ;

[0032] wherein, is the reference voltage used for fault correction by the system at the next moment after the anomaly occurs and triggers the correction, is the reference voltage at the current moment; is the maximum power point voltage value at the previous moment, is the maximum power point voltage value at the previous two moments, is a correction weight coefficient, used to adjust the correction amplitude.

[0033] To achieve one of the above-mentioned purposes, an embodiment of the present application further provides a photovoltaic module adaptive MPPT optimization system, which comprises a collection module, a prediction module and a correction module; the collection module is used to collect multi-dimensional parameter information of the photovoltaic module through adaptive weight distribution, generate a weighted original signal, and generate a filtered signal through dynamic filtering processing; the prediction module is used to generate a principal component feature vector through principal component analysis based on the filtered signal, calculate the light intensity, calculate the predicted power by using a multi-parameter coupling prediction model; the correction module is used to optimize the control action reference voltage through model predictive control based on the predicted power, adjust the multi-parameter abnormality detection weight based on adaptive fuzzy logic, and perform predictive fault correction to realize maximum power point tracking.

[0034] To achieve one of the above-mentioned purposes, an embodiment of the present application further provides an electronic device, which comprises a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the program to realize the steps of the photovoltaic module adaptive MPPT optimization method.

[0035] To achieve one of the above-mentioned purposes, an embodiment of the present application further provides a storage medium, which stores a computer program, wherein the computer program is executed by a processor to realize the steps of the photovoltaic module adaptive MPPT optimization method.

[0036] Compared with the prior art, the photovoltaic module adaptive MPPT optimization method, system, device and storage medium provided by the present application effectively reduce noise interference and improve the stability of multi-dimensional parameter processing through adaptive weight distribution and dynamic Gaussian filtering; the power output is accurately predicted by combining principal component analysis and a multi-parameter coupling prediction model; the reference voltage is optimized through model predictive control to realize rapid and accurate maximum power point tracking; the adaptive fuzzy logic enhances the abnormality detection and fault correction capability and improves the system robustness. Compared with the traditional method, the present application significantly improves the power tracking efficiency and system reliability in complex environments, and is suitable for distributed photovoltaic and micro-grid applications. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 is the overall flowchart of the photovoltaic module adaptive MPPT optimization method described in the present application.

[0038] Figure 2 is the architecture schematic diagram of the photovoltaic module adaptive MPPT optimization system described in the present application. DETAILED DESCRIPTION

[0039] The present application will be described in detail below with reference to the specific embodiments shown in the drawings. However, these embodiments do not limit the present application, and the structural, method, or functional changes made by those skilled in the art based on these embodiments are included in the protection scope of the present application.

[0040] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the drawings, in which the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary and are used only to explain the present application, and cannot be understood as limiting the present application.

[0041] In the first embodiment of the present application, the present application provides an adaptive MPPT optimization method for a photovoltaic module, as shown in Figure 1 The method comprises the following steps: S1: acquiring multi-dimensional parameter information of the photovoltaic module through adaptive weight distribution, generating a weighted original signal, and generating a filtered signal through dynamic filtering processing; S2: based on the filtered signal, generating a principal component feature vector through principal component analysis, calculating the intensity of light, and calculating the predicted power by using a multi-parameter coupling prediction model; S3: based on the predicted power, optimizing the reference voltage of the control action through model predictive control, adjusting the multi-parameter abnormality detection weight based on adaptive fuzzy logic, and performing predictive fault correction to realize maximum power point tracking.

[0042] In one specific embodiment of the present application, the multi-dimensional parameter information of the photovoltaic module is acquired through adaptive weight distribution, specifically, the voltage, current, temperature, and impedance of the photovoltaic module are acquired, and the weighted original signal is generated through adaptive weight distribution, and the formula is:

[0043] wherein, is the weighted original signal, is the voltage, is the current, is the temperature, is the impedance, t is a time variable, and is used to represent the value of the multi-dimensional parameter of the photovoltaic module at a certain time, is the corresponding weight, which is dynamically adjusted based on the parameter change rate, , and the adjustment formula of the weight is:

[0044] ;

[0045] wherein, is the corresponding weight; represents , , , , is the value of the i-th parameter corresponding to the previous moment; is the weight adjustment sensitivity coefficient, used to adjust the degree of response of the weight to the parameter change; j is the index, traversing the voltage, current, temperature and impedance four parameters, ranging from 1 to 4.

[0046] It should be noted that the system monitors the operating state of the photovoltaic module in real time through high-precision sensors, acquires key parameters such as voltage, current, temperature and impedance, and comprehensively characterizes the dynamic characteristics of the photovoltaic system. These parameters are collected at fixed time intervals (determined by experiments) to form time series data, providing a basis for subsequent signal processing and optimal control. The collection process needs to ensure high precision and low delay to adapt to the rapid changes of external conditions such as light intensity and environmental temperature.

[0047] Further, the adaptive weight allocation integrates multi-dimensional parameters through a weighted original signal formula to generate a comprehensive signal . This signal reflects the relative contribution of each parameter to the system state, and the design of the weight aims to dynamically adjust the influence of each parameter to adapt to the operating characteristics of the photovoltaic system under different conditions. The weight adjustment is based on the parameter change rate , through exponential function and normalization processing, so that the weight can quickly respond to the dynamic change of the parameter, for example, when the light intensity suddenly changes, the current weight automatically increases, enhancing the attention to key parameters, thereby improving the quality of the signal.

[0048] It should be noted that the generated weighted original signal is used as the input of the dynamic filtering process, which reduces noise and enhances signal stability, providing a high-quality data basis for principal component analysis and power prediction. The value of the weight adjustment sensitivity coefficient can be determined by analyzing the response speed of the system to sudden changes in light intensity, ensuring that the sensitivity of the weight adjustment matches the dynamic characteristics of the system, and is usually adjusted based on the dynamic response characteristics of the photovoltaic system (such as light change frequency), ensuring that the sensitivity of the weight to the parameter change is moderate. In the experiment, the appropriate value can be selected by comparing the signal quality (such as signal-to-noise ratio) under different values to balance the response speed and stability. In addition, the selection of the collected parameters (voltage, current, temperature, impedance) takes into account the core influencing factors of the photovoltaic system operation, and the introduction of impedance especially enhances the monitoring ability of the component aging or connection problem, providing key information support for subsequent anomaly detection and fault correction.

[0049] In one specific embodiment of the present application, a filtered signal is generated through a dynamic filtering process for reducing noise and enhancing the stability of maximum power point tracking; the dynamic filtering process uses dynamic Gaussian filtering, and the formula is:

[0050] ;

[0051] wherein, is the filtered signal generated after dynamic Gaussian filtering at time t; n is the index of a certain time point within the sampling window, and N is the size of the sampling window; is the sampling time interval; is the dynamic variance, which is adjusted based on the light intensity and temperature change rate, and the formula is: ;

[0052] wherein, is the reference variance; is the light intensity, and the initial value is the calibration value, is the light intensity at the previous time point, is the standard light intensity; is the temperature change weight coefficient, which is used to balance the influence of temperature on filtering; is the temperature at the previous time point.

[0053] It should be noted that the dynamic filtering step is aimed at smoothing the weighted original signal to reduce the influence of environmental noise and measurement error, thereby enhancing the stability of MPPT. Gaussian filtering uses weighted average within the time window, combined with Gaussian kernel function, effectively suppresses high-frequency noise, while retaining the main trend of the signal, and is suitable for the dynamic characteristics of photovoltaic system under rapid changes of light intensity or temperature.

[0054] Further, the dynamic variance is the core parameter of Gaussian filtering, which is dynamically adjusted according to the light intensity change rate and the temperature change rate . When the light or temperature changes significantly, it is increased to relax the filtering range and retain more dynamic information; while in stable environment, it is reduced to enhance the noise suppression effect. This adaptive mechanism ensures that the filtering process can adapt to the complexity of the photovoltaic system operating environment.

[0055] Further, the reference variance can be adjusted according to the noise suppression effect of the filter, and the temperature change weight coefficient can be set by analyzing the degree of influence of temperature change on the system. In the experiment, the filtering effect (such as signal smoothness and dynamic response) under different light and temperature change scenarios can be analyzed, and the value that can balance noise suppression and information retention is selected. The sampling window size N and the time interval are adjusted according to the hardware performance and real-time requirements, and the best value is usually determined through simulation to meet the real-time control requirements.

[0056] It should be noted that the system receives the weighted original signal As input, the high-quality is generated by filtering processing, providing stable input data for subsequent PCA and power prediction. The rate of change of the intensity of illumination and temperature directly affects the filtering parameters, and is linked to the subsequent illumination intensity solving step, ensuring the system's rapid response to environmental changes.

[0057] In one specific implementation scenario of the present application, the filtering processing is implemented on an embedded platform (such as a DSP or ARM microcontroller), using C or MATLAB to develop real-time algorithms. The hardware needs to support fast data sampling and floating-point operations to ensure the real-time nature of the filtering.

[0058] In one specific embodiment of the present application, the principal component feature vector is generated by principal component analysis, and the intensity of illumination is solved, and the predicted power is calculated using a multi-parameter coupled prediction model, specifically, the principal component feature vector is represented as:

[0059] ;

[0060] wherein, is the principal component feature vector at the current time, is the principal component matrix trained by offline analysis of historical photovoltaic system operation data, is the voltage and current obtained based on the filtered signal; the calculation formula of the predicted power is: ;

[0061] wherein, is the predicted power at the next time, is the current control action, is a multi-parameter coupled prediction function trained by learning historical photovoltaic system operation data and corresponding control actions, which can capture the nonlinear relationship between multiple variables and predict the future power output of the system.

[0062] It should be noted that the principal component analysis and power prediction steps aim to extract key features from the filtered multi-dimensional parameters and predict the future power output of the photovoltaic system based on this to optimize the MPPT performance. This step generates a feature vector through PCA, and uses the formula to map the filtered voltage, current, temperature and impedance to a low-dimensional feature space. The principal component matrix is trained by offline analysis of historical photovoltaic system operation data, which can effectively extract the correlation between parameters, reduce redundant information, and improve the computational efficiency of subsequent prediction and control. The implementation of PCA needs to ensure the standardization (zero mean, unit variance) of the input data to ensure the accuracy of feature extraction.

[0063] Further, the principal component matrix is obtained by offline analysis and training of historical photovoltaic system operation data through a principal component analysis method, used to extract key features from multi-dimensional parameters (voltage, current, temperature, impedance) to generate a low-dimensional feature vector , and the specific generation process is as follows:

[0064] Collect historical photovoltaic system operation data, including voltage and impedance. These data are collected by high-precision sensors at fixed time intervals, covering various operating conditions, such as different light intensities, temperature ranges, and component aging or partial shading, etc. Data collection needs to ensure sample diversity, reflecting the dynamic characteristics of photovoltaic systems in actual environments. The data need to be preprocessed, including removing outliers (such as invalid data caused by sensor failure) and standardization (adjusting each parameter to zero mean and unit variance) to eliminate dimensional differences and improve analysis accuracy.

[0065] The preprocessed historical data is processed by the principal component analysis method, and the specific steps include: first, construct a time series data set containing voltage, current, temperature and impedance to form a matrix, where each row represents a data vector at a time point, and each column corresponds to a parameter. Then, calculate the covariance matrix of the data matrix to analyze the correlation between parameters. Based on the covariance matrix, extract the eigenvalues and corresponding eigenvectors by eigenvalue decomposition method. The eigenvectors are sorted according to the eigenvalues, and the eigenvectors with larger eigenvalues correspond to the main variation direction of the data. Select the first several eigenvectors to form the principal component matrix . Each column of the matrix is an eigenvector, which is used to map the original multi-dimensional parameters to a low-dimensional feature space.

[0066] Preferably, the principal component matrix The training process can be completed in an offline environment, and can use computing software or programming tools, such as MATLAB's PCA function, Python's scikit-learn library or R language's statistical analysis package. After training, is stored in the control unit (such as embedded microcontroller or DSP chip) of the photovoltaic system in the form of a fixed matrix, used for real-time calculation of the feature vector. To ensure the applicability of the matrix, it can be retrained according to the new operation data to adapt to the aging of photovoltaic components or environmental changes.

[0067] Further, the prediction function is obtained by training historical photovoltaic system operation data and corresponding control actions through a machine learning method, used to capture the nonlinear relationship between voltage, current, temperature, impedance, light intensity and control action, etc. to predict the power output at the next time The generation process of this function first involves the collection of historical data, including at least several weeks to months of photovoltaic system operation records, covering voltage, current, temperature, impedance, light intensity, control actions and corresponding actual power output. These data are collected in real-time by sensors and labeled with environmental conditions to ensure sample diversity. The data preprocessing steps include outlier removal (such as removing noisy data by threshold filtering), normalization, and feature engineering to improve the generalization ability of the model.

[0068] Preferably, the training of the function uses a supervised learning framework, which can use neural network models (such as multi-layer perceptron or long short-term memory network) or regression algorithms (such as support vector regression) to handle the coupling effects between multiple variables. The specific training method includes: dividing the preprocessed historical data into training set and validation set, defining the loss function as mean square error (MSE), and iteratively training the model parameters by gradient descent optimizer. During the training process, the model needs to optimize the hyperparameters (such as the number of hidden layer neurons or regularization coefficients) to avoid overfitting. Training can be done in an offline environment using programming tools such as TensorFlow or scikit-learn library in Python. After training, the model parameters are stored in the system controller in the form of model parameters for real-time inference. The update mechanism of the model includes periodic retraining (such as every quarter based on new data) to adapt to the aging of photovoltaic components or environmental changes. The implementation of this function ensures the accuracy of the prediction, supporting the dynamic decision-making process of MPPT optimization.

[0069] Further, the current control action is a key input variable in the photovoltaic component adaptive maximum power point tracking (MPPT) optimization method, representing the control instruction applied to the photovoltaic system at time t, used to adjust the working point of the system to achieve maximum power output. In photovoltaic power generation systems, the control action usually refers to specific parameters such as reference voltage adjustment, duty cycle change, or inverter control signal, which directly affect the output power curve of the photovoltaic component. By dynamically modifying , the system can respond to environmental changes (such as light intensity or temperature fluctuations) and guide the working point to approach the maximum power point. Specifically, can be represented as a vector or scalar, for example in a direct-current-direct-current converter, can be defined as the incremental value of the duty cycle; in the model predictive control framework, includes multiple control sequences within the future prediction time domain. The value range of The amplitude of the control action is usually limited to within 5% of the open-circuit voltage of the photovoltaic module to avoid system instability. The generation of the control action relies on the outputs of the upstream modules, such as the principal component feature vector and the illumination intensity solution, ensuring seamless integration with the multi-parameter coupled prediction model. In the implementation process, Real-time calculation and application by the embedded controller support a closed-loop feedback mechanism to improve the accuracy and response speed of MPPT.

[0070] In one specific implementation scenario of the present application, the historical data needs to cover multiple working conditions (such as sunny days, cloudy days, and partial shading), and a training set is constructed by collecting photovoltaic system operation data for at least several months. Data preprocessing includes removing outliers and normalization, and supervised learning methods are used for training, with the goal of minimizing the error between predicted power and actual power. Model development and training are performed using Python machine learning libraries such as TensorFlow or Scikit-learn, and after training, the model parameters are fixed to the embedded system. PCA and the prediction model need to run on a computing platform with floating-point operation capability, and DSP or embedded GPU can be used. Software development can use C or Python, and model inference needs to be optimized to meet real-time requirements.

[0071] It should be noted that the system relies on the filtered signal and as input to ensure the quality of the input data for feature extraction. The predicted power directly serves the subsequent MPC step, providing a key basis for the optimized reference voltage. In addition, the inclusion of impedance improves the characterization of the system's dynamic characteristics, providing support for anomaly detection and fault correction.

[0072] In one specific implementation of the present application, the calculation formula for solving the illumination intensity is:

[0073] ;

[0074] where, is the instantaneous power, is the impedance correction factor, is the power value under standard test conditions, is the temperature value under standard test conditions, is the temperature compensation factor, used to correct the nonlinear effect of temperature on the solution of illumination intensity; the calculation formula of the impedance correction factor is:

[0075] ;

[0076] where, is the dynamic impedance influence coefficient, is a reference impedance influence coefficient, is an impedance estimation sensitivity coefficient, is a photovoltaic module impedance value under standard test conditions, is a photovoltaic module impedance at the previous time; is an interaction influence coefficient; the calculation formula of the temperature compensation factor is:

[0077]

[0078] wherein, α represents a temperature attenuation index, used to represent the nonlinear influence of temperature on the illumination intensity solution; and The solution is realized by iterative solution, and the specific steps include: initializing is the illumination intensity at the previous time or a preset initial value; based on the current , the is calculated; based on the current and , the is calculated; through iterative calculation, until is less than a preset convergence threshold or reaches a maximum number of iterations; wherein, is the initial illumination intensity in the iterative solution process, is the illumination intensity value obtained by the n-th iteration in the iterative solution process, is the impedance correction factor obtained by the n-th iteration in the iterative solution process, is the illumination intensity value obtained by the n+1-th iteration in the iterative solution process, is the absolute value of the difference between the illumination intensity values obtained by adjacent two iterations, used to judge whether the iteration converges.

[0079] It should be noted that the illumination intensity solution step aims to accurately estimate the current illumination intensity of the photovoltaic system through the operating parameters, providing key environmental information for subsequent power prediction and control optimization. The instantaneous power in the formula integrates the filtered voltage and current and the impedance correction factor, reflecting the actual operating state of the photovoltaic module. The power under standard test conditions and the temperature provide a reference, ensuring the comparability of the solution results.

[0080] Further, in the calculation formula of the impedance correction factor, the dynamic impedance influence coefficient considers the dynamic influence of impedance change, enhancing the adaptability to the aging or connection problems of the module. The temperature compensation factor corrects the nonlinear influence of temperature on the illumination solution in an exponential form, ensuring the estimation accuracy.

[0081] ​Further, since and there is mutual dependence, the solving process is realized by an iterative method. The initial value based on the illumination intensity of the last moment or the preset value, the illumination intensity is updated by iteration and , until convergence or the maximum number of iterations is reached. This iterative strategy effectively handles the nonlinear coupling between parameters and improves the robustness of estimation.

[0082] Further, the parameters are determined by experiment and simulation optimization. In the experiment, the influence of different parameter values on the accuracy of illumination solving (such as the error with the actual illumination data) can be analyzed, and the value that can minimize the error is selected. The convergence threshold and the maximum number of iterations can be set according to the accuracy requirement of the system for the estimation of the illumination intensity, and in this application, the threshold range can be adjusted according to the estimation error and the real-time demand.

[0083] In one specific implementation scenario of the application, the iterative solving needs to be realized on a platform supporting fast floating point operation, by using DSP or microcontroller, the software development uses C language, and the iterative algorithm is optimized to reduce the computational complexity.

[0084] It should be noted that the system relies on the filtered and , and the multi-dimensional parameter acquisition and to ensure the reliability of the input data. The obtained by solving is directly input into the multi-parameter coupling prediction model, which affects the accuracy of power prediction, and provides environmental information support for subsequent MPC and anomaly detection. By dynamically adjusting the impedance and temperature influence, this step enhances the adaptability of the system to complex environments (such as sudden changes in illumination or temperature fluctuations), and provides an accurate environmental parameter basis for MPPT optimization.

[0085] In one specific embodiment of the application, the reference voltage is optimized by model predictive control optimization control action, specifically, the predicted power is maximized by optimizing the objective function, and the formula is:

[0086] ;

[0087] wherein, is the optimal control action, is the prediction time length, is the predicted power at future k time; and the update formula of the reference voltage is:

[0088] ;

[0089] wherein, a reference voltage for the next time instant, a voltage adjustment based on model predictive control optimization, an optimal operating voltage for the next time instant based on model prediction, a reference voltage value for the current time instant.

[0090] It should be noted that the MPC optimization step aims to optimize the reference voltage of the photovoltaic system by dynamically adjusting the control action to achieve the accuracy and stability of MPPT. By determining the optimal control action , and by maximizing the predicted power within the prediction horizon length , the system can prospectively evaluate the impact of different control actions on future power output, thereby selecting the optimal control strategy. The predicted power is based on the principal component feature vector and the current control action , taking into account the coupling effects of multi-dimensional parameters.

[0091] Further, in the reference voltage update through the formula, the voltage adjustment amount is dynamically adjusted according to the optimal operating voltage predicted by the model . The current reference voltage . This update mechanism ensures that the system gradually approaches the maximum power point, while avoiding the risk of traditional MPPT methods (such as perturbation and observation method) falling into local extremum in complex environment. The optimization process of MPC can use numerical optimization algorithms (such as gradient descent or dynamic programming), ensuring computational efficiency and real-time performance.

[0092] In a specific implementation scenario of the present application, the prediction horizon length H is determined by simulation optimization, balancing prediction accuracy and computational complexity, and is usually selected based on the dynamic response time of the system. MPC needs to run on a high-performance computing platform, which can use DSP or FPGA, and software development uses C or MATLAB, and optimization algorithms meet real-time control requirements.

[0093] It should be noted that the system relies on the provided by the power prediction step and the generated by principal component analysis to ensure that the optimization is based on high-quality prediction data. The intensity of light indirectly affects the optimization process through the prediction model, enhancing the system's adaptability to environmental changes. The optimized control action and the updated reference voltage are directly applied to the photovoltaic inverter or control circuit, providing a stable operating state reference for subsequent anomaly detection and fault correction steps.

[0094] In one specific embodiment of the present invention, predictive fault correction is performed by adjusting the multi-parameter anomaly detection weights based on adaptive fuzzy logic. Specifically, anomaly indicators are calculated to detect the system operating status, and the formula is: ;

[0095] in, This represents the temperature change value. This represents the impedance change value. These are weighting coefficients for anomaly detection, used to balance the contributions of power, temperature, and impedance changes to anomaly indicators; The adaptive adjustment is achieved through a fuzzy logic controller. The fuzzy logic controller takes the rate of change of light intensity, the rate of change of temperature, and the current trend of abnormal indicators as inputs, and outputs the corresponding weight adjustment amounts. The fuzzy logic controller achieves adaptive adjustment through the following steps: defining fuzzy sets and membership functions for the input and output variables, and establishing a fuzzy rule base; performing fuzzy inference based on the fuzzification result of the current system state and the fuzzy rule base; and defuzzifying the fuzzy inference result into clear weight adjustment amounts. Update weights to ,in, This represents the anomaly detection weight value from the previous time step. X represents the weight adjustment calculated by the adaptive fuzzy logic controller at the current moment, where X is the index corresponding to P, T, and Z. Correction is triggered when an abnormal indicator exceeds a preset threshold; the correction formula is: ;

[0096] in, This is the reference voltage that the system uses for fault correction at the next moment after an anomaly occurs and triggers correction. This is the reference voltage at the current moment; This represents the maximum power point voltage value at the previous moment. These are the maximum power point voltage values ​​from the previous two moments. The correction weighting coefficient is used to adjust the correction magnitude.

[0097] It should be noted that the adaptive fuzzy logic adjustment and predictive fault correction step aims to ensure the stability and reliability of MPPT by dynamically monitoring the photovoltaic system's operating status, identifying anomalies, and performing corrections. This step comprehensively evaluates the rate of change of power, temperature, and impedance using an anomaly index formula to detect abnormalities in the system's operating status in real time. The anomaly index integrates changes in multiple parameters using a weighted approach, with weights... Dynamic adjustments ensure the system can flexibly respond to differences in the contribution of various parameters to anomaly detection. Example rules of fuzzy logic include: if the rate of change in illumination is high and the anomaly index trend is upward, then the power weight is increased. .

[0098] Further, the weight adjustment is realized by a fuzzy logic controller, which takes the illumination intensity change rate , temperature change rate and abnormal index trend as inputs, and generates the weight adjustment amount through fuzzification, fuzzy rule inference and defuzzification. The fuzzy rule base is designed based on expert experience or historical data, which can capture the nonlinear relationship between environmental changes and system state. The updated weight dynamically reflects the system running characteristics, improving the sensitivity and accuracy of anomaly detection.

[0099] Further, when the abnormal index exceeds the preset threshold, the predictive fault correction is triggered, and the reference voltage is updated by the formula . The correction process combines the historical maximum power point voltage and its trend to quickly adjust the working point to restore the system to a state close to the maximum power point, enhancing robustness. The abnormal threshold can be determined by analyzing the distribution of abnormal indicators under normal operating conditions, using statistical methods (such as mean plus multiple of standard deviation).

[0100] In one specific implementation scenario of the present application, the initial value of the weight coefficient and is optimized through experiments, based on the sensitivity of anomaly detection and correction effect. The fuzzy rule base is designed through historical data analysis or expert consultation to ensure the accuracy of inference. The fuzzy logic controller can be implemented on an embedded microcontroller, and can be developed using C language, with the rule base stored in non-volatile memory.

[0101] It should be noted that the system relies on multi-dimensional parameter acquisition , illumination solution and filtered power data to ensure the reliability of abnormal detection input. The corrected reference voltage coordinates with the MPC step to optimize subsequent control actions, while providing feedback to the system to enhance overall stability. Through dynamic adjustment of fuzzy logic and predictive correction, this step effectively deals with abnormal situations under complex environments, improving the anti-interference ability and running efficiency of the MPPT system.

[0102] In a specific implementation scenario of the present application, to verify the effect of the technical solution, simulation and experimental testing are carried out in the laboratory and actual photovoltaic systems. The simulation adopts the MATLAB / Simulink platform, a model is constructed based on typical light and temperature variation scenarios (sunny day, cloudy day, partial shading), and the MPPT efficiency, response speed and abnormal detection accuracy are verified. The experiment is implemented on a small-scale photovoltaic array (single component or multiple components in series), high-precision sensors and DSP controllers are used, and the test results show that the system can effectively track the maximum power point and quickly correct abnormalities in a dynamic environment.

[0103] Hardware configuration: high-precision sensors and DSP controllers or ARM Cortex-M microcontrollers can be used, which support real-time floating-point operations and fast data sampling. The storage unit needs to support historical data storage (at least several months of data) to train the model. Preferably, a processor that supports real-time floating-point operations can be used, such as a hardware platform based on a digital signal processor or a microcontroller.

[0104] Software implementation: algorithm development is recommended to use C language or MATLAB, and the prediction model and PCA can be solidified to the embedded system after training on the PC. Real-time control software needs to be optimized to meet the millisecond-level response requirement, and RTOS can be used to support multi-task processing. Preferably, a programming language that supports real-time control can be used for development, and a data analysis tool can be used for model training.

[0105] In the second embodiment of the present application, the present application provides a photovoltaic module adaptive MPPT optimization system, which comprises a collection module, a prediction module and a correction module; the collection module is used to collect multi-dimensional parameter information of the photovoltaic module through adaptive weight distribution, generate a weighted original signal, and generate a filtered signal through dynamic filtering processing; the prediction module is used to generate a principal component feature vector through principal component analysis based on the filtered signal, calculate the intensity of light, and calculate the predicted power using a multi-parameter coupling prediction model; the correction module is used to optimize the reference voltage of the control action through model predictive control based on the predicted power, adjust the multi-parameter abnormal detection weight based on adaptive fuzzy logic, and perform predictive fault correction to realize maximum power point tracking.

[0106] In the third embodiment of the present application, the present application provides an electronic device comprising a memory and a processor, characterized in that the memory stores a computer program executable on the processor, and the processor executes the program to implement the steps of the photovoltaic module adaptive MPPT optimization method as described above.

[0107] In the fourth embodiment of the present application, the present application provides a storage medium storing a computer program, wherein the computer program is executed by a processor to implement the steps of the adaptive MPPT optimization method for photovoltaic modules.

[0108] In summary, the adaptive MPPT optimization method and system for photovoltaic modules provided by the present application effectively reduces noise interference and improves the stability of multi-dimensional parameter processing through adaptive weight distribution and dynamic Gaussian filtering; accurately predicts power output by combining principal component analysis and a multi-parameter coupling prediction model; realizes rapid and accurate maximum power point tracking by optimizing the reference voltage through model predictive control; and enhances abnormality detection and fault correction capabilities and improves system robustness through adaptive fuzzy logic. Compared with traditional methods, the present application significantly improves power tracking efficiency and system reliability in complex environments and is suitable for distributed photovoltaic and microgrid applications.

[0109] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described modules can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0110] The modules described as separate components can or can not be physically separated, and the components shown as modules can or can not be physical modules, i.e., they can be located in one place or distributed to multiple network modules. Part or all of the modules can be selected to achieve the purpose of the present embodiment scheme according to actual needs.

[0111] In addition, the functional modules in each embodiment of the present application can be integrated in one processing module, or each module can exist physically alone, or two or more modules can be integrated in one module. The integrated module can be realized in the form of hardware or in the form of hardware plus software function module.

[0112] The integrated module realized in the form of software function module can be stored in a computer-readable storage medium. The software function module is stored in a storage medium, including a plurality of instructions for causing a computer system (which can be a personal computer, a server, or a network system, etc.) or a processor to execute part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes various storage medium capable of storing program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0113] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can still be modified, or some technical features therein can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A photovoltaic module adaptive MPPT optimization method, characterized in that: The method comprises the following steps: The multi-dimensional parameter information of the photovoltaic module is collected through adaptive weight allocation to generate a weighted original signal, and a filtered signal is generated through dynamic filtering processing; Based on the filtered signal, a principal component feature vector is generated through principal component analysis, the illumination intensity is solved, and the predicted power is calculated using a multi-parameter coupling prediction model; Based on the predicted power, the optimal reference voltage of the control optimization control action is predicted through the multi-parameter coupling prediction model, the multi-parameter abnormality detection weight is adjusted based on adaptive fuzzy logic, and predictive fault correction is performed to achieve maximum power point tracking; The optimal reference voltage of the control optimization control action is predicted through the multi-parameter coupling prediction model, which comprises: The predicted power is maximized by optimizing the objective function, and the formula is: ； wherein, is the current control action, is the principal component feature vector at the current time, t is the time variable, is the optimal control action, is the prediction horizon length, is the predicted power at future k time. The update formula of the reference voltage is: ； wherein, is a reference voltage for the next time instant, is a voltage adjustment based on model predictive control optimization, is a model predictive based optimal operating voltage for the next time instant, is a reference voltage value for the current time instant.

2. The photovoltaic module adaptive MPPT optimization method according to claim 1, wherein: The multi-dimensional parameter information of the photovoltaic module is collected through adaptive weight allocation, which comprises: The voltage, current, temperature and impedance of the photovoltaic module are collected, and the weighted original signal is generated through adaptive weight allocation, and the formula is: ； wherein, is a weighted original signal, is a voltage, is a current, is a temperature, is an impedance, t is a time variable, used to represent the value of the multi-dimensional parameter of the photovoltaic module at a certain time, is a corresponding weight, dynamically adjusted based on the parameter change rate, The adjustment formula of the weight is: ； wherein, are corresponding weights; denotes , , , , is the value of the i-th parameter at the previous time; is a weight adjustment sensitivity coefficient, used to adjust the response degree of the weight to the parameter change; j is an index, traversing the four parameters of voltage, current, temperature and impedance, ranging from 1 to 4.

3. The photovoltaic module adaptive MPPT optimization method of claim 2, wherein: The filtered signal generated through dynamic filtering processing is used to reduce noise and enhance the stability of maximum power point tracking; The dynamic filtering processing adopts dynamic Gaussian filtering, and the formula is: ； wherein, is a filtered signal generated after dynamic Gaussian filtering processing at time t; n is an index of a certain time at a sampling window, and N is a sampling window size; is a sampling time interval; is a dynamic variance, which is adjusted based on the light intensity and the temperature change rate, and the formula is: ； wherein, is the reference variance; is the light intensity, the initial value is the calibration value, is the light intensity at the previous moment, is the standard light intensity; is the temperature change weight coefficient, used to balance the influence of temperature on filtering; is the temperature at the previous moment.

4. The photovoltaic module adaptive MPPT optimization method of claim 3, wherein: The principal component feature vector is generated through principal component analysis, the illumination intensity is solved, and the predicted power is calculated using a multi-parameter coupling prediction model, which comprises: The principal component feature vector is represented as: ； wherein, is a principal component feature vector of the current time instant, is a principal component matrix trained by offline analysis of historical photovoltaic system operation data, is a voltage and current based on the filtered signal; The calculation formula of the predicted power is: ； wherein, is the predicted power for the next time instant, is the current control action, is a multi-parameter coupling prediction function trained by learning historical photovoltaic system operation data and corresponding control actions, capable of capturing the nonlinear relationship between multiple variables to predict the future power output of the system.

5. The photovoltaic module adaptive MPPT optimization method of claim 4, wherein: The calculation formula of the illumination intensity is: ； wherein, P is the instantaneous power, Z is the impedance correction factor, Pstd is the power value under standard test conditions, Tstd is the temperature value under standard test conditions, Tc is the temperature compensation factor, used to correct the non-linear effect of temperature on the illumination intensity solution; The calculation formula of the impedance correction factor is: ； wherein, is a dynamic impedance influence coefficient, is a reference impedance influence coefficient, is an impedance estimation sensitivity coefficient, is a photovoltaic module impedance value under standard test conditions, is a photovoltaic module impedance at the previous time instant; is an interaction influence coefficient; The calculation formula of the temperature compensation factor is: ； Wherein, α represents the temperature attenuation index, which is used to represent the nonlinear influence of temperature on the solution of illumination intensity; and The solution of the equations is achieved by iterative solving, the specific steps including: Initialization the light intensity of the previous time or a preset initial value; based on the current , calculate ; Based on the current and , calculate ; by iterative computation until less than a preset convergence threshold or a maximum number of iterations is reached. wherein, is the initial value of the illumination intensity in the iterative solving process, is the illumination intensity value obtained in the n-th iteration in the iterative solving process, is the impedance correction factor obtained in the n-th iteration in the iterative solving process, is the illumination intensity value obtained in the n+1-th iteration in the iterative solving process, is the absolute value of the difference between the illumination intensity values obtained in two adjacent iterations, used to determine whether the iteration converges.

6. The photovoltaic module adaptive MPPT optimization method of claim 1, wherein: The multi-parameter abnormality detection weight is adjusted based on adaptive fuzzy logic, and predictive fault correction is performed, which comprises: An abnormality index is calculated to detect the system operating state, and the formula is: ； wherein, is a temperature change value, is an impedance change value, is a weight coefficient for anomaly detection, for balancing the contribution of power, temperature and impedance change to the anomaly index; The The adaptive adjustment is achieved through a fuzzy logic controller, which takes the rate of change of light intensity, the rate of change of temperature, and the current trend of abnormal indicators as inputs, and outputs the corresponding weighted adjustment amount. The fuzzy logic controller achieves adaptive adjustment through the following steps: The fuzzy set and membership function of the input variable and output variable are defined, and a fuzzy rule base is established; Fuzzy reasoning is performed according to the fuzzification result of the current system state and the fuzzy rule base; Defuzzifying the fuzzy inference result into a crisp weight adjustment amount ; The updated weight is wherein, is the abnormality detection weight value of the previous time, is the weight adjustment amount calculated by the adaptive fuzzy logic controller at the current time, and X is an index corresponding to P, T, and Z. When the abnormality index exceeds the preset threshold, the correction is triggered, and the correction formula is: ； wherein, is the reference voltage used by the system for fault correction at the next time instant after an anomaly has occurred and triggered a correction, is the reference voltage at the current time instant; is the maximum power point voltage value at the previous time instant, is the maximum power point voltage value at the previous two time instants, is a correction weight coefficient used to adjust the correction amplitude.

7. A photovoltaic module adaptive MPPT optimization system applied to the photovoltaic module adaptive MPPT optimization method of claim 1, characterized in that: It comprises a collection module, a prediction module and a correction module; The collection module is used to collect the multi-dimensional parameter information of the photovoltaic module through adaptive weight allocation to generate a weighted original signal, and a filtered signal is generated through dynamic filtering processing; The prediction module is used to generate a principal component feature vector through principal component analysis based on the filtered signal, solve the illumination intensity, and calculate the predicted power using a multi-parameter coupling prediction model; The correction module is used to predict the optimal reference voltage of the control optimization control action through the multi-parameter coupling prediction model based on the predicted power, adjust the multi-parameter abnormality detection weight based on adaptive fuzzy logic, and perform predictive fault correction to achieve maximum power point tracking.

8. An electronic device comprising a memory and a processor, characterized in that: The memory stores a computer program that can run on the processor, and the program running on the processor implements the steps of the photovoltaic module adaptive MPPT optimization method in any one of claims 1-6.

9. A storage medium, storing a computer program, characterized in that: The computer program is executed by the processor to implement the steps of the photovoltaic module adaptive MPPT optimization method in any one of claims 1-6.

Images