A method and system for predicting the lifetime of a photovoltaic module
By using dynamic Gamma process and Gaussian process regression interpolation, combined with sliding window maximum likelihood estimation, the data sparsity problem in photovoltaic module lifetime prediction is solved, enabling accurate prediction of photovoltaic module lifetime and adapting to the degradation characteristics throughout its entire life cycle.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- LANZHOU UNIVERSITY OF TECHNOLOGY
- Filing Date
- 2025-07-02
- Publication Date
- 2026-05-22
AI Technical Summary
Existing technologies struggle to accurately predict the remaining lifespan of photovoltaic modules throughout their entire lifecycle, especially in the absence of long-term continuous data, and existing models are not adapted to the degradation characteristics of photovoltaic modules.
A lifetime prediction method based on dynamic Gamma process is adopted, which combines Gaussian process regression (GPR) interpolation and sliding window maximum likelihood estimation to update the remaining lifetime distribution model parameters of photovoltaic modules in real time, thus adapting to the problem of sparse or missing degradation data of photovoltaic modules.
It enables accurate prediction of the remaining lifetime of photovoltaic modules in the absence of long-term data, improves the flexibility and accuracy of prediction, adapts to multi-scale degradation behavior, and enhances the adaptability and prediction accuracy of the model.
Smart Images
Figure CN120409056B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of photovoltaic module lifespan prediction technology, and more specifically to a method and system for predicting the lifespan of photovoltaic modules. Background Technology
[0002] As the core component of a photovoltaic (PV) power generation system, PV modules experience a gradual performance degradation as their service life increases and random factors influence their operation. This accumulated degradation affects the reliability of PV power generation. Furthermore, due to the relatively low degradation rate of PV modules, it is difficult to collect long-term data to confirm the degradation path and lifetime. Therefore, it is necessary to establish a stochastic degradation model to characterize the unstable and fuzzy characteristics of PV module performance degradation over time, in order to estimate the remaining lifetime of PV modules and improve their operational reliability.
[0003] Currently, the methods for predicting the remaining lifespan of photovoltaic modules based on data modeling are basically to establish a performance degradation model of photovoltaic modules using stochastic processes. This method first requires sufficient degradation data, and considering some uncontrollable factors, it is difficult to obtain continuous and complete degradation data in actual engineering. Secondly, the model selected to describe the degradation process of the module is not suitable for the entire life cycle of the photovoltaic module.
[0004] Therefore, how to solve the above-mentioned technical problems is an urgent issue that needs to be addressed by those skilled in the art. Summary of the Invention
[0005] In view of this, in order to at least partially solve the above-mentioned technical problems, the present invention provides a method and system for predicting the lifetime of photovoltaic modules. This method can not only adapt to the entire life cycle of photovoltaic modules, but also overcome the problem of sparse or missing degradation data in the actual operation of photovoltaic modules.
[0006] To achieve the above objectives, the present invention adopts the following technical solution:
[0007] In a first aspect, the present invention provides a method for predicting the lifetime of photovoltaic modules, comprising the following steps:
[0008] Based on the dynamic Gamma process degradation model of photovoltaic modules, the remaining lifetime distribution function of photovoltaic modules at the time of test is determined;
[0009] Acquire photovoltaic module degradation data and perform interpolation using GPR;
[0010] Based on the interpolated photovoltaic module degradation data, the parameters of the remaining lifetime distribution function of the photovoltaic module at the time of test are updated in real time by combining the sliding window and the maximum likelihood estimation method.
[0011] The remaining lifetime distribution model with updated parameters is used to predict the lifetime of photovoltaic modules.
[0012] Secondly, this application provides a photovoltaic module lifetime prediction system, which applies any of the photovoltaic module lifetime prediction methods described above, including:
[0013] The remaining lifetime distribution function determination module is used to determine the remaining lifetime distribution function of the photovoltaic module at the time of test based on the dynamic Gamma process degradation model of the photovoltaic module.
[0014] The degradation data acquisition and processing module is used to acquire photovoltaic module degradation data and perform interpolation processing using GPR;
[0015] The lifetime distribution function parameter update module is used to update the parameters of the remaining lifetime distribution function of the photovoltaic module at the time of test in real time based on the interpolated photovoltaic module degradation data, combined with the sliding window and maximum likelihood estimation method.
[0016] The remaining lifetime prediction module uses the updated remaining lifetime distribution model to predict the lifetime of photovoltaic modules.
[0017] As can be seen from the above technical solution, this invention provides a photovoltaic module lifetime prediction method and system by integrating GPR and dynamic Gamma processes. This method balances prediction accuracy and real-time performance, adapting to the entire lifecycle of photovoltaic modules. It effectively addresses the shortcomings of current data modeling methods, reduces the limitations of current photovoltaic module lifetime prediction, and makes the lifetime prediction method more suitable for practical applications. Compared with existing technologies, the advantages of this application specifically include:
[0018] 1) This application constructs a remaining lifetime prediction model for the time to be predicted based on the dynamic Gamma process, which can still achieve relatively accurate remaining lifetime estimation in scenarios where long-term continuous data is lacking.
[0019] 2) Interpolate degraded data based on GPR, and perform interpolation based on the anisotropic squared exponent (RBF) kernel and... The kernel constructs a covariance function to ensure interpolation smoothness while retaining abrupt changes in the degradation process, thereby enhancing prediction flexibility, improving interpolation accuracy, adapting to multi-scale degradation behavior, and facilitating marginal likelihood optimization learning.
[0020] 3) Based on the degradation trajectory of photovoltaic modules, the parameters of the Gamma degradation model are updated in real time by combining the sliding window and maximum likelihood estimation methods. On this basis, the remaining lifetime of photovoltaic modules is predicted in real time. The sliding window mechanism improves the sensitivity of the model to short-term dynamic changes, and the maximum likelihood estimation achieves local optimal parameter estimation within each window, thereby realizing real-time dynamic updating of parameters and enhancing the tracking modeling capability. This improves the adaptability and prediction accuracy of the model in non-stationary degradation scenarios.
[0021] Other features and advantages of the invention will be set forth in the following description, and the objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims and drawings.
[0022] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description
[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0024] Figure 1 This is a flowchart of the photovoltaic module lifetime prediction method of the present invention;
[0025] Figure 2 This is an example diagram of the photovoltaic module lifetime prediction process of the present invention;
[0026] Figure 3 A schematic diagram of the hardware components for applying the photovoltaic module lifetime prediction method of the present invention;
[0027] Figure 4(a) is a schematic diagram of the process of handling the cumulative power degradation of photovoltaic modules in the first monitoring of the present invention;
[0028] Figure 4(b) is a schematic diagram of the process for handling the cumulative power degradation of photovoltaic modules in the second monitoring of the present invention;
[0029] Figure 4(c) is a schematic diagram of the process for handling the cumulative power degradation of photovoltaic modules in the third monitoring of this invention;
[0030] Figure 5 A schematic diagram of the fitting test provided for an example of the present invention;
[0031] Figure 6 A diagram illustrating the real-time update process of degradation model parameters provided in this invention example;
[0032] Figure 7 The real-time monitoring process and results of photovoltaic module degradation provided in this invention example;
[0033] Figure 8 The probability distribution of photovoltaic module lifetime failure time is provided for an example of the present invention. Detailed Implementation
[0034] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0035] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0036] In one embodiment, the present invention discloses a method for predicting the lifetime of photovoltaic modules, such as... Figure 1 The steps include:
[0037] Based on the dynamic Gamma process degradation model of photovoltaic modules, the remaining lifetime distribution function of photovoltaic modules at the time of test is determined;
[0038] Acquire photovoltaic module degradation data and perform interpolation using GPR;
[0039] Based on the interpolated photovoltaic module degradation data, the parameters of the remaining lifetime distribution function of the photovoltaic module at the time of test are updated in real time by combining the sliding window and the maximum likelihood estimation method.
[0040] The remaining lifetime distribution model with updated parameters is used to predict the lifetime of photovoltaic modules.
[0041] In this embodiment, the construction of lifetime distribution and the acquisition of component degradation data can be performed simultaneously, or the data can be acquired first and then the lifetime distribution function can be constructed.
[0042] In one embodiment, the remaining lifetime distribution function of the photovoltaic module at the time of measurement is first determined based on the dynamic Gamma process degradation model of the photovoltaic module.
[0043] This embodiment establishes a function related to the equipment performance degradation process based on the standard Gamma process. When the degradation during equipment service is described as... The degradation model based on the Gamma process is shown in the following equation:
[0044]
[0045] in, The value represents the degradation increment of the Gamma process; k represents the shape parameter. The scale parameter is represented by 'a', which represents the time variable used to describe the service life of the photovoltaic module, and 'a' ≥ 0.
[0046] The Gamma process assumes that the degradation process of photovoltaic modules is strictly monotonic, making it suitable for characterizing the monotonic degradation trend of equipment failure caused by the continuous accumulation of degradation. This application selects cumulative power degradation as a characteristic quantity to describe the performance degradation of photovoltaic modules, which can well explain the mechanism of module service life changes over time. Specifically, the cumulative power degradation is calculated based on the initial output power of the target photovoltaic module and its output power at the current moment.
[0047] The cumulative power degradation of photovoltaic (PV) modules generally increases over time during their service life. Furthermore, the data preprocessing considers the temporal uncertainty of the degradation process due to the combined effects of various factors, including natural environment and human activity. Generally, a cumulative power degradation of 20% of the initial power is used as the threshold for module failure. However, in actual prediction, this failure threshold can be adjusted based on the specific engineering requirements of the module's service life.
[0048] Furthermore, the performance failure of photovoltaic modules is defined as first-pass failure. Once the degradation process X(t) of the module reaches the failure threshold ξ, it can be considered to have failed. The failure time is defined as the moment when the sample degradation trajectory of X(t) first exceeds ξ, and the lifetime T is the length of time from when the equipment is first put into use and first reaches the failure threshold. Thus, the following equation can be obtained:
[0049]
[0050] t represents the time T corresponding to the first time the cumulative power degradation reaches the failure threshold, and s represents any time before time t.
[0051] When X(t) is a Gamma process, based on the monotonicity of the degradation process sample trajectory, the lifetime distribution function is:
[0052]
[0053] In the formula, Let ξ represent the probability that the lifetime random variable T is less than or equal to a certain time point t, k(t) represent the shape parameter of the dynamic Gamma process at time t, ξ represent the failure threshold, and k represent the shape parameter. The scale parameter is represented by γ(k, x), which represents the upper incomplete Gamma function, defined as:
[0054]
[0055] In the formula, k is the shape parameter of the Gamma process, which reflects the degree of acceleration of the degradation process. The larger the value, the faster the degradation. x is the normalized time, which reflects the relative degradation time remaining before the current component fails. It is related to the current degree of degradation and the scale parameter.
[0056] Furthermore, based on the measured time t0, the remaining lifetime distribution function of the photovoltaic module is determined. Using the remaining lifetime distribution function constructed from the dynamic Gamma process model, and combined with the real-time estimation results of the model parameters, the expected lifetime value is further derived.
[0057] If a component has been operating up to the current time t0, its remaining lifetime distribution is calculated based on its lifetime distribution function:
[0058]
[0059] In the formula, This represents the cumulative probability that a photovoltaic module will fail before time t. This represents the cumulative probability that the component will fail before the current time t0. This represents the conditional probability that a known component will fail before time t, provided it has not yet failed at time t0.
[0060] For the performance degradation process described by the Gamma process, a distribution function for the time when the degradation process first reaches the failure threshold ξ can be established based on the component's performance degradation value x0 = X(t0) < ξ at time t0 and the incremental independence characteristic of the Gamma process:
[0061]
[0062] When t0=0, the remaining lifespan of the photovoltaic module is equivalent to its service life.
[0063] After obtaining the aforementioned photovoltaic module remaining lifetime prediction model, the target photovoltaic module is predicted using this model. To address the sparsity and fragmentation issues of the input degradation data, this invention proposes to adaptively update the model parameters based on the processed data of the cumulative power degradation of the target photovoltaic module at the current moment. This ensures that the entire remaining lifetime prediction process is based on the degradation trajectory of the target module itself, and that the influence of random effects on the module degradation process is considered during the data preprocessing stage.
[0064] In one embodiment, photovoltaic module degradation data is acquired, and for missing or sparse degradation data, GPR is used for interpolation.
[0065] The core idea of GPR is to estimate the distribution of missing data by constructing a covariance function. For a given dataset of the following form... It can be used for any new point in time. Predict its degradation metric X( ):
[0066]
[0067] In the formula, This indicates the new time point of the interpolation. Indicates a point in time The corresponding degradation data, where D represents the photovoltaic module degradation data. Representative New Point The covariance function of the degraded data D, where C represents the covariance matrix of the degraded data, and y represents the degraded value of the degraded data. , Represents the nth time point The corresponding degradation value, This represents the sampling time of the nth training data point, where n represents the total number of training data points.
[0068] In this embodiment, to ensure that the interpolation can smooth the global trend while preserving local abrupt changes, the present invention preferably combines the anisotropic squared exponent (RBF) kernel with... Construct the covariance function by weighted superposition of kernels:
[0069]
[0070] In the formula, Let represent two arbitrary time input points belonging to the training sample time set, used to construct the kernel function in Gaussian process regression. For amplitude hyperparameters, Using the feature length scale, adaptive learning is performed by maximizing the marginal log-likelihood to achieve optimal interpolation performance across the entire sample space.
[0071] Maximizing the marginal log-likelihood is:
[0072]
[0073] In the formula, K represents the covariance matrix constructed from the training data, whose elements are the kernel function values between each sampling point, and n represents the number of training data points, i.e., the total number of observed degenerate data points. This represents a logarithmic function with the natural constant e as its base.
[0074]
[0075] In the formula, The magnitude hyperparameter of the radial basis function (RBF) kernel is used to control the overall contribution of the kernel function to the covariance matrix. This represents the characteristic length scale of the radial basis function kernel, used to measure the range of influence of the similarity between data on the covariance. express The magnitude hyperparameter of the kernel represents its contribution intensity; express The length scale of the kernel is used to control the smoothness and sensing range of the kernel function. The standard deviation of the noise term reflects the intensity of observation error in degraded data and is often added to the diagonal as the "noise kernel" in the covariance function.
[0076] In this way, GPR can interpolate missing data while smoothing out the effects of noise.
[0077] Furthermore, based on this, this application provides a closed-form expression for the prediction variance to quantify the uncertainty of the interpolation results. The calculation formula is as follows:
[0078]
[0079] In the formula, For the covariance of the new sample itself, it can be seen from the above formula that if If there is a lack of nearby observation points or the sample noise is high, the prediction variance will increase; conversely, it will decrease. This provides a theoretical basis for the subsequent construction of confidence intervals and active point supplementation.
[0080] In one embodiment, a combined sliding window and maximum likelihood estimation method are used to adaptively update the parameters of the remaining lifetime distribution function of the photovoltaic module at the time of test in real time.
[0081] Sliding window fitting, as a local fitting method, divides the dataset into segments and fits the target distribution parameters separately within each window, thereby obtaining the dynamic change pattern of the data over time or other variables. It is characterized by its simplicity, strong dynamism, and high robustness. This application uses the sliding window method to dynamically fit the cumulative power degradation data of photovoltaic modules to obtain a better understanding of the degradation trend.
[0082] Furthermore, for the data within the sliding window, maximum likelihood estimation is used to estimate the parameters k and y of the Gamma process. ,include:
[0083] Determine the maximum likelihood function for the data within the sliding window;
[0084] Taking the logarithm yields the log-likelihood function;
[0085] Maximize the log-likelihood function to update parameters in real time.
[0086] Assuming the degradation data within the time window is The weight of each data point is w(t)i If the maximum likelihood function is... for:
[0087]
[0088] In the formula, m is the number of data items in the sliding window, and i is the index of the data items in the sliding window. This represents the difference between adjacent degenerate data within the sliding window. , This represents the weight of the i-th data point. Let k and k represent the sampling time of the i-th degenerate data point within the sliding window. These represent the shape parameter and scale parameter in the remaining lifetime distribution function, respectively.
[0089] Taking the logarithm yields the log-likelihood function. By maximizing the log-likelihood function, we can obtain k and... The estimated value.
[0090] In an alternative embodiment, for parameters k and The estimate, by taking the partial derivatives of each and setting them to zero, yields the following equation:
[0091]
[0092] In the formula: Let represent a logarithmic function with base e. ; Represents the Digamma function, which is defined as: .
[0093] By solving the above equations, the parameter estimates of the degradation process at the current time t0 can be obtained. and .
[0094] Dynamically updated using maximum likelihood estimation and It can be used to input the original model to achieve the goal of real-time prediction of the remaining lifespan of photovoltaic modules.
[0095] The expected value and variance of the remaining lifetime R(t0) are:
[0096]
[0097]
[0098] The above two equations, used to quantify the statistical characteristics of the remaining lifetime of the component at the current time t0, are standard derivations based on the dynamic Gamma process. Their function is as follows:
[0099] E represents the expected remaining lifespan of the photovoltaic module, starting from the current time t0 and continuing to operate until failure.
[0100] Var represents the magnitude of the uncertainty in the remaining life prediction, i.e., the range of fluctuation in the confidence interval.
[0101] These two indicators provide quantitative support for maintenance decisions and can be used to set the optimal preventive maintenance time window or adjust thresholds in intelligent operation and maintenance strategies.
[0102] In one exemplary embodiment, the overall photovoltaic module lifetime prediction process refers to Figure 2 ,include:
[0103] S1. Input degradation data;
[0104] S2 and GPR interpolation processing; Gaussian smoothing to remove noise;
[0105] S3. Whether the degradation increment verification is passed. If not, return to re-perform the interpolation process; if yes, execute S4. In this embodiment, the degradation increment verification is to check the temporal monotonicity and physical consistency of the degradation sequence obtained by GPR regression interpolation. This includes eliminating or correcting non-physical jumps or regressions caused by overfitting, edge effects, etc., in the interpolation to ensure that the interpolation data conforms to the mathematical assumptions of Gamma process modeling.
[0106] S4. Set up a sliding window and initialize the Gamma distribution parameters;
[0107] S5, MLE updates parameters in real time, checks if new data has been added, if so, returns to S2 to re-enter for interpolation processing; otherwise, executes S6.
[0108] S6. Output the latest model parameters;
[0109] S7. Perform real-time lifetime prediction.
[0110] To facilitate real-time lifetime prediction of this application, one embodiment discloses a photovoltaic module lifetime prediction system. This system applies any of the photovoltaic module lifetime prediction methods described above, including:
[0111] The remaining lifetime distribution function determination module is used to determine the remaining lifetime distribution function of the photovoltaic module at the time of test based on the dynamic Gamma process degradation model of the photovoltaic module; the degradation data acquisition and processing module is used to acquire the degradation data of the photovoltaic module and perform interpolation processing using GPR; the lifetime distribution function parameter update module is used to update the parameters of the remaining lifetime distribution function of the photovoltaic module at the time of test in real time based on the interpolated photovoltaic module degradation data, combined with the sliding window and maximum likelihood estimation method; the remaining lifetime prediction module uses the parameter-updated remaining lifetime distribution model to predict the lifetime of the photovoltaic module.
[0112] Preferably, a transmission module is also included for transmitting the lifetime prediction results of the photovoltaic module at the time of test to the cloud platform.
[0113] Alternatively, a memory may be provided, internally equipped with a processor, which executes the aforementioned photovoltaic module lifetime prediction method and transmits the prediction results to the operation and maintenance center control room via a transmitter using a wireless network (4G / 5G). See details below. Figure 3 .
[0114] To verify the effectiveness of this application, the following simulation experiments are conducted:
[0115] Based on the output power degradation data of photovoltaic modules in a certain photovoltaic array over 12 years, a photovoltaic module lifetime prediction model based on GPR and dynamic Gamma process is constructed, and the remaining lifetime is adaptively updated to verify the feasibility of the proposed method. The remaining lifetime is known to be 7.4 years.
[0116] To simulate the sparsity and fragmentation of data records during actual operation, the recording interval of the degradation data was artificially extended to 0.25 years. Simultaneously, prediction results from the previous 5, 6, and 7 years were used to simulate the real-time update process of lifetime prediction failure, denoted as the first monitoring, second monitoring, and third monitoring, respectively. The real-time processing and fitting results of the cumulative power degradation data at different monitoring times are shown in Figures 4(a), 4(b), and 4(c). The final monitoring data was fitted and validated, and the results are as follows: Figure 5 As shown. Figure 5 The results show that in the tests of each segment, the sample data points are close to a straight line, and the degraded data meet the data requirements of the selected model.
[0117] Subsequently, the model is used to predict the remaining lifespan of the photovoltaic module in real time. First, the model parameters are updated in real time using the module's degradation data. The parameter update process is as follows: Figure 6 As shown in the diagram, the shape parameter and scale parameter characterize the cumulative characteristics of the degradation process and the change in degradation intensity per unit time, respectively. Specifically, the shape parameter reflects the degradation rate and trend; an increase indicates accelerated degradation, reflecting the long-term stability of the system. The scale parameter describes the magnitude of degradation per unit time; an increase indicates enhanced degradation intensity, reflecting the influence of external environment or operating conditions. The dynamic adjustment of both parameters allows the model to flexibly adapt to different stages of the degradation process, accurately describing the complex degradation behavior of photovoltaic modules and providing a reliable basis for health status assessment and remaining lifetime prediction.
[0118] Based on the obtained model parameters, the real-time remaining lifetime of photovoltaic (PV) modules is predicted using the concept of first-time arrival (LTA). The failure threshold for PV modules is set at ξ=20%, meaning that the module is considered to have failed when the cumulative power degradation rate reaches 20%. Based on the real-time estimates of the model parameters and the probability density function of the remaining lifetime, the real-time monitoring results of the degradation process of the PV modules at each monitoring time point and the probability distribution of the remaining lifetime can be obtained, as shown below. Figure 7 and Figure 8 As shown. In the remaining lifetime prediction process, if t0=0, the predicted result is the actual remaining lifetime; conversely, the actual remaining lifetime of the photovoltaic module can also be obtained by combining the current monitoring time.
[0119] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.
[0120] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for predicting the lifetime of a photovoltaic module, characterized in that, include: Based on the dynamic Gamma process degradation model of photovoltaic modules, the remaining lifetime distribution function of photovoltaic modules at the time of test is determined; Photovoltaic module degradation data was acquired and interpolated using GPR; and the anisotropic squared exponent kernel was then used for interpolation. The covariance function is constructed by weighting and superimposing the kernels: In the formula, Let represent two arbitrary time input points belonging to the training sample time set, used to construct the kernel function in Gaussian process regression. Using the feature length scale, adaptive learning is performed by maximizing the edge log-likelihood. This represents the magnitude hyperparameter of the radial basis function kernel. This represents the characteristic length scale of the radial basis function kernel. express The amplitude hyperparameter of the kernel, express The length scale of the nucleus; Based on the interpolated photovoltaic module degradation data, the parameters of the remaining lifetime distribution function of the photovoltaic module at the time of test are updated in real time by combining the sliding window and the maximum likelihood estimation method. The remaining lifetime distribution model with updated parameters is used to predict the lifetime of photovoltaic modules.
2. The photovoltaic module lifetime prediction method according to claim 1, characterized in that, Based on the dynamic Gamma process degradation model of photovoltaic modules, the remaining lifetime distribution function of photovoltaic modules is determined, including: Based on the dynamic Gamma process degradation model of photovoltaic modules, the time T at which the cumulative power degradation first reaches the failure threshold is calculated; Determine the lifetime distribution function of the photovoltaic module based on time T; Combined with the time to be measured t 0, determine the distribution function of the remaining lifetime of the photovoltaic module.
3. The photovoltaic module lifetime prediction method according to claim 1 or 2, characterized in that, The expression for the remaining lifetime distribution function of a photovoltaic module is: In the formula, t This represents the time T corresponding to the moment when the cumulative power degradation first reaches the failure threshold. t 0 represents the moment of the remaining lifetime to be measured. γ ( k, x ) represents the incomplete Gamma function. ξ Indicates the failure threshold. Indicates the component at time t A performance degradation value of 0 k Represents shape parameters; Represents the scale parameter.
4. The photovoltaic module lifetime prediction method according to claim 3, characterized in that, γ ( k , x The expression is: In the formula, k The shape parameters of the Gamma process, x For normalized time.
5. The photovoltaic module lifetime prediction method according to claim 1, characterized in that, The interpolation method is as follows: In the formula, This indicates the new time point of the interpolation. Indicates a point in time The corresponding degradation data, D This indicates photovoltaic module degradation data. Representative New Point With degraded data D The covariance function, C The covariance matrix represents the degraded data, and y represents the degraded values of the degraded data. , Represents the nth time point The corresponding degradation value, This represents the sampling time of the nth training data point, where n represents the total number of training data points.
6. The photovoltaic module lifetime prediction method according to claim 1, characterized in that, Real-time parameter updates include: Determine the maximum likelihood function for the data within the sliding window; Taking the logarithm yields the log-likelihood function; Maximize the log-likelihood function to update parameters in real time.
7. The photovoltaic module lifetime prediction method according to claim 6, characterized in that, Maximum likelihood function for: In the formula, m is the number of data items in the sliding window, and i is the index of the data items in the sliding window. This represents the difference between adjacent degenerate data within the sliding window. This represents the weight of the i-th data point. Indicates the first [number]th ... i The sampling time of each degraded data point, k and These represent the shape parameter and scale parameter in the remaining lifetime distribution function, respectively.
8. A photovoltaic module lifetime prediction system, characterized in that, The photovoltaic module lifetime prediction method according to any one of claims 1-7 includes: The remaining lifetime distribution function determination module is used to determine the remaining lifetime distribution function of the photovoltaic module at the time of test based on the dynamic Gamma process degradation model of the photovoltaic module. The degradation data acquisition and processing module is used to acquire photovoltaic module degradation data and perform interpolation processing using GPR; The lifetime distribution function parameter update module is used to update the parameters of the remaining lifetime distribution function of the photovoltaic module at the time of test in real time based on the interpolated photovoltaic module degradation data, combined with the sliding window and maximum likelihood estimation method. The remaining lifetime prediction module uses the updated remaining lifetime distribution model to predict the lifetime of photovoltaic modules.
9. The photovoltaic module lifetime prediction system according to claim 8, characterized in that, It also includes a transmission module for transmitting the lifetime prediction results of the photovoltaic module at the time of testing to the cloud platform.