Intelligent photovoltaic battery health assessment system and method based on deep learning

Abstract

The present invention discloses an intelligent photovoltaic cell health assessment system and method based on deep learning, comprising the following steps: S1. obtaining a standardized data set that meets modeling requirements; S2. inputting the standardized data into a continuous-time dynamic health state modeling system based on Neural Ordinary Differential Equations; S3. performing a global search and adaptive dynamic adjustment of model parameters; S4. utilizing the optimized photovoltaic cell health state model to predict and assess the current health state of the photovoltaic cell in real time, and outputting a health state assessment result reflecting the dynamic health state of the photovoltaic cell; and S5. transmitting the health state assessment result to an assessment feedback module, which classifies and grades the health state of the photovoltaic cell according to a pre-set health state grading standard. The present invention can quantify the weights of the key influencing factors of photovoltaic cells at different time points, thereby realizing state-based self-adjustment of feature weights.

Description

Technical Field

[0001] The present invention relates to the field of battery technology, and in particular to a deep learning-based intelligent photovoltaic battery health assessment system and method. Background Art

[0002] With the rapid development of renewable energy, photovoltaic power generation, as an important component of green energy, is being widely used in industrial parks, residential buildings and large-scale photovoltaic power stations. However, due to the long-term exposure of photovoltaic cell modules to complex environments, their performance will gradually degrade or even fail. Therefore, how to timely and accurately assess the health status of photovoltaic cells to achieve preventive maintenance has become one of the key technical issues in current photovoltaic operation and maintenance management.

[0003] Existing photovoltaic cell health assessment methods can be divided into two main categories: one is based on statistical analysis of electrical performance indicators, mainly through offline data collection of voltage, current, and power to conduct trend analysis and evaluate component performance degradation; the other is through the establishment of physical models such as equivalent circuit models and thermodynamic models, combined with some real-time data to infer the internal state of the battery. It has certain reference value in engineering practice, but still has obvious shortcomings.

[0004] First, traditional statistical analysis methods rely on static data and cannot adequately characterize the temporal evolution of photovoltaic cells, making it difficult to dynamically track and predict degradation trends. Second, physical modeling methods usually require a lot of prior knowledge, with complex model parameters, long debugging cycles, and weak generalization capabilities, making it difficult to adapt to the diversity and complexity of equipment operating states in large-scale scenarios. In addition, most traditional methods use manually set threshold judgment mechanisms, which lack adaptability and make it difficult to accurately identify early degradation signals, resulting in delayed maintenance responses.

[0005] In recent years, some studies have attempted to introduce machine learning algorithms into photovoltaic health assessment. However, commonly used regression or classification models still mainly process discrete time series and fail to capture the dynamic evolution characteristics of photovoltaic cell status in the continuous time domain. In addition, the model training process lacks an efficient global optimization mechanism and is prone to falling into local optimality, affecting the assessment accuracy and stability.

[0006] In summary, existing photovoltaic cell health assessment technologies have limitations in modeling accuracy, dynamics, parameter optimization capabilities, and early degradation detection capabilities. They are unable to meet the current intelligent photovoltaic operation and maintenance system's comprehensive assessment requirements for real-time, accuracy, and foresight. There is an urgent need for a new assessment method that can simultaneously model continuous dynamics, has adaptive capabilities, and can perform global optimization to improve the intelligence level of photovoltaic cell operation and maintenance management. Summary of the Invention

[0007] One purpose of the present invention is to propose an intelligent photovoltaic battery health assessment system and method based on deep learning. The present invention can quantify the weights of key influencing factors of photovoltaic cells at different time points and realize state-based feature weight self-adjustment.

[0008] A method for evaluating the health of a photovoltaic cell based on deep learning according to an embodiment of the present invention includes the following steps:

[0009] S1. Real-time collection of raw data generated by photovoltaic cells during operation and preprocessing to obtain a standardized dataset that meets modeling requirements;

[0010] S2. Input the standardized data into a continuous-time dynamic health state modeling system based on neural ordinary differential equations. The continuous-time dynamic health state modeling system generates a photovoltaic cell health state model that describes the continuous evolution of the photovoltaic cell's internal state over time through training.

[0011] S3. Introducing a jellyfish swarm optimization algorithm during the training of the photovoltaic cell health status model, the algorithm performs a global search and adaptive dynamic adjustment of model parameters to achieve optimal settings for the model parameters and ultimately achieve a globally optimal solution for the photovoltaic cell health status model within a multidimensional parameter space.

[0012] S4. Use the optimized PV cell health model to predict and evaluate the current health status of the PV cell in real time, and output a health status assessment result reflecting the dynamic health status of the PV cell.

[0013] S5. The health status assessment results are transmitted to the assessment feedback module. The assessment feedback module classifies and grades the health status of the photovoltaic cells according to a pre-set health status grading standard and generates a corresponding assessment report and preventive maintenance recommendations.

[0014] Optionally, the S1 includes the following steps:

[0015] S11. Real-time collection of raw data generated by photovoltaic cells during operation, setting a sampling time window to collect data to form the original data set : ;

[0016] in, Indicates at time The collected PV cell operation data, Indicates at time The collected photovoltaic cell voltage, Indicates at time The collected photovoltaic cell current, Indicates at time The collected photovoltaic cell temperature, Indicates at time The light irradiance, is the total number of sampling moments;

[0017] S12. For the original dataset Data denoising is performed to remove random interference and high-frequency noise in the original data, and adjacent data interpolation is used to fill in the missing values ​​in the original data set. The data is estimated based on the average value of two adjacent valid observation points in the time series. Then, outlier removal is performed, and the valid interval range of each physical parameter is set. The observation data that is not within the valid interval range is removed. The valid interval range is set according to the normal physical state boundary of the photovoltaic cell operation. Finally, normalization is performed to convert the original physical parameter values ​​into interval values ​​according to a unified linear mapping rule. The dimensionless standard value within is constructed as a standardized data set ;

[0018] S13. Each observation data in the standardized data set consists of five dimensions: sampling time, standardized photovoltaic cell voltage, standardized photovoltaic cell current, standardized photovoltaic cell temperature, and standardized light irradiance. Each standardized data is uniformly represented as an ordered data tuple containing the five indicators of sampling time, photovoltaic cell voltage, photovoltaic cell current, photovoltaic cell temperature, and light irradiance.

[0019] Optionally, the S2 includes the following steps:

[0020] S21. Construct a photovoltaic cell health state model based on a standardized data set, and introduce a two-variable joint modeling mechanism into the neural ordinary differential equation structure. The two-variable joint modeling mechanism includes the photovoltaic cell health state variable and photovoltaic cell degradation state variables , the photovoltaic cell health state variable and photovoltaic cell degradation state variable are used to characterize the functional state and degradation trend state of the photovoltaic cell respectively, and a simultaneous differential equation model is established: ;

[0021] in, Indicates time The input feature vector of the standardized dataset, 、 The parameters are the neural network trainable parameter groups 、 The neural network function, is the dynamic attention weight coefficient vector;

[0022] S22. Define the dynamic weight adjustment mechanism of the degradation state and construct the time-adaptive dynamic attention weight coefficient vector ;

[0023] S23. Based on the simultaneous differential equation model constructed in step S21, input the initial state 、 , for continuous time intervals Perform forward integral solution to obtain the joint health status prediction sequence ,in, Indicates at time The estimated value of the photovoltaic cell functional state obtained by prediction, Indicates at time The estimated degradation trend state obtained by prediction;

[0024] S24. The predicted joint health status prediction sequence Reflects the current operating status and potential aging trend of photovoltaic cells.

[0025] Optionally, the S23 includes the following steps:

[0026] S231. Photovoltaic cell degradation state variables In the dynamic modeling process, the time-varying influence of input features on the degradation state is considered, and a time-adaptive dynamic attention weight coefficient vector is constructed. : ;

[0027] in, 、 、 、 Respectively indicate time The effects of photovoltaic cell voltage, current, temperature and light irradiance on photovoltaic cell degradation state variables The relative influence weight of

[0028] S232. Time-adaptive dynamic attention weight coefficient vector By feature importance evaluation function Dynamically generated: ;

[0029] in, Based on the neural network parameters The constructed nonlinear mapping function takes the current time as input Photovoltaic cell degradation state variables With the normalized input feature vector , the output is the unnormalized feature importance vector, which is normalized by Softmax to generate the dynamic attention weight coefficient vector ;

[0030] S233. Neural Network Mapping Function Parameter set The dynamic attention weight coefficient vector generated by end-to-end training through back-propagation mechanism in the process of photovoltaic cell health state modeling At every moment The internal parameters are updated in real time to achieve dynamic identification and weight adjustment of the impact of different characteristic parameters of photovoltaic cells on the degradation process;

[0031] S234. Dynamic attention weight coefficient vector Applied to the degenerate state differential equation in the simultaneous differential equation model, so that the degradation state variable of the photovoltaic cell The evolution process of is simultaneously affected by the normalized input feature vector and its dynamic attention weight coefficient vector A differential modeling mechanism that reflects the nonlinear degradation characteristics and characteristic sensitive dynamics of photovoltaic cells is constructed based on the combined influence of the two factors.

[0032] Optionally, S3 includes the following steps:

[0033] S31. In the training process of photovoltaic cell health status model, the jellyfish swarm optimization algorithm is introduced to optimize the parameters in the neural ordinary differential equation structure and set the optimization parameter set. ;

[0034] S32. Set the optimization parameters Encoded as the position vector of the jellyfish individual ,in , is the number of individuals in the jellyfish population, constructing the initial jellyfish population ;

[0035] S33. Based on the current population Each jellyfish The corresponding neural ordinary differential equation structure is used to calculate the health status prediction error of the neural ordinary differential equation structure on the training set and construct the loss function: ;

[0036] in, 、 Respectively represent A jellyfish individual at a time The predicted functional state and degradation state of photovoltaic cells, 、 is the real label data;

[0037] S34. Based on the behavior mechanism of the jellyfish swarm optimization algorithm, the position of each jellyfish is updated using the active motion and tidal drift mechanisms: ;

[0038] in, Indicates the The jellyfish individuals in The parameter vector for the iteration, is the optimal individual parameter vector with the smallest prediction error in the current population, is the disturbance vector simulating tidal flow, 、 To control the adjustment coefficient of active motion and drift intensity, 、 is a uniformly distributed random number;

[0039] S35. In each round of iteration, based on the loss function Sort and select all jellyfish individuals and retain the current global optimal parameter vector : ;

[0040] S36. Repeat the jellyfish swarm optimization process until the maximum number of iterations or the loss function convergence condition is met, and finally output the optimal parameter set , the optimal parameter set is applied to each module of the neural ordinary differential equation structure to construct a photovoltaic cell health status model with a global optimal solution.

[0041] Optionally, the S4 includes the following steps:

[0042] S41. Comprehensive evaluation index based on photovoltaic cell health , and calculate the health status assessment value by fusing the joint health status prediction sequence: ;

[0043] in, Indicates that the photovoltaic cell is at the time The health status assessment value, 、 are the evaluation weight coefficients of functional status and degradation status, respectively;

[0044] S42. Extract the classification standards of different health levels of photovoltaic cells from historical monitoring data, and use the health status assessment value The health level classification rules are defined as follows:

[0045] Normal state: If , indicating that the photovoltaic cell is in good operating condition and fully functional, and it is recommended to continue normal operation;

[0046] Mild degenerative state: If , indicating that the photovoltaic cells have initial aging phenomena, and it is recommended to conduct extended operation cycle testing and non-invasive optical inspection;

[0047] Moderately degraded state: If , indicating that the photovoltaic cells have experienced significant performance degradation. It is recommended to conduct distributed component testing in combination with on-site temperature data and electrical performance analysis to estimate their remaining lifespan;

[0048] Severe degeneration: If , indicating that there may be structural or material failure in the photovoltaic cell. It is recommended to immediately stop using the component and arrange manual inspection and component replacement;

[0049] S43. The health status evaluation value of the photovoltaic cell at each moment Corresponding to the health level classification standard, the output includes the health status assessment results of the current health level and health trend indicators.

[0050] Optionally, the S5 includes the following steps:

[0051] S51. Based on the photovoltaic cell sampling period in the full cycle The health status assessment trend sequence within the dataset is used to extract the health status change trajectory and calculate the degradation rate index: ;

[0052] in, Indicates the average rate of decrease of the health status assessment value of the photovoltaic cell per unit time, which is used to determine whether there is a risk of rapid degradation;

[0053] S52. Based on the classification results and degradation rate indicators Generate health status assessment reports;

[0054] S53. Generate preventive maintenance recommendations based on the combination of the health level and the degradation rate indicator.

[0055] Optionally, the health status assessment report includes the current health level, the health status trend curve of the most recent complete cycle, and the impact weights of key features.

[0056] Optionally, the preventive maintenance suggestion includes a recommended monitoring cycle, whether to enable manual inspection, and whether to perform replacement warning marking.

[0057] A deep learning-based intelligent photovoltaic battery health assessment system is used to implement a deep learning-based intelligent photovoltaic battery health assessment method, including the following modules:

[0058] Data acquisition and preprocessing module, used to collect raw data of photovoltaic cell operation and complete preprocessing to generate standardized data sets;

[0059] The health state modeling module builds a photovoltaic cell health state model based on the Neural Ordinary Differential Equation to characterize the continuous time evolution of the photovoltaic cell operating state;

[0060] The parameter optimization module performs global optimization on the parameters in the photovoltaic cell health status model based on the jellyfish swarm optimization algorithm and outputs the optimal parameter set;

[0061] The state prediction and assessment module uses the optimized photovoltaic cell health state model to predict the current state of the photovoltaic cell and generate health state assessment results;

[0062] The status feedback and decision-making module classifies and grades the health status of photovoltaic cells according to the set health status classification standards and degradation trends, and generates assessment reports and preventive maintenance recommendations.

[0063] The beneficial effects of the present invention are:

[0064] The present invention constructs a continuous-time modeling framework based on neural ordinary differential equations. By introducing a two-variable joint modeling mechanism of healthy state variables and degradation state variables, simultaneous differential equations are established to perform high-precision modeling of the evolution process of the functional state and degradation trend state of photovoltaic cells. The neural ordinary differential equations can model the state that changes continuously in time with arbitrary resolution, avoiding the limitation of fixed time steps, and improving the adaptability of the model to non-uniform sampling data. It is suitable for photovoltaic cell component systems with complex operating environments and slow state changes.

[0065] The present invention introduces the jellyfish swarm optimization algorithm into the training process of neural ordinary differential equations, and performs global search optimization on a multidimensional parameter set including health state function parameters, degradation state function parameters, and attention mechanism network parameters. By simulating the active motion and tidal drift behavior of jellyfish, efficient search is achieved in the complex loss function space, avoiding the problem of gradient descent method easily falling into local optimality, and significantly improving the stability of training convergence and the consistency of model performance.

[0066] The present invention proposes a time-adaptive dynamic attention weight mechanism. By constructing a neural network mapping function with degraded state variables and input feature vectors as input, the attention coefficient is dynamically generated at each moment. The weights of the key influencing factors of photovoltaic cells at different time points can be quantified, and the state-based feature weight self-adjustment is realized, so that the model has stronger discrimination and generalization capabilities when facing complex disturbance factors from multiple sources. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0068] Figure 1 This is a flowchart of a deep learning-based intelligent photovoltaic battery health assessment system and method proposed in the present invention. DETAILED DESCRIPTION

[0069] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0070] refer to Figure 1 , a smart photovoltaic battery health assessment method based on deep learning, including the following steps:

[0071] S1. Real-time collection of raw data generated by photovoltaic cells during operation and preprocessing to obtain a standardized dataset that meets modeling requirements;

[0072] S2. Input the standardized data into a continuous-time dynamic health state modeling system based on neural ordinary differential equations. The continuous-time dynamic health state modeling system generates a photovoltaic cell health state model that describes the continuous evolution of the photovoltaic cell's internal state over time through training.

[0073] S3. Introducing a jellyfish swarm optimization algorithm during the training of the photovoltaic cell health status model, the algorithm performs a global search and adaptive dynamic adjustment of model parameters to achieve optimal settings for the model parameters and ultimately achieve a globally optimal solution for the photovoltaic cell health status model within a multidimensional parameter space.

[0074] S4. Use the optimized PV cell health model to predict and evaluate the current health status of the PV cell in real time, and output a health status assessment result reflecting the dynamic health status of the PV cell.

[0075] S5. The health status assessment results are transmitted to the assessment feedback module. The assessment feedback module classifies and grades the health status of the photovoltaic cells according to a pre-set health status grading standard and generates a corresponding assessment report and preventive maintenance recommendations.

[0076] In this embodiment, S1 includes the following steps:

[0077] S11. Real-time collection of raw data generated by photovoltaic cells during operation, setting a sampling time window to collect data to form the original data set : ;

[0078] in, Indicates at time The collected PV cell operation data, Indicates at time The collected photovoltaic cell voltage, Indicates at time The collected photovoltaic cell current, Indicates at time The collected photovoltaic cell temperature, Indicates at time The light irradiance, is the total number of sampling moments;

[0079] S12. For the original dataset Data denoising is performed to remove random interference and high-frequency noise in the original data, and adjacent data interpolation is used to fill in the missing values ​​in the original data set. The data is estimated based on the average value of two adjacent valid observation points in the time series. Then, outlier removal is performed, and the valid interval range of each physical parameter is set. The observation data that is not within the valid interval range is removed. The valid interval range is set according to the normal physical state boundary of the photovoltaic cell operation. Finally, normalization is performed to convert the original physical parameter values ​​into interval values ​​according to a unified linear mapping rule. The dimensionless standard value within is constructed as a standardized data set ;

[0080] S13. Each observation data in the standardized data set consists of five dimensions: sampling time, standardized photovoltaic cell voltage, standardized photovoltaic cell current, standardized photovoltaic cell temperature, and standardized light irradiance. Each standardized data is uniformly represented as an ordered data tuple containing the five indicators of sampling time, photovoltaic cell voltage, photovoltaic cell current, photovoltaic cell temperature, and light irradiance.

[0081] In this embodiment, S2 includes the following steps:

[0082] S21. Construct a photovoltaic cell health state model based on a standardized data set, and introduce a two-variable joint modeling mechanism into the neural ordinary differential equation structure. The two-variable joint modeling mechanism includes the photovoltaic cell health state variable and photovoltaic cell degradation state variables , the photovoltaic cell health state variable and photovoltaic cell degradation state variable are used to characterize the functional state and degradation trend state of the photovoltaic cell respectively, and a simultaneous differential equation model is established: ;

[0083] in, Indicates time The input feature vector of the standardized dataset, 、 The parameters are the neural network trainable parameter groups 、 The neural network function, is the dynamic attention weight coefficient vector;

[0084] 、 The construction specifically includes the following steps:

[0085] Neural Network Function Used to construct photovoltaic cell health status variables The differential change relationship of the photovoltaic cell is the current health status. , degenerate state , normalized input feature vector and the current moment , the output is the rate of change of the health state variable: ;

[0086] Neural Network Function It is composed of a multi-layer feedforward neural network, and the parameter set is ,in 、 Respectively The weight matrix and bias vector of the layer, is the number of network layers;

[0087] The neural network function Used to construct photovoltaic cell degradation state variables The differential change relationship of , normalized input feature vector , dynamic attention weight coefficient vector and the current moment , the output is the rate of change of the degenerate state variable: ;

[0088] Neural Network Function It is also composed of a multi-layer feedforward neural network, and the parameter set is ,in 、 Respectively The weight matrix and bias vector of the layer, the number of network layers and the structure can be adaptively set according to the training data set.

[0089] Neural Network Function 、 In both cases, nonlinear activation functions are used for feature transformation, and activation functions are used after each layer of neurons. Processing linear output, the activation function is: ;

[0090] in, is the hyperbolic tangent function, To modify the linear unit function, the selected activation function is used to enhance the nonlinear modeling capability.

[0091] Two neural network functions 、 Collaborative training is performed through the back-propagation mechanism, and its outputs are used as differential control terms of the photovoltaic cell health state variables and degradation state variables in the neural ordinary differential equation model.

[0092] S22. Define the dynamic weight adjustment mechanism of the degradation state and construct the time-adaptive dynamic attention weight coefficient vector ;

[0093] S23. Based on the simultaneous differential equation model constructed in step S21, input the initial state 、 , for continuous time intervals Perform forward integral solution to obtain the joint health status prediction sequence ,in, Indicates at time The estimated value of the photovoltaic cell functional state obtained by prediction, Indicates at time The estimated degradation trend state obtained by prediction;

[0094] S24. The predicted joint health status prediction sequence Reflects the current operating status and potential aging trend of photovoltaic cells.

[0095] In this embodiment, S23 includes the following steps:

[0096] S231. Photovoltaic cell degradation state variables In the dynamic modeling process, the time-varying influence of input features on the degradation state is considered, and a time-adaptive dynamic attention weight coefficient vector is constructed. : ;

[0097] in, 、 、 、 Respectively indicate time The effects of photovoltaic cell voltage, current, temperature and light irradiance on photovoltaic cell degradation state variables The relative influence weight of

[0098] S232. Time-adaptive dynamic attention weight coefficient vector By feature importance evaluation function Dynamically generated: ;

[0099] in, Based on the neural network parameters The constructed nonlinear mapping function takes the current time as input Photovoltaic cell degradation state variables With the normalized input feature vector , the output is the unnormalized feature importance vector, which is normalized by Softmax to generate the dynamic attention weight coefficient vector ;

[0100] S233. Neural Network Mapping Function Parameter set The dynamic attention weight coefficient vector generated by end-to-end training through back-propagation mechanism in the process of photovoltaic cell health state modeling At every moment The internal parameters are updated in real time to achieve dynamic identification and weight adjustment of the impact of different characteristic parameters of photovoltaic cells on the degradation process;

[0101] S234. Dynamic attention weight coefficient vector Applied to the degenerate state differential equation in the simultaneous differential equation model, so that the degradation state variable of the photovoltaic cell The evolution process of is simultaneously affected by the normalized input feature vector and its dynamic attention weight coefficient vector A differential modeling mechanism that reflects the nonlinear degradation characteristics and characteristic sensitive dynamics of photovoltaic cells is constructed based on the combined influence of the two factors.

[0102] In this embodiment, S3 includes the following steps:

[0103] S31. In the training process of photovoltaic cell health status model, the jellyfish swarm optimization algorithm is introduced to optimize the parameters in the neural ordinary differential equation structure and set the optimization parameter set. ;

[0104] S32. Set the optimization parameters Encoded as the position vector of the jellyfish individual ,in , is the number of individuals in the jellyfish population, constructing the initial jellyfish population ;

[0105] S33. Based on the current population Each jellyfish The corresponding neural ordinary differential equation structure is used to calculate the health status prediction error of the neural ordinary differential equation structure on the training set and construct the loss function: ;

[0106] in, 、 Respectively represent A jellyfish individual at a time The predicted functional state and degradation state of photovoltaic cells, 、 is the real label data;

[0107] S34. Based on the behavior mechanism of the jellyfish swarm optimization algorithm, the position of each jellyfish is updated using the active motion and tidal drift mechanisms: ;

[0108] in, Indicates the The jellyfish individuals in The parameter vector for the iteration, is the optimal individual parameter vector with the smallest prediction error in the current population, is the disturbance vector simulating tidal flow, 、 To control the adjustment coefficient of active motion and drift intensity, 、 is a uniformly distributed random number;

[0109] S35. In each round of iteration, based on the loss function Sort and select all jellyfish individuals and retain the current global optimal parameter vector : ;

[0110] S36. Repeat the jellyfish swarm optimization process until the maximum number of iterations or the loss function convergence condition is met, and finally output the optimal parameter set , the optimal parameter set is applied to each module of the neural ordinary differential equation structure to construct a photovoltaic cell health status model with a global optimal solution.

[0111] In this embodiment, S4 includes the following steps:

[0112] S41. Comprehensive evaluation index based on photovoltaic cell health , and calculate the health status assessment value by fusing the joint health status prediction sequence: ;

[0113] in, Indicates that the photovoltaic cell is at the time The health status assessment value, 、 are the evaluation weight coefficients of functional status and degradation status, respectively;

[0114] S42. Extract the classification standards of different health levels of photovoltaic cells from historical monitoring data, and use the health status assessment value The health level classification rules are defined as follows:

[0115] Normal state: If , indicating that the photovoltaic cell is in good operating condition and fully functional, and it is recommended to continue normal operation;

[0116] Mild degenerative state: If , indicating that the photovoltaic cells have initial aging phenomena, and it is recommended to conduct extended operation cycle testing and non-invasive optical inspection;

[0117] Moderately degraded state: If , indicating that the photovoltaic cells have experienced significant performance degradation. It is recommended to conduct distributed component testing in combination with on-site temperature data and electrical performance analysis to estimate their remaining lifespan;

[0118] Severe degeneration: If , indicating that there may be structural or material failure in the photovoltaic cell. It is recommended to immediately stop using the component and arrange manual inspection and component replacement;

[0119] S43. The health status evaluation value of the photovoltaic cell at each moment Corresponding to the health level classification standard, the output includes the health status assessment results of the current health level and health trend indicators.

[0120] In this embodiment, S5 includes the following steps:

[0121] S51. Based on the photovoltaic cell sampling period in the full cycle The health status assessment trend sequence within the dataset is used to extract the health status change trajectory and calculate the degradation rate index: ;

[0122] in, Indicates the average rate of decrease of the health status assessment value of the photovoltaic cell per unit time, which is used to determine whether there is a risk of rapid degradation;

[0123] S52. Based on the classification results and degradation rate indicators Generate health status assessment reports;

[0124] S53. Generate preventive maintenance recommendations based on the combination of the health level and the degradation rate indicator.

[0125] In this embodiment, the health status assessment report includes the current health level, the health status trend curve of the most recent complete cycle, and the impact weights of key features.

[0126] In this embodiment, the preventive maintenance suggestion includes a recommended monitoring cycle, whether to enable manual inspection, and whether to perform replacement warning marking.

[0127] A deep learning-based intelligent photovoltaic battery health assessment system is used to implement a deep learning-based intelligent photovoltaic battery health assessment method, including the following modules:

[0128] Data acquisition and preprocessing module, used to collect raw data of photovoltaic cell operation and complete preprocessing to generate standardized data sets;

[0129] The health state modeling module builds a photovoltaic cell health state model based on the Neural Ordinary Differential Equation to characterize the continuous time evolution of the photovoltaic cell operating state;

[0130] The parameter optimization module performs global optimization on the parameters in the photovoltaic cell health status model based on the jellyfish swarm optimization algorithm and outputs the optimal parameter set;

[0131] The state prediction and assessment module uses the optimized photovoltaic cell health state model to predict the current state of the photovoltaic cell and generate health state assessment results;

[0132] The status feedback and decision-making module classifies and grades the health status of photovoltaic cells according to the set health status classification standards and degradation trends, and generates assessment reports and preventive maintenance recommendations.

[0133] Example 1

[0134] During routine data collection, the operation monitoring system of a large ground-based photovoltaic power station in City A discovered that the output power of the 14th group of components in the 7th column in the east zone (number: D7G14) fluctuated slightly compared to the rest of the components in the string. The frequency of power fluctuations increased abnormally (more than seven deviations exceeding 3% in the past two hours). This deviation did not trigger the inverter protection mechanism but entered the gray-zone anomaly window of the operation and maintenance monitoring.

[0135] The intelligent health assessment system activates the model of the present invention and analyzes the data of the component D7G14 for the last 24 hours. The system retrieves the continuous data of the component in real time, collecting a total of 288 data points. The data format is (timestamp, voltage V, current A, temperature °C, irradiance W / m²). The following is a portion of the original record:

[0136] Table 1 Partial original records of component D7G14

[0137] Timestamp Voltage (V) Current (A) Temperature (℃) Irradiance (W / m²) 2024-08-13 08:45:00 38.72 4.96 47.6 845 2024-08-13 12:10:00 38.21 4.52 53.1 916 2024-08-13 15:05:00 37.35 4.01 55.2 888 2024-08-13 18:35:00 38.02 4.37 49.8 691 2024-08-14 05:50:00 36.90 3.94 46.4 402

[0138] The data is fed into the continuous-time modeling module constructed using neural network ordinary differential equations. The system automatically loads the initial values ​​of the historical state variables h0=0.81 and z0=0.23 for component number D7G14, and constructs a differential prediction path based on the current input eigenvector. It predicts that the component health status score will drop rapidly to 0.39 in the next three hours, and the degradation variable will grow exponentially.

[0139] At the same time, the dynamic attention mechanism analyzes the factors affecting the state of the component, and the system calculates the attention weight of the input feature at the current moment as:

[0140] αV(t)=0.16 (voltage); αI(t)=0.11 (current); αT(t)=0.48 (temperature); αG(t)=0.25 (irradiance). The system determines that temperature anomalies have a significant impact on the degradation trend.

[0141] The system refers to the established health status threshold standards, determines that the current status has changed from "mild degradation" to "moderate degradation", and automatically generates a health status assessment report containing the following content:

[0142] Current health score S(t) = 0.41; degradation rate Rdeg = 0.026 / h in the last 12 hours; recommended maintenance level: Level 2; recommended measures: component thermal imaging inspection + electrical performance testing.

[0143] The system pushes the above reports and suggestions to the PDA devices of front-line inspection personnel through the internal operation and maintenance platform of the main control room.

[0144] Frontline inspection personnel reviewed the system assessment recommendations and began their on-site inspection. Upon arriving at the D7G14 module site, they used an infrared thermal imager to detect a non-uniform hotspot in the lower left corner of the module, with a temperature 12.7°C higher than the surrounding area. Subsequently, they conducted an open-circuit voltage test: the actual voltage was 37.22V, approximately 1.3V lower than the average voltage of 38.56V for adjacent modules in the same group, and the current was 4.08A. This initially led to the diagnosis of localized cracking or solder delamination in the module.

[0145] The front-line inspection personnel completed the on-site test and checked the "accept the recommendation to stop use" in the system. The system automatically marked the component as "high risk to be replaced" and recorded the fault identification time as "2024081406:29".

[0146] To verify the advantages of the method in the actual scenario, the project team deployed a comparative test between the traditional LSTM model and the empirical rule method on the same day, selecting the same input data. The test results are as follows:

[0147] Table 2 Test results of the present invention model and the comparative model

[0148] Model Type Fault identification time Initial health score Health decline warning time Alarm advance Method of the present invention 06:29 0.81 06:29 - LSTM prediction method 08:10 0.79 08:06 97 minutes lag Based on empirical threshold method No alarm No rating none Unidentified fault

[0149] In addition, a comparative analysis was conducted on 25 degraded component events identified by the proposed model and the comparison model within one week of continuous deployment:

[0150] Table 3 Comparative analysis of evaluation indicators of the present model and the comparative model

[0151] Evaluation Metrics Method of the present invention LSTM model empirical method Fault identification accuracy 96.2% 83.1% 68.7% Advance warning time (average) 2.3 hours 0.6 hours 0 hours Health score error (mean square) 0.014 0.029 Unable to evaluate Response time (model calculation time) 5.4 seconds 19.8 seconds 0.3 seconds

[0152] At the monthly summary meeting, the power plant's chief operations engineer pointed out that since its deployment, the intelligent health assessment system has identified 38 early-stage degraded components, reduced unplanned outages by 11 times, and avoided series inefficiencies caused by hidden component failures. The system has also triggered 65 effective maintenance recommendations, with an adoption rate of 94%.

[0153] The present invention constructs a continuous-time modeling framework based on neural ordinary differential equations. By introducing a two-variable joint modeling mechanism of healthy state variables and degradation state variables, simultaneous differential equations are established to perform high-precision modeling of the evolution process of the functional state and degradation trend state of photovoltaic cells. The neural ordinary differential equations can model the state that changes continuously in time with arbitrary resolution, avoiding the limitation of fixed time steps, and improving the adaptability of the model to non-uniform sampling data. It is suitable for photovoltaic cell component systems with complex operating environments and slow state changes.

[0154] The present invention introduces the jellyfish swarm optimization algorithm into the training process of neural ordinary differential equations, and performs global search optimization on a multidimensional parameter set including health state function parameters, degradation state function parameters, and attention mechanism network parameters. By simulating the active motion and tidal drift behavior of jellyfish, efficient search is achieved in the complex loss function space, avoiding the problem of gradient descent method easily falling into local optimality, and significantly improving the stability of training convergence and the consistency of model performance.

[0155] The present invention proposes a time-adaptive dynamic attention weight mechanism. By constructing a neural network mapping function with degraded state variables and input feature vectors as input, the attention coefficient is dynamically generated at each moment. The weights of the key influencing factors of photovoltaic cells at different time points can be quantified, and the state-based feature weight self-adjustment is realized, so that the model has stronger discrimination and generalization capabilities when facing complex disturbance factors from multiple sources.

[0156] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A smart photovoltaic battery health assessment method based on deep learning, characterized in that: The steps include: S1. Real-time collection of raw data generated by photovoltaic cells during operation and preprocessing to obtain a standardized dataset that meets modeling requirements; S2. Input the standardized data into a continuous-time dynamic health state modeling system based on neural ordinary differential equations. The continuous-time dynamic health state modeling system generates a photovoltaic cell health state model that describes the continuous evolution of the photovoltaic cell's internal state over time through training. The S2 comprises the following steps: S21. Construct a photovoltaic cell health state model based on a standardized data set, and introduce a two-variable joint modeling mechanism into the neural ordinary differential equation structure. The two-variable joint modeling mechanism includes the photovoltaic cell health state variable and photovoltaic cell degradation state variables , the photovoltaic cell health state variable and photovoltaic cell degradation state variable are used to characterize the functional state and degradation trend state of the photovoltaic cell respectively, and a simultaneous differential equation model is established: ; in, Indicates time The input feature vector of the standardized dataset, 、 The parameters are the neural network trainable parameter groups 、 The neural network function, is the dynamic attention weight coefficient vector; S22. Define the dynamic weight adjustment mechanism of the degradation state and construct the time-adaptive dynamic attention weight coefficient vector ; S23. Based on the simultaneous differential equation model constructed in step S21, input the initial state 、 , for continuous time intervals Perform forward integral solution to obtain the joint health status prediction sequence ,in, Indicates at time The estimated value of the photovoltaic cell functional state obtained by prediction, Indicates at time The estimated degradation trend state obtained by prediction; S24. The predicted joint health status prediction sequence Reflect the current operating status and potential aging trend of photovoltaic cells; S3. Introducing a jellyfish swarm optimization algorithm during the training of the photovoltaic cell health status model, the algorithm performs a global search and adaptive dynamic adjustment of model parameters to achieve optimal settings for the model parameters and ultimately achieve a globally optimal solution for the photovoltaic cell health status model within a multidimensional parameter space. S4. Use the optimized PV cell health model to predict and evaluate the current health status of the PV cell in real time, and output a health status assessment result reflecting the dynamic health status of the PV cell. S5. The health status assessment results are transmitted to the assessment feedback module. The assessment feedback module classifies and grades the health status of the photovoltaic cells according to a pre-set health status grading standard and generates a corresponding assessment report and preventive maintenance recommendations.

2. The method for evaluating the health of a smart photovoltaic battery based on deep learning according to claim 1, characterized in that: Said S1 comprises the following steps: S11. Real-time collection of raw data generated by photovoltaic cells during operation, setting a sampling time window to collect data to form the original data set : ; in, Indicates at time The collected PV cell operation data, Indicates at time The collected photovoltaic cell voltage, Indicates at time The collected photovoltaic cell current, Indicates at time The collected photovoltaic cell temperature, Indicates at time The light irradiance, is the total number of sampling moments; S12. For the original dataset Data denoising is performed to remove random interference and high-frequency noise in the original data, and adjacent data interpolation is used to fill in the missing values ​​in the original data set. The data is estimated based on the average value of two adjacent valid observation points in the time series. Then, outlier removal is performed, and the valid interval range of each physical parameter is set. The observation data that is not within the valid interval range is removed. The valid interval range is set according to the normal physical state boundary of the photovoltaic cell operation. Finally, normalization is performed to convert the original physical parameter values ​​into interval values ​​according to a unified linear mapping rule. The dimensionless standard value within is constructed as a standardized data set ; S13. Each observation data in the standardized data set consists of five dimensions: sampling time, standardized photovoltaic cell voltage, standardized photovoltaic cell current, standardized photovoltaic cell temperature, and standardized light irradiance. Each standardized data is uniformly represented as an ordered data tuple containing the five indicators of sampling time, photovoltaic cell voltage, photovoltaic cell current, photovoltaic cell temperature, and light irradiance.

3. The method for evaluating the health of a smart photovoltaic battery based on deep learning according to claim 1, wherein: The S23 includes the following steps: S231. Photovoltaic cell degradation state variables In the dynamic modeling process, the time-varying influence of input features on the degradation state is considered, and a time-adaptive dynamic attention weight coefficient vector is constructed. : ; in, 、 、 、 Respectively indicate time The effects of photovoltaic cell voltage, current, temperature and light irradiance on photovoltaic cell degradation state variables The relative influence weight of S232. Time-adaptive dynamic attention weight coefficient vector By feature importance evaluation function Dynamically generated: ; in, Based on the neural network parameters The constructed nonlinear mapping function takes the current time as input Photovoltaic cell degradation state variables With the normalized input feature vector , the output is the unnormalized feature importance vector, which is normalized by Softmax to generate the dynamic attention weight coefficient vector ; S233. Neural Network Mapping Function Parameter set The dynamic attention weight coefficient vector generated by end-to-end training through back-propagation mechanism in the process of photovoltaic cell health state modeling At every moment The internal parameters are updated in real time to achieve dynamic identification and weight adjustment of the impact of different characteristic parameters of photovoltaic cells on the degradation process; S234. Dynamic attention weight coefficient vector Applied to the degenerate state differential equation in the simultaneous differential equation model, so that the degradation state variable of the photovoltaic cell The evolution process of is simultaneously affected by the normalized input feature vector and its dynamic attention weight coefficient vector A differential modeling mechanism that reflects the nonlinear degradation characteristics and characteristic sensitive dynamics of photovoltaic cells is constructed based on the combined influence of the two factors.

4. The method for evaluating the health of a smart photovoltaic battery based on deep learning according to claim 2, wherein: The S3 includes the following steps: S31. In the training process of photovoltaic cell health status model, the jellyfish swarm optimization algorithm is introduced to optimize the parameters in the neural ordinary differential equation structure and set the optimization parameter set. ; S32. Set the optimization parameters Encoded as the position vector of the jellyfish individual ,in , is the number of individuals in the jellyfish population, constructing the initial jellyfish population ; S33. Based on the current population Each jellyfish The corresponding neural ordinary differential equation structure is used to calculate the health status prediction error of the neural ordinary differential equation structure on the training set and construct the loss function: ; in, 、 Respectively represent A jellyfish individual at a time The predicted functional state and degradation state of photovoltaic cells, 、 is the real label data; S34. Based on the behavioral mechanism of the jellyfish swarm optimization algorithm, the position of each jellyfish individual is updated using active motion and tidal drift mechanisms. ; S35. In each round of iteration, based on the loss function Sort and select all jellyfish individuals and retain the current global optimal parameter vector : ; S36. Repeat the jellyfish swarm optimization process until the maximum number of iterations or the loss function convergence condition is met, and finally output the optimal parameter set , the optimal parameter set is applied to each module of the neural ordinary differential equation structure to construct a photovoltaic cell health status model with a global optimal solution.

5. The method for evaluating the health of a smart photovoltaic battery based on deep learning according to claim 4, characterized in that: The S4 comprises the following steps: S41. Based on comprehensive evaluation indicators of photovoltaic cell health , and calculate the health status assessment value by fusing the joint health status prediction sequence: ; in, Indicates that the photovoltaic cell is at the time The health status assessment value, 、 are the evaluation weight coefficients of functional status and degradation status, respectively; S42. Extract the classification standards of different health levels of photovoltaic cells from historical monitoring data, and use the health status assessment value The health level classification rules are defined as follows: Normal state: If , indicating that the photovoltaic cell is in good operating condition and fully functional, and it is recommended to continue normal operation; Mild degenerative state: If , indicating that the photovoltaic cells have initial aging phenomena, and it is recommended to conduct extended operation cycle testing and non-invasive optical inspection; Moderately degraded state: If , indicating that the photovoltaic cells have experienced significant performance degradation. It is recommended to conduct distributed component testing in combination with on-site temperature data and electrical performance analysis to estimate their remaining lifespan; Severe degeneration: If , indicating that there may be structural or material failure in the photovoltaic cell. It is recommended to immediately stop using the component and arrange manual inspection and component replacement; S43. The health status evaluation value of the photovoltaic cell at each moment Corresponding to the health level classification standard, the output includes the health status assessment results of the current health level and health trend indicators.

6. The method for evaluating the health of a smart photovoltaic battery based on deep learning according to claim 5, characterized in that: The S5 comprises the following steps: S51. Based on the photovoltaic cell sampling period in the whole cycle The health status assessment trend sequence within the dataset is used to extract the health status change trajectory and calculate the degradation rate index: ; in, Indicates the average rate of decrease of the health status assessment value of the photovoltaic cell per unit time, which is used to determine whether there is a risk of rapid degradation; S52. Based on the classification results and degradation rate indicators Generate health status assessment reports; S53. Generate preventive maintenance recommendations based on the combination of the health level and the degradation rate indicator.

7. The method for evaluating the health of a smart photovoltaic battery based on deep learning according to claim 6, characterized in that: The health status assessment report includes the current health level, the health status trend curve of the most recent complete cycle, and the impact weights of key features.

8. The method for evaluating the health of a smart photovoltaic battery based on deep learning according to claim 6, wherein: The preventive maintenance recommendations include recommended monitoring cycles, whether to enable manual inspections, and whether to perform replacement warning markings.

9. A deep learning-based intelligent photovoltaic battery health assessment system, used to execute the deep learning-based intelligent photovoltaic battery health assessment method according to any one of claims 1 to 8, characterized in that: Includes the following modules: Data acquisition and preprocessing module, used to collect raw data of photovoltaic cell operation and complete preprocessing to generate standardized data sets; The health state modeling module builds a photovoltaic cell health state model based on the Neural Ordinary Differential Equation to characterize the continuous time evolution of the photovoltaic cell operating state; The parameter optimization module performs global optimization on the parameters in the photovoltaic cell health status model based on the jellyfish swarm optimization algorithm and outputs the optimal parameter set; The state prediction and assessment module uses the optimized photovoltaic cell health state model to predict the current state of the photovoltaic cell and generate health state assessment results; The status feedback and decision-making module classifies and grades the health status of photovoltaic cells according to the set health status classification standards and degradation trends, and generates assessment reports and preventive maintenance recommendations.

Images