Energy flow dynamic simulation management system based on digital twinning

Abstract

The application relates to an energy flow dynamic simulation management system based on digital twinning, and relates to the technical field of energy digital management, and comprises the following: a monitoring and power generation and transmission prediction module, which is used for monitoring the parameter distribution in the environment of a photovoltaic component and predicting power generation and transmission parameters; a dynamic twinning modeling module, which is used for constructing a photovoltaic energy flow model based on digital twinning according to the power generation and transmission parameters, and obtaining a photovoltaic energy flow model sequence; a twinning difference compensation module, which is used for processing the twinning difference of historical and real-time photovoltaic energy flow models, and obtaining compensated twinning difference after compensation; and a damage prediction and identification management module, which is used for obtaining a historical parameter distribution sequence, combining the compensated twinning difference to predict the damage probability of the photovoltaic component, and identifying and managing the real-time photovoltaic energy flow model. The application solves the problem that traditional photovoltaic power generation digital simulation management only focuses on real-time updating of power generation states and does not focus on faults caused by photovoltaic hot spot effects, thereby causing insufficient early warning of potential damage to photovoltaic components.

Description

Technical Field

[0001] This application relates to the field of digital energy management, and in particular to a dynamic simulation management system for energy flow based on digital twins. Background Technology

[0002] With the continuous development of energy digital management technology, the intelligent management of photovoltaic power generation systems has an increasingly significant impact on energy utilization efficiency. Precise control of fault risks such as photovoltaic hot spot effect has become a key technical challenge to improve the reliability of photovoltaic systems.

[0003] Currently, traditional photovoltaic power generation digital simulation management focuses on updating the real-time power generation status, making it difficult to take into account the potential faults of photovoltaic modules caused by hot spot effects. This can lead to delayed early warning of module damage and insufficient management accuracy, which not only reduces the operational stability of the photovoltaic power generation system but also increases the risk of equipment maintenance costs and energy loss. Summary of the Invention

[0004] To address the aforementioned technical issues, this application provides a digital twin-based energy flow dynamic simulation management system, which improves upon the current situation in traditional photovoltaic power generation digital simulation management where the focus is solely on real-time updates of power generation status and neglects faults caused by photovoltaic hot spot effects, resulting in insufficient early warning of potential component damage and low management sophistication.

[0005] The embodiments of this application disclose the following technical solutions:

[0006] This application provides an energy flow dynamic simulation management system based on digital twins, the system comprising:

[0007] The monitoring and power generation / transmission prediction module is used to monitor and acquire the distribution of irradiance parameters, cloud shadow parameters, and temperature distribution of the photovoltaic module environment, and to predict and acquire the power generation and transmission parameters of the photovoltaic module.

[0008] The dynamic twin modeling module is used to construct a dynamically updated photovoltaic energy flow model based on digital twins according to the power generation parameters and transmission parameters, and obtain a photovoltaic energy flow model sequence.

[0009] The twin difference compensation module is used to process and obtain the twin difference between the historical photovoltaic energy flow model and the real-time photovoltaic energy flow model within the photovoltaic energy flow model sequence, and to compensate for the twin difference to obtain the compensated twin difference.

[0010] The damage prediction and identification management module is used to acquire historical light parameter distribution sequences, historical cloud shadow parameter distribution sequences, and historical temperature distribution sequences, and combine them with the compensated twin difference degree to predict the damage probability of photovoltaic modules, obtain the damage probability, identify the real-time photovoltaic energy flow model, and manage it.

[0011] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0012] This application proposes a dynamic energy flow simulation management system based on digital twins. First, a monitoring and power generation / transmission prediction module divides the photovoltaic module environment into multiple monitoring areas and collects data on irradiance, cloud cover, and temperature distributions. The power generation and transmission parameters are then predicted and output by a power generation / transmission predictor. Next, a dynamic twin modeling module constructs a digital twin-based photovoltaic energy flow model based on the obtained power generation and transmission parameters and updates it at a preset frequency to form a photovoltaic energy flow model sequence. Then, a twin difference compensation module calculates the difference between the historical photovoltaic energy flow model and the real-time photovoltaic energy flow model, and combines this with the prediction error rate to obtain the compensated twin difference degree. Finally, a damage prediction and identification management module integrates environmental parameters and the corrected twin difference degree to predict the damage probability and performs collaborative operations on the identification management of the real-time photovoltaic energy flow model, achieving dynamic simulation of photovoltaic energy flow and precise control of module damage risk.

[0013] This technical solution addresses the problem in traditional photovoltaic energy management that focuses solely on real-time updates of power generation status while neglecting faults caused by photovoltaic hot spot effects. This is achieved through a combination of steps including comprehensive monitoring and intelligent prediction of environmental parameters across multiple regions, dynamic construction and sequence tracking of digital twin models, compensation and correction of model differences, and multi-dimensional damage prediction. The solution resolves the issue of delayed early warning of module damage and insufficient management precision caused by the photovoltaic hot spot effect. It realizes dynamic simulation and precise management of the entire photovoltaic energy flow process, providing scientific support for photovoltaic module operation and maintenance decisions. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 A schematic diagram of the structure of the energy flow dynamic simulation management system based on digital twin provided in the embodiments of this application;

[0016] Figure 2 A schematic diagram illustrating the process of obtaining compensated twin discrepancies by combining the prediction error rate, as provided in an embodiment of this application;

[0017] The components represented by each number in the attached diagram are explained below:

[0018] Monitoring and power generation and transmission prediction module 01, dynamic twin modeling module 02, twin difference compensation module 03, damage prediction and identification management module 04. Detailed Implementation

[0019] This application provides a digital twin-based energy flow dynamic simulation management system to address the technical problem in existing photovoltaic power generation digital simulation management that only focuses on real-time updates of power generation status and does not pay attention to faults caused by photovoltaic hot spot effects, resulting in insufficient early warning of potential damage to photovoltaic modules and low management precision.

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0022] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0023] Examples, as shown in the appendix Figure 1 As shown, this application provides a dynamic energy flow simulation management system based on digital twins, the system comprising the following modules:

[0024] The monitoring and power generation / transmission prediction module 01 is used to monitor and acquire the distribution of light parameters, cloud shadow parameters, and temperature distribution of the photovoltaic module environment, and to predict and acquire the power generation and transmission parameters of the photovoltaic module.

[0025] In this embodiment of the application, in the scenario of dynamic simulation management of photovoltaic energy flow, in order to accurately obtain the power generation parameters and transmission parameters of photovoltaic modules, it is necessary to obtain accurate distribution of irradiance parameters, cloud shadow parameters, temperature distribution, and corresponding power generation parameters and transmission parameters by monitoring the environmental parameters of photovoltaic modules and applying power generation and transmission predictors.

[0026] Specifically, multiple monitoring areas within the photovoltaic module environment are first determined, which serve as the basis for parameter collection.

[0027] Furthermore, by deploying corresponding sensing devices in these monitoring areas, the illumination parameters, cloud shadow parameters, and temperature in each area are monitored in real time, thereby integrating them to form a comprehensive distribution of illumination parameters, cloud shadow parameters, and temperature covering the entire photovoltaic module environment, so as to fully reflect the environmental status of the photovoltaic module.

[0028] Meanwhile, considering the losses during power transmission, predicting only generation parameters cannot meet the needs of energy flow management; therefore, it is necessary to predict transmission parameters simultaneously. Based on this, a power generation and transmission predictor is constructed in advance. This predictor includes a generation prediction branch and a transmission prediction branch, which predict generation and transmission parameters respectively.

[0029] Furthermore, the monitored distributions of illumination parameters, cloud shadow parameters, and temperature are input into a pre-built power generation and transmission predictor. The predictor processes the input environmental parameters through its internally trained power generation and transmission prediction branches, and finally outputs the corresponding power generation and transmission parameters.

[0030] This step, through monitoring environmental parameters in multiple regions of the photovoltaic module and applying a dedicated power generation and transmission predictor, provides accurate power generation and transmission data support for the subsequent construction of a photovoltaic energy flow model based on digital twins, ensuring the accuracy and effectiveness of dynamic simulation management of energy flow.

[0031] The monitoring and power generation / transmission prediction module 01 in the system provided in this application embodiment includes:

[0032] Identify multiple monitoring areas within the photovoltaic module environment;

[0033] The system monitors and acquires illumination parameters, cloud shadow parameters, and temperature within multiple monitoring areas, obtaining the distribution of illumination parameters, cloud shadow parameters, and temperature.

[0034] The illumination parameter distribution, cloud shadow parameter distribution, and temperature distribution are input into a pre-constructed power generation and transmission predictor to obtain power generation parameters and transmission parameters.

[0035] In this embodiment of the application, in order to accurately obtain the power generation and transmission parameters of photovoltaic modules, it is necessary to conduct scientific environmental parameter monitoring and construct a power generation and transmission prediction model. By combining multi-regional monitoring with intelligent prediction, the power generation and transmission status of the photovoltaic system can be accurately controlled. At the same time, transmission loss factors are considered to ensure the comprehensiveness and practicality of the prediction results.

[0036] Specifically, the first step is to divide the environment of the photovoltaic modules into multiple monitoring zones. That is, based on the layout characteristics, installation density, and distribution of environmental influencing factors of the photovoltaic modules, the environment in which the photovoltaic modules are located is divided into multiple independent monitoring zones.

[0037] For example, for large-scale photovoltaic power plants, monitoring areas can be divided according to the array, with each area covering a certain number of photovoltaic panels, to ensure that the environmental parameters in each area are representative and can reflect local characteristics.

[0038] Furthermore, multi-regional parameter monitoring will be conducted. Corresponding sensing devices, such as light sensors, temperature sensors, and cloud shadow monitoring devices, will be deployed in each monitoring area to collect real-time light parameters, cloud shadow parameters, and temperature in each area.

[0039] Among them, the illumination parameters include illumination intensity (unit: W / m²). 2 ), illumination duration (unit: h), spectral distribution (the proportion of light intensity at different wavelengths), etc., are used to quantify the light energy input received by photovoltaic modules.

[0040] In addition, cloud shadow parameters include cloud coverage (the percentage of the monitored area that is blocked by cloud shadows), cloud shadow movement speed (unit: m / s), and cloud shadow thickness (reflecting the degree of cloud blockage of sunlight), which are used to assess the dynamic impact of cloud shadows on sunlight.

[0041] In addition, the temperature is the surface temperature of the photovoltaic module (unit: °C), which directly affects the power generation efficiency and operating status of the module. It needs to be monitored separately from the ambient temperature to accurately reflect the thermal state of the module itself.

[0042] Furthermore, these environmental parameters obtained from the tests are summarized and analyzed to form a distribution of illumination parameters, cloud shadow parameters, and temperature covering the entire photovoltaic module environment, so as to comprehensively reflect the environmental status of the photovoltaic module.

[0043] For example, a photovoltaic power station is divided into 10 monitoring zones, each equipped with a light sensor, an infrared thermometer, and a cloud shadow monitoring camera. Monitoring showed that the light intensity in zone 1 was 750 W / m². 2 The cumulative sunshine duration was 3.2 hours, cloud cover was 15%, cloud movement speed was 0.8 m / s, and the component surface temperature was 42℃; the irradiance in area 5 was 520 W / m². 2The cloud cover was 40%, and the component surface temperature was 38°C. The integrated distribution of light parameters, cloud cover parameters, and temperature clearly showed the environmental differences in different areas.

[0044] Meanwhile, considering the unavoidable losses such as conductor resistance loss and joint loss during power transmission, relying solely on power generation parameters cannot fully reflect the actual transmission and utilization of photovoltaic energy. Therefore, it is necessary to predict power transmission parameters simultaneously.

[0045] Therefore, it is necessary to build a power generation and transmission forecaster in advance. This forecaster includes a power generation forecasting branch and a transmission forecasting branch, which predict power generation parameters and transmission parameters respectively.

[0046] The steps for pre-constructing the "power generation and transmission predictor" in the system provided in this application embodiment include:

[0047] Based on the photovoltaic module operation data over a historical period, sample light parameter distribution sets, sample cloud shadow parameter distribution sets, and sample temperature distribution sets were collected. The power generation and transmission of the photovoltaic modules were also collected, and the sample power generation parameter sets and sample transmission parameter sets were obtained by labeling.

[0048] Based on machine learning, power generation prediction branch and power transmission prediction branch are constructed;

[0049] Using the sample illumination parameter distribution set, sample cloud shadow parameter distribution set, and sample temperature distribution set as input training data, and the sample power generation parameter set and sample power transmission parameter set as output supervision data, respectively, the power generation prediction branch and the power transmission prediction branch are iteratively supervised and trained. After convergence, the power generation and transmission predictor is obtained.

[0050] In this embodiment of the application, in order to ensure that the power generation and transmission predictor can accurately predict the power generation parameters and transmission parameters, the predictor needs to be built and trained based on historical operating data, and the mapping relationship between environmental parameters and power generation and transmission parameters is learned through machine learning algorithms to achieve accurate prediction.

[0051] First, we carried out the work of collecting and labeling sample data.

[0052] Specifically, from the historical operation records of photovoltaic modules, operation data under different time periods and environmental conditions are extracted, and sample light parameter distribution sets containing light intensity distribution, light duration distribution and other light parameters at each historical moment are collected, sample cloud shadow parameter distribution sets containing cloud coverage distribution, cloud shadow movement speed distribution and other cloud parameters at each historical moment are collected, and sample temperature distribution sets containing module surface temperature distribution at each historical moment are collected.

[0053] Simultaneously, the power generation and transmission power of photovoltaic modules within the corresponding time period are collected, and these power data are organized and labeled to form a sample power generation parameter set and a sample transmission parameter set.

[0054] For example, historical operating data from the past two years is collected, divided into monthly time periods, with 800 samples collected each month. Among them, the sample illumination parameter distribution set contains the illumination intensity matrix of each region (such as a 10×10 region matrix, where each element is the illumination intensity of the corresponding region), the sample cloud shadow parameter distribution set contains the cloud shadow coverage percentage of each region, and the sample temperature distribution set contains the component surface temperature values ​​of each region.

[0055] In addition, the corresponding sample power generation parameter set is the total power generation at each time (unit: kW), and the sample transmission parameter set is the transmission power after deducting losses (unit: kW).

[0056] Furthermore, power generation prediction and transmission prediction branches are constructed based on machine learning, both of which use gradient boosting tree models as their basic framework.

[0057] The power generation prediction branch takes the distribution of solar radiation parameters, cloud shadow parameters, and temperature distribution as input features and outputs the predicted power generation value; the power transmission prediction branch, on the basis of the same input features, additionally incorporates parameters such as transmission line length and conductor type, and outputs the predicted power transmission value after taking losses into account.

[0058] Furthermore, sample data is used to train the two prediction branches. The sample sets of illumination, cloud shadow, and temperature parameters are used as input, and the sample sets of power generation parameters and power transmission parameters are used as supervision data. The prediction error is minimized by iteratively adjusting the model parameters.

[0059] During training, the root mean square error (RMSE) is used as the evaluation metric. When the change in RMSE of the validation set is less than 0.01 over 30 consecutive iterations, the model is considered to have converged and training is stopped.

[0060] For example, the power generation prediction branch converges after 120 training rounds, and the predicted power generation for a certain sample deviates from the actual value by only 2.3%; the power transmission prediction branch converges after 150 training rounds, and the prediction error for power transmission loss is controlled within 3%.

[0061] Finally, the real-time monitored distributions of light, cloud shadow, and temperature are input into the power generation and transmission forecaster. The power generation forecast branch outputs power generation parameters, and the transmission forecast branch outputs transmission parameters, thus achieving accurate quantification of the power generation capacity and actual transmission efficiency of photovoltaic modules under the current environmental conditions.

[0062] For example, at a certain moment, the distribution of light parameters in the photovoltaic module environment was monitored as follows: the light intensity in each area was concentrated between 680-820 W / m². 2 The average sunshine duration has reached 4.5 hours; the cloud shadow parameters are as follows: the cloud coverage rate in the northeast region is 12%, the cloud shadow movement speed is 0.6 m / s, and there is no cloud shadow in the other regions; the temperature distribution is as follows: the surface temperature of the components is between 37-41℃, with an average temperature of 39℃.

[0063] Furthermore, after inputting the above parameters into the power generation and transmission forecaster, the power generation forecast branch outputs a power generation parameter of 1050kW, and the transmission forecast branch outputs a transmission parameter of 980kW (including 70kW of transmission loss).

[0064] The predicted results deviated from the actual power generation of 1045kW and transmission power of 978kW collected during operation by 0.48% and 0.20%, respectively, which fully verified the accuracy of the prediction and laid a solid data foundation for the dynamic construction of the photovoltaic energy flow model.

[0065] The power generation and transmission parameters obtained through the above steps comprehensively reflect the power generation and transmission status of the photovoltaic modules, providing high-quality data support for the subsequent construction of a photovoltaic energy flow model based on digital twins, and ensuring the scientificity and reliability of energy flow dynamic simulation management.

[0066] The dynamic twin modeling module 02 is used to construct a dynamically updated photovoltaic energy flow model based on digital twins according to the power generation parameters and transmission parameters, and obtain a photovoltaic energy flow model sequence.

[0067] In this embodiment of the application, in the scenario of dynamic simulation management of photovoltaic energy flow, in order to achieve accurate simulation and dynamic tracking of photovoltaic energy flow, it is necessary to construct and dynamically update the photovoltaic energy flow model based on digital twin technology, combined with power generation parameters and transmission parameters, to form a model sequence to reflect the real-time changes of energy flow.

[0068] Specifically, the core data foundation is the obtained power generation and transmission parameters, and a photovoltaic energy flow model is constructed based on digital twin technology.

[0069] The photovoltaic energy flow model uses digital mapping to present the power generation process, power transmission process, and energy flow status of photovoltaic modules in the form of a virtual model, achieving accurate replication of the energy flow of photovoltaic modules.

[0070] Furthermore, to ensure that the model can match the actual operating status of the photovoltaic modules in real time, the constructed photovoltaic energy flow model is dynamically updated at a preset frequency.

[0071] The preset frequency can be set according to the operating characteristics and management needs of the photovoltaic modules, such as updating every 10 minutes, to ensure that the model can reflect the changes in power generation and transmission parameters in a timely manner.

[0072] Based on this, new photovoltaic energy flow models are continuously generated through dynamic updates, ultimately forming a sequence of photovoltaic energy flow models. This sequence includes photovoltaic energy flow models at different time points, fully recording the evolution of energy flow over time.

[0073] This step involves constructing a dynamically updated photovoltaic energy flow model based on digital twins and forming a sequence. This provides continuous and accurate model support for subsequent steps such as analyzing the differences between historical and real-time photovoltaic energy flow models and predicting the probability of damage to photovoltaic modules, ensuring the timeliness and comprehensiveness of dynamic simulation management of photovoltaic energy flow.

[0074] The dynamic twin modeling module 02 in the system provided in this application embodiment includes:

[0075] Based on the power generation and transmission parameters, a photovoltaic energy flow model is constructed using digital twins.

[0076] The photovoltaic energy flow model is dynamically updated according to a preset frequency to obtain a photovoltaic energy flow model sequence.

[0077] In this embodiment of the application, in order to achieve dynamic simulation and full-process tracking of photovoltaic energy flow, it is necessary to rely on digital twin technology, using power generation parameters and transmission parameters as core data, to construct a dynamically updatable photovoltaic energy flow model and form a model sequence, thereby accurately mapping the energy flow state of photovoltaic modules and providing a continuous and reliable model foundation for subsequent difference analysis and damage prediction.

[0078] First, we will construct a photovoltaic energy flow model. Specifically, using the power generation and transmission parameters output by the power generation and transmission forecaster as key inputs, and combining them with the physical structural parameters of the photovoltaic modules, we will construct a virtual photovoltaic energy flow model based on existing digital twin technology.

[0079] The photovoltaic energy flow model uses digital mapping to present the entire process of energy generation, transmission, and loss in physical photovoltaic modules from the power generation end to the power transmission end in a visualized virtual form, achieving accurate replication of energy flow path, energy magnitude, and real-time status.

[0080] For example, when the power generation parameter is 1050kW and the transmission parameter is 980kW (including 70kW loss), the photovoltaic energy flow model built based on digital twin technology will clearly display the power generation distribution of each photovoltaic array in the virtual scene, such as 120kW power generation in area 1 and 110kW power generation in area 2.

[0081] Meanwhile, dynamic arrows are used to mark the transmission path of energy from components to combiner boxes, inverters and then to transmission lines, and the transmission power of each segment of the line is marked, such as the total output power of the combiner box is 1050kW, the output power of the inverter is 1030kW, and the received power at the end of the transmission line is 980kW, so as to fully restore the real-time status of energy flow.

[0082] Furthermore, to ensure that the photovoltaic energy flow model can match the dynamic changes of photovoltaic modules in real time, the model needs to be dynamically updated at a preset frequency to accurately capture changes in energy flow caused by fluctuations in environmental parameters or changes in equipment operating status.

[0083] The preset frequency setting needs to comprehensively consider the operating characteristics and management requirements of the photovoltaic modules, and can be set to a fixed time interval or a triggered update. At the same time, through dynamic updates, the photovoltaic energy flow model can reflect changes in energy flow in the photovoltaic modules caused by environmental changes or changes in equipment status in a timely manner.

[0084] For example, if the preset update frequency is once every 10 minutes, after the photovoltaic energy flow model is built at 10:00, at 10:10 it is detected that the increase in light intensity causes the power generation parameter to increase to 1100kW and the transmission parameter to change to 1025kW. At this time, the model will automatically update the power generation distribution, power value on the transmission path and loss ratio in the virtual scene to generate the photovoltaic energy flow model at 10:10. At 10:20, due to cloud shadow shading, the power generation parameter drops to 980kW and the transmission parameter changes to 920kW. The model is updated again and a new state model is generated.

[0085] Based on this, through continuous dynamic updates, photovoltaic energy flow models at different time points will form an ordered sequence of photovoltaic energy flow models.

[0086] The photovoltaic energy flow model sequence fully records the evolution of energy flow over time, including both historical photovoltaic energy flow models (such as models at 10:00 and 10:10) and real-time updated current photovoltaic energy flow models (such as models at 10:20), providing a continuous data source for analyzing the changing trends of energy flow and comparing the differences between models at different times.

[0087] For example, a photovoltaic power station's photovoltaic energy flow model sequence contains 144 models in one day. By comparing the models from 8:00 to 12:00, it can be found that as sunlight intensifies, the power values ​​of the power generation areas in the models gradually increase, and the energy flow density along the transmission path increases synchronously. However, from 14:00 to 16:00, due to the influence of cloud shadows, the power generation in some areas of the photovoltaic energy flow model fluctuates, and the transmission parameters also show dynamic changes, thus intuitively reflecting the real-time impact of environmental factors on energy flow.

[0088] The twin difference compensation module 03 is used to process and obtain the twin difference between the historical photovoltaic energy flow model and the real-time photovoltaic energy flow model within the photovoltaic energy flow model sequence, and to compensate for the twin difference to obtain the compensated twin difference.

[0089] In this embodiment of the application, in the scenario of dynamic simulation management of photovoltaic energy flow, in order to accurately reflect the changes of photovoltaic energy flow model over time and eliminate the interference of prediction error on difference analysis, it is necessary to calculate the difference degree between historical and real-time photovoltaic energy flow models and perform compensation processing in combination with prediction error rate to obtain a more reliable compensated twin difference degree, so as to provide accurate difference data support for subsequent component damage probability prediction.

[0090] Specifically, historical photovoltaic energy flow models and real-time photovoltaic energy flow models are first extracted from the photovoltaic energy flow model sequence.

[0091] Among them, the historical photovoltaic energy flow model can select several historical node models that are closest to the current time in the sequence, while the real-time photovoltaic energy flow model is the latest updated model at the current time. By comparing the two, the changes of the photovoltaic energy flow model in the time dimension can be reflected.

[0092] Furthermore, the difference between the real-time photovoltaic energy flow model and the historical photovoltaic energy flow model is calculated to obtain the twin difference. The difference calculation is based on the comparison of core parameters in the models, including the distribution of power generation parameters, the distribution of power transmission parameters, and power losses along the energy flow path. By quantitatively analyzing the degree of deviation between the two models on these parameters, the specific value of the twin difference is obtained.

[0093] At the same time, the error rate of the predicted power generation parameters and transmission parameters is obtained as a compensation coefficient.

[0094] The error rate originates from the deviation generated by the power generation and transmission forecaster during the forecasting process, i.e., the proportion of deviation between the predicted value and the actual value. By statistically summing the error rates of power generation parameters and transmission parameters within the current forecasting period, this error rate is incorporated as a compensation coefficient into the difference correction process.

[0095] Based on this, the twin discrepancy is compensated according to the compensation coefficient to obtain the compensated twin discrepancy. That is, the twin discrepancy is added to the compensation coefficient to eliminate the problem of model differences caused by prediction errors being ignored.

[0096] This step extracts historical and real-time photovoltaic energy flow models and calculates the degree of difference, then compensates for it by combining the prediction error rate. This not only accurately captures the real changes of the model over time, but also takes into account the abnormal deviations caused by errors in the prediction process, effectively improving the reliability of the degree of difference data and providing a more accurate analytical basis for subsequent prediction of component damage probability based on degree of difference.

[0097] As attached Figure 2 As shown, the twin difference compensation module 03 in the system provided in this application embodiment includes:

[0098] Historical photovoltaic energy flow models and real-time photovoltaic energy flow models are extracted from the photovoltaic energy flow model sequence;

[0099] Calculate the difference between the real-time photovoltaic energy flow model and the historical photovoltaic energy flow model to obtain the twin difference.

[0100] Obtain the error rate of the predicted power generation parameters and transmission parameters, and use it as a compensation coefficient;

[0101] The twin difference is compensated according to the compensation coefficient to obtain the compensated twin difference.

[0102] In this embodiment of the application, in order to accurately quantify the dynamic changes of the photovoltaic energy flow model over time and eliminate the interference of power generation and transmission parameter prediction errors on the difference analysis, it is necessary to obtain a more realistic compensated twin difference degree by extracting historical and real-time photovoltaic energy flow models, calculating the difference degree, and introducing compensation coefficient correction, so as to provide reliable difference data support for subsequent photovoltaic module damage probability prediction.

[0103] Specifically, the first step is to extract the model. This involves extracting historical and real-time photovoltaic energy flow models from the photovoltaic energy flow model sequence.

[0104] Among them, the historical photovoltaic energy flow model can be selected from the photovoltaic energy flow model sequence that is closest to the current time node model, while the real-time photovoltaic energy flow model is the latest updated model at the current time. By comparing the two, the change trajectory of the model in the time dimension can be intuitively reflected.

[0105] For example, if the photovoltaic energy flow model sequence is updated every 10 minutes, the model at 10:00 is extracted from the sequence as the historical photovoltaic energy flow model, and the model at 10:10 is used as the real-time photovoltaic energy flow model for subsequent difference calculation.

[0106] Furthermore, the difference between the real-time photovoltaic energy flow model and the historical photovoltaic energy flow model is calculated to obtain the twin difference.

[0107] The calculation of the difference is based on a quantitative comparison of core parameters in the photovoltaic energy flow model, including the distribution of power generation parameters, the distribution of power transmission parameters, and the matching degree of energy flow paths. The specific value of the twin difference is obtained by arithmetically averaging the deviations of these parameters.

[0108] For example, compared with the historical photovoltaic energy flow model at 10:00, the real-time photovoltaic energy flow model at 10:10 shows an average decrease of 5% in power generation in each region, an increase of 3% in transmission loss, and a 2% shift in energy flow path due to local power changes. The twin difference is calculated as (5%+3%+2%) / 3=3.33%, resulting in a twin difference of 3.33%.

[0109] At the same time, the error rate of the predicted power generation parameters and transmission parameters is obtained as a compensation coefficient.

[0110] The error rate originates from the prediction deviation of the power generation and transmission forecaster, i.e., the proportion of deviation between the predicted power generation and transmission parameters and the actual values. By statistically summing the error rates of power generation and transmission parameters within the current prediction period, this sum is incorporated as a compensation coefficient into the difference correction process to eliminate the implicit impact of prediction errors on model difference.

[0111] For example, within a certain prediction period, the power generation parameter output by the power generation and transmission predictor is 1050kW, while the actual power generation parameter is 1030kW, with an error rate of 1.9% (|1050-1030| / 1030×100%≈1.9%); the predicted transmission parameter is 980kW, while the actual transmission parameter is 960kW, with an error rate of 2.1% (|980-960| / 960×100%≈2.1%). Then, the compensation coefficient is 1.9%+2.1%=4%.

[0112] Furthermore, the twin difference degree is compensated according to the compensation coefficient to obtain the compensated twin difference degree.

[0113] Specifically, the twin difference degree is added to the compensation coefficient to correct the problem that the model difference caused by prediction error is ignored, so as to ensure that the difference degree can truly reflect the comprehensive deviation between the photovoltaic module and the virtual model.

[0114] For example, if the twin difference is 10% and the compensation coefficient is 5%, then the compensated twin difference is 10% + 5% = 15%; if the twin difference is 8% and the compensation coefficient is 3%, then the compensated twin difference is 11%.

[0115] This step, by calculating and compensating for the differences between the historical photovoltaic energy flow model and the real-time photovoltaic energy flow model, not only accurately captures the dynamic changes of the model over time, but also fully considers the implicit biases caused by prediction errors, effectively improving the reliability of the difference data.

[0116] For example, the twin discrepancy between the real-time photovoltaic energy flow model and the historical model of a photovoltaic power station is 8%, and the total error rate of the concurrent power generation and transmission forecast is 4%. After compensation, the compensated twin discrepancy is 12%. This result is consistent with the deviation (11.8%) of physical state changes caused by sudden environmental changes in actual photovoltaic modules, verifying the accuracy of the compensation treatment.

[0117] The damage prediction and identification management module 04 is used to acquire historical light parameter distribution sequences, historical cloud shadow parameter distribution sequences, and historical temperature distribution sequences, combine them with the compensated twin difference degree, predict the damage probability of photovoltaic modules, obtain the damage probability, identify the real-time photovoltaic energy flow model, and manage it.

[0118] In this embodiment of the application, in order to accurately assess the damage risk of photovoltaic modules and achieve targeted management of real-time photovoltaic energy flow models, it is necessary to combine historical environmental parameter sequences and compensation twin differences to predict the damage probability through multi-dimensional analysis, thereby completing model identification and providing a scientific basis for photovoltaic module operation and maintenance decisions.

[0119] Specifically, the historical illumination parameter distribution sequence, historical cloud shadow parameter distribution sequence, and historical temperature distribution sequence are first obtained within the most recent preset time range.

[0120] The preset time range can be set according to the damage accumulation characteristics of photovoltaic modules. The historical parameter distribution sequence contains dynamic change data of sunlight, cloud shadows and temperature in each time period, which can reflect the long-term impact of environmental factors on the modules.

[0121] Furthermore, the aforementioned historical parameter distribution sequence is input into the photovoltaic energy fluctuation analyzer, and the energy fluctuation degree is output. This energy fluctuation degree is used to quantify the overall fluctuation range of environmental parameters, reflecting the degree of energy flow instability caused by changes in sunlight, cloud cover, and temperature. The higher the value, the more significant the impact of the environment on the module.

[0122] Based on this, energy fluctuation degree is used to correct the compensation twin difference degree, thus obtaining the corrected twin difference degree. By eliminating the influence of natural environmental fluctuations on the differences in the photovoltaic energy flow model, the difference degree focuses more on the deviations caused by changes in the state of the photovoltaic modules themselves, thereby improving the accuracy of subsequent damage prediction.

[0123] Meanwhile, based on the historical cloud shadow parameter distribution sequence and the historical temperature distribution sequence, the cumulative maximum difference cloud shadow parameter and the cumulative maximum difference temperature are calculated.

[0124] Among them, the cumulative maximum difference cloud shadow parameter is the difference between the maximum and minimum values ​​of the cloud shadow parameter within a preset time, and the cumulative maximum difference temperature is the maximum difference in temperature during the same period. Both are used to evaluate the cumulative damage effect of cloud shadow shading and sudden temperature changes on photovoltaic modules.

[0125] Furthermore, a first damage probability is obtained based on the cumulative maximum difference cloud shadow parameter and the cumulative maximum difference temperature classification, and a second damage probability is obtained based on the corrected twin difference classification.

[0126] The first damage probability focuses on the risk of physical damage caused by sudden changes in environmental parameters, while the second damage probability reflects the potential damage caused by the performance degradation of photovoltaic modules behind the differences in photovoltaic energy flow models; the two are combined to obtain the final damage probability.

[0127] Finally, the real-time photovoltaic energy flow model is labeled according to the damage probability to enable targeted management. Color coding is used for labeling to facilitate quick identification of photovoltaic modules corresponding to models with high damage risk, allowing for timely maintenance.

[0128] This step, by integrating historical environmental parameters with model difference data, predicts the probability of damage from both environmental and performance change dimensions, and achieves refined management through model identification. This not only ensures the comprehensiveness of damage assessment, but also provides precise guidance for the efficient operation and maintenance of photovoltaic modules, enhancing the practical value of dynamic energy flow simulation management.

[0129] The damage prediction and identification management module 04 in the system provided in this application embodiment includes:

[0130] Obtain the historical solar radiation parameter distribution sequence, historical cloud shadow parameter distribution sequence, and historical temperature distribution sequence within the most recent preset time range, input them into the photovoltaic energy fluctuation analyzer, and output the energy fluctuation degree.

[0131] The energy fluctuation degree is used to correct the compensated twin difference degree to obtain the corrected twin difference degree;

[0132] Based on the historical cloud shadow parameter distribution sequence and the historical temperature distribution sequence, the cumulative maximum difference cloud shadow parameter and the cumulative maximum difference temperature are calculated.

[0133] The first damage probability is obtained by classifying the cumulative maximum difference cloud shadow parameter and the cumulative maximum difference temperature. The second damage probability is obtained by classifying the corrected twin difference degree. The damage probability is calculated and the real-time photovoltaic energy flow model is identified.

[0134] In this embodiment of the application, in order to accurately assess the damage risk of photovoltaic modules caused by environmental fluctuations and changes in their own state, and to achieve targeted management of the real-time photovoltaic energy flow model, it is necessary to combine historical environmental parameter sequences and compensated twin differences, and through multi-dimensional analysis and correction, finally predict the damage probability and complete model identification, so as to provide a scientific basis for the operation and maintenance decision of photovoltaic modules.

[0135] Specifically, the historical illumination parameter distribution sequence, historical cloud shadow parameter distribution sequence, and historical temperature distribution sequence are first obtained within the most recent preset time range.

[0136] The preset time range can be set according to the damage accumulation characteristics of photovoltaic modules, such as the past 72 hours or the past week. The historical parameter distribution sequence includes dynamic change data of sunlight, cloud shadows and temperature in each time period to comprehensively reflect the long-term impact of environmental factors on the modules.

[0137] Furthermore, the historical parameter distribution sequence is input into the photovoltaic energy fluctuation analyzer, and the energy fluctuation degree is obtained as the output.

[0138] The training steps for the "photovoltaic energy fluctuation analyzer" in the system provided in this application embodiment include:

[0139] Based on the historical operating data of photovoltaic modules, sample light parameter distribution sequence sets, sample cloud shadow parameter distribution sequence sets, and sample temperature distribution sequence sets are collected, combined to obtain a sample input dataset, and the average energy fluctuation amplitude after different sample input data is collected and labeled to obtain a sample energy fluctuation set.

[0140] Construct a photovoltaic energy fluctuation analyzer based on machine learning;

[0141] The photovoltaic energy fluctuation analyzer is trained under supervision using the sample input dataset and sample energy fluctuation set, and training is completed after convergence.

[0142] In this embodiment of the application, in order to ensure that the photovoltaic energy fluctuation analyzer can accurately quantify the magnitude of energy flow changes caused by environmental parameter fluctuations, the analyzer needs to be built and trained based on historical operating data. The mapping relationship between the environmental parameter sequence and the energy fluctuation degree is learned through machine learning algorithms to achieve accurate prediction of energy fluctuation degree.

[0143] Specifically, the first step is to collect and label sample data. Operational data under different time periods and environmental conditions are extracted from the historical operation records of photovoltaic modules. Sample sets of light parameter distribution sequences, cloud shadow parameter distribution sequences, and temperature distribution sequences are collected, and these are combined to form the sample input dataset.

[0144] Simultaneously, the average fluctuation amplitude of energy flow within the corresponding time period is collected, and these fluctuation data are organized and labeled to form a sample energy fluctuation set.

[0145] For example, historical operational data from the past two years is collected, divided into quarterly periods, with 1000 samples collected per quarter. The sample input dataset contains the distribution sequence of light, cloud cover, and temperature over three days for each sample, and the sample energy fluctuation set is the average fluctuation amplitude of energy flow within the corresponding time period, such as 5%, 3%, 6%, etc.

[0146] Furthermore, a machine learning-based photovoltaic energy fluctuation analyzer is constructed. This analyzer uses a long short-term memory network model as its basic framework, leveraging its advantage in processing time-series data to capture the correlation between the dynamic changes in environmental parameter sequences and energy volatility.

[0147] Specifically, the model structure includes an input layer, an LSTM layer, a fully connected layer, and an output layer. The input layer receives the temporal features of the sample input dataset, the LSTM layer extracts the long-term dependencies in the sequence, the fully connected layer maps the features into a fixed-dimensional vector, and the output layer outputs the predicted value of energy volatility.

[0148] Furthermore, the photovoltaic energy fluctuation analyzer is trained under supervision using a sample input dataset and a sample energy fluctuation set. The sample input dataset is used as the input training data, and the sample energy fluctuation set is used as the output supervision data. The prediction error is minimized by iteratively adjusting the model parameters.

[0149] During training, mean squared error was used as the loss function, and the initial learning rate was set to 0.001. Every 50 iterations, the learning rate was reduced to 1 / 10 of the previous rate.

[0150] Meanwhile, the sample data is divided into training and validation sets in a 7:3 ratio. The validation set is used to monitor the model performance in real time. If the loss function value of the validation set no longer decreases after 20 consecutive rounds, the model is considered to have converged and training is stopped.

[0151] For example, the photovoltaic energy fluctuation analyzer converged after 150 rounds of training, and the average deviation between the predicted energy fluctuation of the validation set and the actual labeled value was controlled within 0.3%, which met the accuracy requirements.

[0152] Finally, by inputting the historical solar radiation parameter distribution sequence, historical cloud shadow parameter distribution sequence, and historical temperature distribution sequence within the most recent preset time range into the trained photovoltaic energy fluctuation analyzer, the corresponding energy fluctuation degree can be obtained as output, providing an accurate quantitative basis for subsequent correction of twin differences.

[0153] Furthermore, the energy fluctuation degree output by the photovoltaic energy fluctuation analyzer is used to correct the twin difference degree to obtain the corrected twin difference degree.

[0154] Specifically, the formula for calculating the corrected twin difference can be expressed as "Corrected twin difference = Compensated twin difference - Energy fluctuation". By eliminating the interference of natural fluctuations in environmental parameters on model differences, the model bias that may be caused by abnormal factors such as hot spot damage to photovoltaic modules is highlighted, so that the corrected twin difference more accurately reflects the abnormal changes in the state of the module itself.

[0155] For example, if the compensation twin difference is 16% and the energy fluctuation output by the photovoltaic energy fluctuation analyzer is 4.2%, then the corrected twin difference is 16% - 4.2% = 11.8%. This value can focus on the model bias caused by damage to photovoltaic modules such as hot spot effect.

[0156] Meanwhile, based on the historical cloud shadow parameter distribution sequence and the historical temperature distribution sequence, the cumulative maximum difference cloud shadow parameter and the cumulative maximum difference temperature are calculated.

[0157] Among them, the cumulative maximum difference cloud shadow parameter is the difference between the maximum and minimum values ​​in the historical cloud shadow parameter distribution sequence within a preset time range, used to quantify the extreme fluctuation range of cloud shadow parameters; the cumulative maximum difference temperature is the difference between the maximum and minimum values ​​in the historical temperature distribution sequence during the same period, used to measure the extreme range of temperature changes. Combining the two can assess the cumulative impact of sudden changes in environmental factors on photovoltaic modules, especially the risk of hot spot effects caused by the combined effects of local shading and temperature changes due to rapid changes in cloud shadow.

[0158] For example, in a 3-day historical cloud shadow parameter distribution sequence, the maximum cloud shadow coverage is 45% and the minimum is 5%, then the cumulative maximum difference in cloud shadow parameters is 45% - 5% = 40%; in a historical temperature distribution sequence, the maximum component surface temperature is 43℃ ​​and the minimum is 26℃, then the cumulative maximum difference in temperature is 43℃ ​​- 26℃ = 17℃.

[0159] Based on this, the first damage probability is obtained by classifying the cumulative maximum difference cloud shadow parameters and the cumulative maximum difference temperature, and the second damage probability is obtained by classifying the corrected twin difference degree, and then the damage probability is calculated.

[0160] The system provided in this application embodiment includes the following steps: "obtaining a first damage probability based on the cumulative maximum difference cloud shadow parameter and the cumulative maximum difference temperature, obtaining a second damage probability based on the corrected twin difference degree, and calculating the damage probability".

[0161] The cumulative maximum difference cloud shadow parameter and the cumulative maximum difference temperature are input into the damage probability classifier to classify and obtain the first damage probability. The damage probability classifier is constructed based on a decision tree and is constructed using the sample cumulative maximum difference cloud shadow parameter set, the sample cumulative maximum difference temperature set, and the sample first damage probability set.

[0162] Based on historical photovoltaic power generation data, the proportion of photovoltaic modules damaged when the corrected twin difference occurs is obtained, and a second damage probability is obtained.

[0163] The damage probability is calculated based on the first damage probability and the second damage probability.

[0164] In this embodiment of the application, in order to comprehensively assess the damage risk of photovoltaic modules from two dimensions, namely changes in environmental parameters and abnormal states of the modules themselves, it is necessary to obtain a damage probability that can comprehensively reflect the possibility of module damage through classification calculation and fusion processing, so as to provide an accurate basis for the identification and operation and maintenance decision of real-time photovoltaic energy flow model.

[0165] First, calculate the first damage probability. The cumulative maximum difference cloud shadow parameters and cumulative maximum difference temperature are input into the damage probability classifier, and the first damage probability is obtained through classification by the classifier.

[0166] The damage probability classifier is built based on the decision tree algorithm in the existing technology. Its core is to learn the association rules between changes in environmental parameters and component damage through historical sample data.

[0167] Specifically, the first step is to collect sample data. This involves extracting the cumulative maximum difference cloud shadow parameter set and the cumulative maximum difference temperature set for different time periods from the historical operation records of photovoltaic modules, and simultaneously labeling the damage probability of photovoltaic modules caused by changes in environmental parameters within the corresponding time periods to form the first damage probability set of the samples.

[0168] For example, historical operating data of photovoltaic modules over the past three years were collected, divided into monthly periods, and 800 valid samples were extracted for each month. For each sample, the statistical period was 72 hours, and the cumulative maximum difference in cloud shadow parameters and the cumulative maximum difference in temperature within that period were calculated.

[0169] At the same time, by combining the infrared detection report and operation and maintenance records for that period, the probability of hot spot damage caused by cloud shadows and temperature changes was marked.

[0170] For example, if the cloud cover rate of a certain sample group is as high as 50% and as low as 8% within 72 hours, the cumulative maximum difference cloud cover parameter is 42%. The highest surface temperature of the component during the same period is 45°C and the lowest is 24°C, so the cumulative maximum difference temperature is 21°C. In addition, 6 mild hot spots and 2 moderate hot spots appear among 100 components in this area. Then the first damage probability corresponding to this sample group is marked as 8% ((6+2) / 100×100%).

[0171] In addition, in another set of samples, the cloud shadow coverage rate was 20% at the highest and 5% at the lowest within 72 hours, the cumulative maximum difference cloud shadow parameter was 15%, the component surface temperature was 35℃ at the highest and 28℃ at the lowest, the cumulative maximum difference temperature was 7℃, and no hot spot damage occurred in this area, so the probability of the first damage was marked as 0%.

[0172] Finally, by extracting the cumulative maximum difference cloud shadow parameter and the cumulative maximum difference temperature from all monthly samples in the above manner and labeling the corresponding first damage probability, a sample first damage probability set containing the first damage probability values ​​of each sample is formed.

[0173] For example, the first damage probability set of this sample includes different values ​​such as 8%, 0%, 5%, and 12%, which correspond to the hot spot damage probability under different combinations of cumulative maximum difference cloud shadow parameters and cumulative maximum difference temperature, providing single-dimensional output supervision data for subsequent model training.

[0174] Furthermore, the decision tree model is trained using the cumulative maximum difference cloud shadow parameter set and the cumulative maximum difference temperature set as input features, and the first damage probability set of the samples as output labels, in order to construct a damage probability classifier.

[0175] Specifically, the CART (Classification and Regression Tree) algorithm is used as the basic architecture to construct a decision tree model. This architecture constructs a tree structure through recursive splitting nodes. Each internal node corresponds to a splitting threshold for a feature (cumulative maximum difference cloud shadow parameter or cumulative maximum difference temperature), and the leaf nodes correspond to the classification result of the first damage probability.

[0176] Furthermore, the cumulative maximum difference cloud shadow parameter set, the cumulative maximum difference temperature set, and the first damage probability set of the samples are divided in a ratio of 8:1:1. For example, 2400 sets are selected from 3000 sets of samples as the training set, 300 sets as the validation set, and 300 sets as the test set to ensure that the samples cover the environmental parameter characteristics under different seasons and weather conditions, and to ensure the comprehensiveness and representativeness of the data distribution.

[0177] Furthermore, during the model training phase, the cumulative maximum difference cloud shadow parameters and cumulative maximum difference temperature of the training set samples are input into the model. The optimal splitting features and thresholds are determined by calculating the Gini impurity, and the decision tree structure is gradually constructed.

[0178] Meanwhile, using the classification error between the predicted first damage probability and the sample label value as the optimization objective, overfitting risk is reduced through pruning operations (pre-pruning limits the tree depth to 8 layers, and post-pruning removes redundant branches), thus gradually reducing the classification error. A minimum number of split samples is set at 20; splitting stops when the number of samples at a node falls below this value.

[0179] Meanwhile, the model performance is monitored in real time using the validation set. If the classification accuracy of the validation set increases by less than 0.5% over 15 consecutive rounds, the model is considered to have converged and training is stopped.

[0180] Furthermore, during the testing phase, the test set is input into the trained model. If the classification accuracy of the first damage probability is not less than 85%, and the recall rate of the high-risk category (first damage probability ≥ 10%) is not less than 80%, then the damage probability classifier is deemed qualified and can be put into use.

[0181] Finally, the trained damage probability classifier can accurately output the corresponding first damage probability based on the input cumulative maximum difference cloud shadow parameters and cumulative maximum difference temperature.

[0182] Furthermore, the second damage probability is calculated. That is, based on the photovoltaic power generation data over a historical period, the proportion of damage to photovoltaic modules when the twin discrepancy occurs is statistically corrected, and this is used as the second damage probability.

[0183] Specifically, a dataset containing records of corrected twin discrepancies and corresponding module damage states is extracted from the historical operation database of photovoltaic modules. First, historical records with deviations from the current corrected twin discrepancy value within ±1% are selected to form a related sample group.

[0184] Furthermore, the second damage probability is calculated as the ratio of the number of confirmed damage records (such as hot spots and performance degradation) of photovoltaic modules in the associated sample group to the total number of records in the sample group. The specific calculation formula can be expressed as: "Second damage probability = (Number of damage records in the associated sample group / Total number of records in the associated sample group) × 100%".

[0185] For example, if the current corrected twin difference is 11%, and 200 records with a corrected twin difference in the range of 10%-12% are selected from the historical operating data, and 25 of these records correspond to components with damage, then the probability of the second damage is 25 / 200×100%=12.5%.

[0186] Based on this, the final damage probability is calculated according to the first damage probability and the second damage probability. That is, a weighted summation method is used for fusion, and the specific calculation formula can be expressed as "damage probability = (first damage probability × environmental induced damage weight) + (second damage probability × component self-induced damage weight)".

[0187] The first damage probability reflects the damage risk caused by changes in environmental parameters, while the second damage probability reflects the damage risk caused by abnormal component status. The weights can be set according to the proportion of the two types of damage causes in historical operation and maintenance data. For example, the weight of environmental induced damage is 0.4, and the weight of component self-induced damage is 0.6.

[0188] For example, if the first damage probability is 9% and the second damage probability is 12.5%, then the final damage probability is 9% × 0.4 + 12.5% ​​× 0.6 = 3.6% + 7.5% = 11.1%. This result comprehensively reflects the damage probability of photovoltaic modules under the combined effect of changes in environmental parameters and abnormal states of the modules themselves, providing a quantitative basis for the identification of real-time photovoltaic energy flow models.

[0189] Furthermore, the real-time photovoltaic energy flow model is labeled, and targeted management is carried out based on the labeling results. That is, based on the calculated damage probability, the real-time photovoltaic energy flow model is labeled with a risk level to intuitively distinguish the damage risk of photovoltaic modules corresponding to different models, providing clear guidance for operation and maintenance management.

[0190] Specifically, a multi-level labeling rule is set based on the numerical range of the damage probability. For example, when the damage probability is <5%, it is labeled "green - low risk", indicating that the current damage risk of the photovoltaic module is extremely low and can be monitored according to the regular cycle; when the damage probability is 5% ≤ damage probability <15%, it is labeled "yellow - medium risk", indicating that the inspection frequency of photovoltaic modules in the corresponding area should be increased, and hot spot signs should be given special attention; when the damage probability is ≥15%, it is labeled "red - high risk", and special testing should be arranged immediately and an emergency maintenance plan should be formulated to avoid the damage from expanding.

[0191] Simultaneously, the identification results are associated with and stored in the real-time photovoltaic energy flow model to form a complete record containing model parameters, damage probability, risk identification, and generation time, which is then incorporated into the photovoltaic energy flow model management database.

[0192] In addition, the management platform can automatically filter high-risk photovoltaic energy flow models based on the identification results and push them to the operation and maintenance terminal to help operation and maintenance personnel prioritize the handling of high-risk areas. At the same time, it can regularly count the distribution of photovoltaic energy flow models with different risk levels and generate damage risk trend reports to provide data support for the long-term maintenance strategy of photovoltaic modules.

[0193] For example, when the damage probability corresponding to a real-time photovoltaic energy flow model is 11.1%, it is marked as "yellow-medium risk" according to the preset multi-level identification rules. The photovoltaic module area associated with the model is automatically added to the key inspection list, and the energy flow path of the area is highlighted in yellow in the virtual simulation interface. This makes it easier for operation and maintenance personnel to quickly locate the component locations that need attention, and realizes refined management of the photovoltaic energy flow model.

[0194] The embodiments of this application, through the specific implementation methods described above, achieve the following technical effects:

[0195] This application proposes a dynamic simulation management system for energy flow based on digital twins. First, the monitoring area of ​​the photovoltaic module environment is determined. Sensors are deployed to acquire the distribution of irradiance, cloud shadow, and temperature parameters in each monitoring area. These environmental parameters are input into a pre-constructed power generation and transmission predictor to obtain accurate power generation and transmission parameters. Next, based on digital twin technology, a photovoltaic energy flow model is constructed according to the power generation and transmission parameters and dynamically updated at a preset frequency to form a continuous sequence of photovoltaic energy flow models, achieving real-time mapping of the energy flow. Then, historical and real-time photovoltaic energy flow models are extracted from this model sequence, and the twin difference between the two is calculated. This difference is then compensated by incorporating the prediction error rate of the power generation and transmission parameters to obtain a compensated twin difference. Simultaneously, energy fluctuations are analyzed using historical environmental parameter sequences to correct the compensated twin difference. Combined with the cumulative maximum difference in cloud shadow parameters and the cumulative maximum difference in temperature, the first and second damage probabilities are predicted respectively. These are then fused to obtain the final damage probability. Finally, the real-time photovoltaic energy flow model is identified and managed to achieve precise operation and maintenance.

[0196] The system provided in this application, through the technical solution of "environmental monitoring - model building - difference compensation - damage prediction - identification management", solves the problem that traditional photovoltaic power generation digital simulation management only focuses on real-time updates of power generation status and does not pay attention to faults caused by photovoltaic hot spot effects. It realizes dynamic simulation of photovoltaic energy flow and precise control of damage risks, providing support for improving the energy efficiency and safety of photovoltaic systems.

[0197] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0198] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0199] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A dynamic simulation management system for energy flow based on digital twins, characterized in that, The system includes: The monitoring and power generation / transmission prediction module is used to monitor and acquire the distribution of irradiance parameters, cloud shadow parameters, and temperature distribution of the photovoltaic module environment, and to predict and acquire the power generation and transmission parameters of the photovoltaic module. The dynamic twin modeling module is used to construct a dynamically updated photovoltaic energy flow model based on digital twins according to the power generation parameters and transmission parameters, and obtain a photovoltaic energy flow model sequence. The twin discrepancy compensation module is used to process and obtain the twin discrepancy between the historical photovoltaic energy flow model and the real-time photovoltaic energy flow model within the photovoltaic energy flow model sequence, and to compensate for the twin discrepancy, including: Historical photovoltaic energy flow models and real-time photovoltaic energy flow models are extracted from the photovoltaic energy flow model sequence; Calculate the difference between the real-time photovoltaic energy flow model and the historical photovoltaic energy flow model to obtain the twin difference. Obtain the error rate of the predicted power generation parameters and transmission parameters, and use it as a compensation coefficient; The twin difference is compensated according to the compensation coefficient to obtain the compensated twin difference. The damage prediction and identification management module is used to acquire historical light parameter distribution sequences, historical cloud shadow parameter distribution sequences, and historical temperature distribution sequences, and combine them with the compensated twin difference degree to predict the damage probability of photovoltaic modules, obtain the damage probability, identify the real-time photovoltaic energy flow model, and manage it.

2. The energy flow dynamic simulation management system based on digital twins according to claim 1, characterized in that, Monitoring and acquiring cloud shadow parameter distribution and photovoltaic module temperature distribution within the photovoltaic module environment, and predicting and acquiring photovoltaic module power generation and transmission parameters, including: Identify multiple monitoring areas within the photovoltaic module environment; The system monitors and acquires illumination parameters, cloud shadow parameters, and temperature within multiple monitoring areas, obtaining the distribution of illumination parameters, cloud shadow parameters, and temperature. The illumination parameter distribution, cloud shadow parameter distribution, and temperature distribution are input into a pre-constructed power generation and transmission predictor to obtain power generation parameters and transmission parameters.

3. The energy flow dynamic simulation management system based on digital twins according to claim 2, characterized in that, The steps for pre-constructing the power generation and transmission predictor include: Based on the photovoltaic module operation data over a historical period, sample light parameter distribution sets, sample cloud shadow parameter distribution sets, and sample temperature distribution sets were collected. The power generation and transmission of the photovoltaic modules were also collected, and the sample power generation parameter sets and sample transmission parameter sets were obtained by labeling. Based on machine learning, power generation prediction branch and power transmission prediction branch are constructed; Using the sample illumination parameter distribution set, sample cloud shadow parameter distribution set, and sample temperature distribution set as input training data, and the sample power generation parameter set and sample power transmission parameter set as output supervision data, respectively, the power generation prediction branch and the power transmission prediction branch are iteratively supervised and trained. After convergence, the power generation and transmission predictor is obtained.

4. The energy flow dynamic simulation management system based on digital twins according to claim 1, characterized in that, Based on the power generation and transmission parameters, a dynamically updated photovoltaic energy flow model is constructed using digital twins, resulting in a photovoltaic energy flow model sequence, including: Based on the power generation and transmission parameters, a photovoltaic energy flow model is constructed using digital twins. The photovoltaic energy flow model is dynamically updated according to a preset frequency to obtain a photovoltaic energy flow model sequence.

5. The energy flow dynamic simulation management system based on digital twins according to claim 1, characterized in that, Obtain historical illumination parameter distribution sequences, historical cloud shadow parameter distribution sequences, and historical temperature distribution sequences. Combine these with the compensated twin difference degree to predict the damage probability of photovoltaic modules, obtain the damage probability, and identify the real-time photovoltaic energy flow model, including: Obtain the historical solar radiation parameter distribution sequence, historical cloud shadow parameter distribution sequence, and historical temperature distribution sequence within the most recent preset time range, input them into the photovoltaic energy fluctuation analyzer, and output the energy fluctuation degree. The energy fluctuation degree is used to correct the compensated twin difference degree to obtain the corrected twin difference degree; Based on the historical cloud shadow parameter distribution sequence and the historical temperature distribution sequence, the cumulative maximum difference cloud shadow parameter and the cumulative maximum difference temperature are calculated. The first damage probability is obtained by classifying the cumulative maximum difference cloud shadow parameter and the cumulative maximum difference temperature. The second damage probability is obtained by classifying the corrected twin difference degree. The damage probability is calculated and the real-time photovoltaic energy flow model is identified.

6. The energy flow dynamic simulation management system based on digital twins according to claim 5, characterized in that, The training steps for the photovoltaic energy fluctuation analyzer include: Based on the historical operating data of photovoltaic modules, sample light parameter distribution sequence sets, sample cloud shadow parameter distribution sequence sets, and sample temperature distribution sequence sets are collected, combined to obtain a sample input dataset, and the average energy fluctuation amplitude after different sample input data is collected and labeled to obtain a sample energy fluctuation set. Construct a photovoltaic energy fluctuation analyzer based on machine learning; The photovoltaic energy fluctuation analyzer is trained under supervision using the sample input dataset and sample energy fluctuation set, and training is completed after convergence.

7. The energy flow dynamic simulation management system based on digital twins according to claim 5, characterized in that, A first damage probability is obtained based on the cumulative maximum difference cloud shadow parameter and the cumulative maximum difference temperature. A second damage probability is obtained based on the corrected twin difference degree. The damage probability is then calculated, including: The cumulative maximum difference cloud shadow parameter and the cumulative maximum difference temperature are input into the damage probability classifier to classify and obtain the first damage probability. The damage probability classifier is constructed based on a decision tree and is constructed using the sample cumulative maximum difference cloud shadow parameter set, the sample cumulative maximum difference temperature set, and the sample first damage probability set. Based on historical photovoltaic power generation data, the proportion of photovoltaic modules damaged when the corrected twin difference occurs is obtained, and a second damage probability is obtained. The damage probability is calculated based on the first damage probability and the second damage probability.

Images