Method and device for detecting self-reduced load operation of a photovoltaic installation

Abstract

The application provides a photovoltaic device self-load reduction operation detection method and device, relates to the photovoltaic power generation technical field, and comprises the following steps: acquiring photovoltaic multidimensional data; dynamically calibrating a single diode physical model based on the photovoltaic multidimensional data, training a deep learning model in combination with photovoltaic physical characteristics, and obtaining dynamic reference power; optimizing random forest model ensemble learning based on the photovoltaic multidimensional data, classifying load reduction characteristics of the photovoltaic device, and obtaining load reduction type results; calculating an initial loss value based on the dynamic reference power and real-time power, correcting the initial loss value, and constructing power loss value; based on the power loss value and the load reduction type results, combining edge computing nodes and cloud model real-time interaction, obtaining an optimized cloud model; and detecting the photovoltaic device in real time according to the optimized cloud model, and triggering real-time alarm when the detection result does not meet a preset performance threshold. The application solves the problem of power station operation and maintenance response lag.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic power generation, in particular to a detection method and device for self-load reduction operation of photovoltaic equipment. BACKGROUND

[0002] In the photovoltaic power generation technology, the traditional monitoring and diagnosis method mainly relies on single electrical data or static model to carry out power monitoring and equipment state detection work. However, this method has significant limitations and is difficult to accurately distinguish between active load reduction and passive load reduction. Influenced by this, the photovoltaic light abandonment rate in the northwest region in 2023 is high, reaching 7.3%. At present, the existing photovoltaic equipment lacks the ability to accurately diagnose and cannot accurately identify the running state of the equipment self-load reduction. Due to the inability to obtain real-time operation information of the equipment, it is also impossible to quickly take effective optimization measures. This leads to the problem of delayed operation and maintenance response of the power station.

[0003] Therefore, there is an urgent need for a detection method and device for self-load reduction operation of photovoltaic equipment to solve the problem of delayed operation and maintenance response of the power station. SUMMARY

[0004] The purpose of the present application is to provide a detection method and device for self-load reduction operation of photovoltaic equipment to improve the above problems. In order to achieve the above purpose, the technical solution adopted by the present application is as follows:

[0005] In a first aspect, the present application provides a detection method for self-load reduction operation of photovoltaic equipment, comprising:

[0006] Obtaining photovoltaic multi-dimensional data, the photovoltaic multi-dimensional data including meteorological data, electrical data, environmental data and equipment health data;

[0007] Calibrating a single diode physical model based on the photovoltaic multi-dimensional data, and training a deep learning model based on the photovoltaic physical characteristics to obtain a dynamic reference power;

[0008] Based on the photovoltaic multi-dimensional data, optimizing the random forest model ensemble learning, classifying the load reduction characteristics of the photovoltaic equipment through the optimized random forest model, and obtaining a load reduction type result;

[0009] Based on the dynamic reference power and the real-time power in the photovoltaic multi-dimensional data, calculating an initial loss value, and constructing an electricity loss value by multi-factor correction of the initial loss value;

[0010] Based on the electricity loss value and the load reduction type result, combining the edge computing node and the cloud model for real-time interaction to obtain an optimized cloud model;

[0011] The photovoltaic equipment is monitored in real time based on the optimized cloud model. When the monitoring results do not meet the preset performance threshold, a real-time alarm is triggered.

[0012] Secondly, this application also provides a detection device for photovoltaic equipment operating under self-derating load, comprising:

[0013] The acquisition module is used to acquire photovoltaic multidimensional data, which includes meteorological data, electrical data, environmental data, and equipment health data.

[0014] The calibration module is used to dynamically calibrate the physical model of a single diode based on photovoltaic multidimensional data, and to perform transfer learning training on the deep learning model in combination with photovoltaic physical characteristics to obtain a dynamic reference power.

[0015] The optimization module is used to optimize the random forest model through ensemble learning based on the photovoltaic multidimensional data. The optimized random forest model is used to classify the load reduction characteristics of photovoltaic equipment to obtain the load reduction type results.

[0016] The calculation module is used to calculate the initial loss value based on the dynamic reference power and the real-time power in the photovoltaic multidimensional data, and to construct the power loss value by performing multi-factor correction on the initial loss value;

[0017] The interaction module is used to obtain an optimized cloud model by combining the power loss value and load reduction type results with the real-time interaction between the edge computing node and the cloud model.

[0018] The detection module is used to perform real-time detection on photovoltaic equipment based on the optimized cloud model. When the detection result does not meet the preset performance threshold, a real-time alarm is triggered.

[0019] The beneficial effects of this invention are as follows:

[0020] This invention uses photovoltaic multidimensional data to dynamically calibrate a single-diode physical model for accurate power generation prediction, avoiding the baseline deviation caused by environmental fluctuations in traditional static models. It optimizes a random forest model using photovoltaic multidimensional data, classifying load shedding characteristics and accurately distinguishing potential load shedding types during operation, thus preventing wasted maintenance resources. Based on dynamic baseline power and real-time power from photovoltaic multidimensional data, it calculates initial loss values ​​and applies multi-factor corrections to reflect the true operating status and loss situation of the photovoltaic system. This solves the problem of existing technologies only calculating power loss, ignoring peak-valley electricity price differences and component lifespan degradation, resulting in significant deviations in economic loss assessment. In summary, this invention solves the problem of delayed power plant operation and maintenance response.

[0021] Other features and advantages of the present application will be set forth in the following description, and in part will be apparent from the description, or can be learned by practice of the application. The objects and other advantages of the present application will be realized and attained by the structure particularly pointed out in the written description and claims thereof as well as the appended drawings. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be considered as limiting the scope. For those skilled in the art, other related drawings can also be obtained without creative labor.

[0023] Figure 1 Flow chart of the detection method for self-reducing load operation of the photovoltaic device described in the embodiments of the present application;

[0024] Figure 2 Structural schematic diagram of the detection device for self-reducing load operation of the photovoltaic device described in the embodiments of the present application.

[0025] In the figure, the marks are: 800, the detection device for self-reducing load operation of the photovoltaic device; 801, the processor; 802, the memory; 803, the multimedia component; 804, the I / O interface; 805, the communication component. DETAILED DESCRIPTION

[0026] In order to make the objects, technical solutions and advantages of the embodiments of the present application more clear, the following will combine the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments. The components of the embodiments of the present application described and shown in the drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0027] It should be noted that: similar labels and letters represent similar items in the following drawings, therefore, once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. Meanwhile, in the description of the present application, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0028] Embodiment 1: The embodiment provides a detection method for self-load reduction operation of a photovoltaic device.

[0029] Referring to Figure 1 , the method comprises steps S1 to S6, comprising:

[0030] S1: acquiring photovoltaic multidimensional data, wherein the photovoltaic multidimensional data comprises meteorological data, electrical data, environmental data and device health data;

[0031] In this step, the photovoltaic multidimensional data is detected by an irradiance sensor, a temperature sensor, a current / voltage collector and an inverter state monitoring unit.

[0032] The meteorological data comprises temperature, light intensity, wind speed and humidity, the electrical data comprises current, voltage and power, the environmental data comprises dust concentration and pollutant concentration, and the device health data comprises device operation time, fault record and maintenance record.

[0033] S2: dynamically calibrating a single-diode physical model based on the photovoltaic multidimensional data, and performing transfer learning training on a deep learning model combined with photovoltaic physical characteristics to obtain a dynamic reference power;

[0034] In this step, the single-diode physical model and the deep learning model are used to enhance the generalization ability in complex environments, so that a high prediction accuracy can be maintained even in extreme weather or sudden device aging.

[0035] To clarify the specific acquisition method of the dynamic reference power, steps S21 to S23 are included in step S2, and specifically:

[0036] S21: dynamically calibrating a single-diode physical model based on real-time temperature and light intensity in the photovoltaic multidimensional data to obtain an optimized single-diode physical model;

[0037] In this step, the single-diode physical model is dynamically calibrated based on real-time temperature, light intensity, real-time irradiance, component temperature and component health in the photovoltaic multidimensional data to optimize model parameters, so that the model can accurately reflect the current photovoltaic system state. By dynamically adjusting the model parameters, the optimized single-diode physical model can more accurately reflect the operation state of the current photovoltaic system under different environmental conditions. Based on the basic physical characteristics of the photovoltaic device, the device aging and other long-term changes are adapted to avoid the reference deviation caused by environmental fluctuations of the traditional static model.

[0038] The single-diode physical model expression is:

[0039] (1);

[0040] In the above formula (1), is the current, is the reverse saturation current at real-time temperature, is the real-time temperature, is the band gap, is the temperature, is the exponential function, is the Boltzmann constant;

[0041] S22: Perform transfer learning training on the deep learning model according to the power historical data and meteorological forecast data in the photovoltaic physical characteristics, build a power prediction model, and predict future power changes through the power prediction model to obtain a predicted power value;

[0042] In this step, transfer learning is used to reduce the large amount of data and computing resources required for training the deep learning model, capture short-term environmental fluctuations, and improve the accuracy and timeliness of power prediction. It solves the problem of long training time and large data demand for deep learning model from scratch.

[0043] S23: Perform weighted fusion based on the optimized single-diode physical model and the predicted power value to generate a dynamic reference power.

[0044] In this step, the output power of the optimized single-diode physical model and the predicted power value are weighted and fused to generate a dynamic reference power. The dynamic reference power combines the accuracy of the physical model and the prediction ability of the deep learning model, and can better reflect the ideal output power of the photovoltaic system under current and future conditions.

[0045] Wherein, when the real-time power in the photovoltaic multi-dimensional data is greater than or equal to 95% of the dynamic reference power, it is determined to be normal operation, and the S3 step is entered for processing. If the real-time power is less than 95%, further analyze the environmental condition changes and equipment failures, and take maintenance or adjustment measures according to the specific situation.

[0046] S3: Optimize the random forest model ensemble learning based on the photovoltaic multi-dimensional data, and classify the load reduction types of the photovoltaic equipment through the optimized random forest model to obtain a load reduction type result;

[0047] In this step, the random forest model is used to accurately distinguish the load reduction types that may occur during operation, avoiding the problem of waste of operation and maintenance resources.

[0048] To clearly specify the specific acquisition method of the load reduction type result, steps S3 include S31 to S34, which are as follows:

[0049] S31: Divide based on the photovoltaic multi-dimensional data to obtain a training subset, the training subset including a training set and a test set;

[0050] S32: performing integrated learning optimization on the hyperparameters of the random forest model according to the grid search and the training subset, and determining an optimal hyperparameter configuration through cross-validation;

[0051] In this step, the hyperparameters of the random forest model are optimized through integrated learning according to the grid search and the training subset, and the performance of each combination is evaluated in combination with k-fold cross-validation. The optimal hyperparameter configuration is determined by combining the results of the grid search and the cross-validation, thereby solving the problem of difficulty in selecting the hyperparameters of the random forest model and avoiding poor model performance due to improper selection of hyperparameters.

[0052] The range of the hyperparameters is determined, including the number of decision trees, the maximum depth of the decision trees, the minimum number of samples required for splitting internal nodes, and the minimum number of samples for leaf nodes. The k-fold cross-validation is specifically performed as follows: the training set is divided into k subsets, k-1 subsets are used for training each time, and the remaining one subset is used for validation. After repeating k times, the average value is taken as the evaluation of the performance of the model.

[0053] S33: initializing the random forest model based on the optimal hyperparameter configuration, and generating multiple decision trees through integrated learning method for double random sampling of features and samples during the training process;

[0054] To clarify the specific acquisition method of the multiple decision trees, steps S331 to S334 are included in step S33, specifically:

[0055] S331: optimizing the hyperparameters of the random forest model according to the grid search to obtain a hyperparameter configuration;

[0056] In this step, the hyperparameter combination with the highest performance index is selected as the hyperparameter configuration for measuring the performance of the model.

[0057] S332: fitting the random forest model based on the hyperparameter configuration and the training set to obtain an optimized random forest model;

[0058] In this step, the optimized hyperparameter configuration is input into the random forest model, and the random forest model is trained through the training set data to obtain an optimized random forest model, which is used to ensure learning of the features and rules in the data.

[0059] S333: evaluating the optimized random forest model based on the test set to obtain a model performance index;

[0060] In this step, the test set is input into the optimized random forest model to obtain the prediction results. The model performance index is calculated by comparing the true labels of the test set with the prediction results, thus solving the problem of overfitting in single decision tree models.

[0061] S334: Evaluate the performance of different hyperparameters based on the model performance metrics and cross-validation, and determine the optimal hyperparameter configuration.

[0062] In this step, the performance of different hyperparameter combinations is comprehensively evaluated based on the model performance metrics and the cross-validation results. The model is then refitted using the entire training set to determine the optimal hyperparameter configuration, ensuring the model's stability and generalization ability.

[0063] S34: Based on the ensemble learning mechanism, perform load reduction feature voting on the prediction results of multiple decision trees, and output the load reduction type classification result.

[0064] In this step, the ensemble learning mechanism aggregates the prediction results of multiple decision trees. For the load reduction feature classification task, each decision tree performs classification prediction on the input sample, and selects the category with the most votes as the final load reduction type classification result.

[0065] The classification results of load reduction types include active load reduction and passive load reduction.

[0066] When the real-time power in the photovoltaic multidimensional data is less than 95% of the dynamic reference power and there is no equipment fault alarm, the active load reduction judgment is entered. When the equipment has a health alarm (such as the module degradation rate > 2% / year), the passive load reduction judgment is entered.

[0067] S4: Calculate the initial loss value based on the dynamic reference power and the real-time power in the photovoltaic multidimensional data, and construct the power loss value by performing multi-factor correction on the initial loss value;

[0068] This step is used to reflect the actual operating status and losses of the photovoltaic system.

[0069] To clarify the specific method for obtaining the power loss value, step S4 includes S41 to S44, specifically:

[0070] S41: Calculate the initial power loss value by performing time integration based on the difference between the dynamic reference power and the real-time power;

[0071] In this step, the expression for the initial power loss value is:

[0072] (2);

[0073] In the above formula (2), This is the initial power loss value. The starting time for integration. The end time of the integration process. This is the difference in integral power. denoted as a tiny time interval in the integral.

[0074] S42: Analyze the change pattern of power loss over time using the long short-term memory network model to obtain the power change rate;

[0075] In this step, the time series data of the power loss value is input into the Long Short-Term Memory (LSTM) network model for training. The hyperparameters are adjusted to optimize the model performance. The trained LSM network model is used to predict the changing trend of the power loss value and calculate the power change rate. By using the LSM network model to capture the dynamic changing pattern of the power loss value, the prediction accuracy of the power loss trend is improved, and the problem that traditional methods are difficult to accurately predict the dynamic changes of power loss is solved.

[0076] S43: Dynamically adjust the initial power loss value based on the power change rate to obtain a corrected power loss value;

[0077] In this step, a correction amount is calculated based on the power change rate, and the initial power loss value is adjusted according to the correction amount to obtain a corrected power loss value. This takes into account the dynamic changes in power loss, making the corrected power loss value closer to the actual loss, and solving the problem that the initial power loss value fails to reflect dynamic changes.

[0078] S44: The corrected power loss value is calculated by weighting the electricity price weight and subsidy coefficient for different time periods to obtain the final power loss value.

[0079] To clarify the further method for obtaining the power loss value, step S44 includes S441 to S443, specifically:

[0080] S441: Based on the electricity price weights and subsidy coefficients for different time periods, a dynamic electricity price model is constructed to measure direct economic losses.

[0081] In this step, the dynamic electricity price model accurately calculates direct economic losses based on electricity prices and subsidy policies at different times, solving the problem that traditional loss assessments do not consider changes in electricity prices and subsidies.

[0082] The dynamic electricity price model expression is as follows:

[0083] (3);

[0084] In the above formula (3), For dynamic electricity pricing models, Basic electricity price, As a weighting factor for electricity prices, This is the subsidy coefficient.

[0085] S442: Based on the lifetime loss acceleration factor and the corrected power loss value, a lifetime loss model is constructed;

[0086] In this step, the lifetime loss for each time period is calculated and constructed based on the lifetime loss acceleration factor and the corrected power loss value for different time periods, resulting in a lifetime loss model. This lifetime loss model is used to quantify the impact of power loss on equipment lifespan, solving the problem that traditional loss assessments do not consider equipment lifetime loss.

[0087] S443: Based on the dynamic electricity price model and the lifetime loss model, a weighted sum is performed to obtain the final electricity loss value.

[0088] This step comprehensively considers losses from both economic and equipment lifespan perspectives, making the final power loss value more comprehensive and accurate, and solving the problem of traditional loss assessments that only consider a single factor such as economics or equipment lifespan.

[0089] S5: Based on the power loss value and load reduction type results, and combined with the real-time interaction between the edge computing node and the cloud model, an optimized cloud model is obtained;

[0090] To clarify the specific method for optimizing the acquisition of cloud models, step S5 includes S51 to S56, specifically:

[0091] S51: Based on the power loss value and the load reduction type result, perform feature engineering analysis and extract key features;

[0092] In this step, the extraction operation specifically involves: extracting key features that have a significant impact on power loss and load reduction type;

[0093] The key features include time-domain features and frequency-domain features. The time-domain features include mean, variance, and peak value, while the frequency-domain features include power spectral density and frequency distribution.

[0094] S52: Train a gradient boosting tree based on the time-domain and frequency-domain features of the key features to obtain a regression prediction model;

[0095] In this step, the strong learning ability of gradient boosting trees is utilized to improve the accuracy of the regression prediction model, thus solving the problem of insufficient accuracy of traditional regression models when dealing with complex data relationships.

[0096] S53: Evaluate the performance of the regression prediction model. When the performance of the regression prediction model meets the prediction performance evaluation index, output the optimized regression prediction model.

[0097] In this step, the regression prediction model is used to make predictions based on the test set to obtain prediction results. The mean square error, root mean square error, and mean absolute error are calculated to evaluate the performance of the prediction results. When the performance of the regression prediction model meets the prediction performance evaluation indicators, the optimized regression prediction model is output.

[0098] S54: Based on the optimized regression prediction model and dynamic scheduling strategy, determine the collaborative division of labor rules between edge computing nodes and cloud models, and output the collaborative division of labor scheme;

[0099] This step addresses the issue of limited processing power of a single computing node, thereby improving the overall performance of the system.

[0100] S55: According to the collaborative division of labor scheme, the edge computing nodes are integrated with the optimized regression prediction model to construct an edge-cloud collaborative computing mechanism;

[0101] In this step, the optimized regression prediction model is deployed to edge computing nodes, and a communication mechanism is established between the edge computing nodes and the cloud model according to the collaborative division of labor scheme. This solves the problem of difficult collaboration between edge computing and cloud computing.

[0102] S56: Based on the edge-cloud collaborative computing mechanism, the edge computing nodes and the cloud model interact in real time to obtain an optimized cloud model.

[0103] In this step, real-time communication between edge computing nodes and cloud models is achieved to ensure timely data transmission. Based on feedback data from edge computing nodes, the parameters of the cloud model are adjusted to optimize model performance.

[0104] S6: Real-time detection of photovoltaic equipment is performed based on the optimized cloud model. When the detection result does not meet the preset performance threshold, a real-time alarm is triggered. Example 2:

[0105] This embodiment provides a detection device for photovoltaic equipment operating under self-derating load, the device comprising:

[0106] The acquisition module is used to acquire photovoltaic multidimensional data, which includes meteorological data, electrical data, environmental data, and equipment health data.

[0107] The calibration module is used to dynamically calibrate the physical model of a single diode based on photovoltaic multidimensional data, and to perform transfer learning training on the deep learning model in combination with photovoltaic physical characteristics to obtain a dynamic reference power.

[0108] To clarify the specific methods for obtaining the calibration module, the following are provided:

[0109] The calibration unit is used to dynamically calibrate the physical model of a single diode based on real-time temperature and light intensity in photovoltaic multidimensional data, so as to obtain an optimized physical model of a single diode.

[0110] The training unit is used to perform transfer learning training on the deep learning model based on the historical power data and weather forecast data in the photovoltaic physical characteristics, to build a power prediction model, and to predict future power changes through the power prediction model to obtain the predicted power value.

[0111] The fusion unit is used to perform weighted fusion based on the optimized single-diode physical model and the predicted power value to generate a dynamic reference power.

[0112] The optimization module is used to optimize the random forest model through ensemble learning based on the photovoltaic multidimensional data. The optimized random forest model is used to classify the load reduction characteristics of photovoltaic equipment to obtain the load reduction type results.

[0113] To clarify the specific methods for obtaining the optimization module, the following are included:

[0114] A partitioning unit is used to partition the photovoltaic multidimensional data to obtain a training subset, wherein the training subset includes a training set and a test set;

[0115] An optimization unit is used to perform ensemble learning optimization of the hyperparameters of the random forest model based on grid search and the training subset, and to determine the optimal hyperparameter configuration through cross-validation.

[0116] The configuration unit is used to initialize the random forest model based on the optimal hyperparameter configuration. During the training process, it performs double random sampling of features and samples through ensemble learning methods to generate multiple decision trees.

[0117] To clarify the specific methods for obtaining configuration units, the following are included:

[0118] An optimization subunit is used to optimize the hyperparameters of the random forest model based on the grid search to obtain the hyperparameter configuration;

[0119] A fitting subunit is used to fit the random forest model based on the hyperparameter configuration and the training set to obtain an optimized random forest model;

[0120] The first evaluation subunit is used to evaluate the optimized random forest model based on the test set and obtain the model performance index.

[0121] The second subunit is used to evaluate the performance of different hyperparameters based on the model performance metrics and cross-validation, and to determine the optimal hyperparameter configuration.

[0122] The voting unit is used to vote on the load reduction features of multiple decision trees based on the ensemble learning mechanism, and output the load reduction type classification result.

[0123] The calculation module is used to calculate the initial loss value based on the dynamic reference power and the real-time power in the photovoltaic multidimensional data, and to construct the power loss value by performing multi-factor correction on the initial loss value;

[0124] The interaction module is used to obtain an optimized cloud model by combining the power loss value and load reduction type results with the real-time interaction between the edge computing node and the cloud model.

[0125] The detection module is used to perform real-time detection on photovoltaic equipment based on the optimized cloud model. When the detection result does not meet the preset performance threshold, a real-time alarm is triggered.

[0126] It should be noted that the specific manner in which each module performs its operation in the apparatus described in the above embodiments has been described in detail in the embodiments of the method, and will not be elaborated here. Example 3:

[0127] Corresponding to the above method embodiments, this embodiment also provides a detection device for photovoltaic equipment self-derating operation. The detection device for photovoltaic equipment self-derating operation described below and the detection method for photovoltaic equipment self-derating operation described above can be referred to in correspondence.

[0128] Figure 2 This is a block diagram illustrating a detection device 800 for self-derating operation of a photovoltaic device according to an exemplary embodiment. Figure 2 As shown, the photovoltaic device self-derating load operation detection device 800 may include: a processor 801 and a memory 802. The photovoltaic device self-derating load operation detection device 800 may also include one or more of the following: a multimedia component 803, an I / O interface 804, and a communication component 805.

[0129] The processor 801 controls the overall operation of the photovoltaic device self-derating operation detection device 800 to complete all or part of the steps in the aforementioned photovoltaic device self-derating operation detection method. The memory 802 stores various types of data to support the operation of the photovoltaic device self-derating operation detection device 800. This data may include, for example, instructions for any application or method operating on the photovoltaic device self-derating operation detection device 800, and application-related data such as contact data, sent and received messages, images, audio, video, etc. The memory 802 can be implemented using any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 802 or transmitted via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. I / O interface 804 provides an interface between processor 801 and other interface modules, such as keyboards, mice, and buttons. These buttons can be virtual or physical. Communication component 805 is used for wired or wireless communication between the photovoltaic device's self-derating load detection device 800 and other devices. Wireless communication includes Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 805 may include a Wi-Fi module, a Bluetooth module, or an NFC module.

[0130] In an exemplary embodiment, the photovoltaic device self-derating operation detection device 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described photovoltaic device self-derating operation detection method. Example 4:

[0131] Corresponding to the above method embodiments, this embodiment also provides a medium. The medium described below can be referred to in conjunction with the detection method for self-derating operation of photovoltaic equipment described above.

[0132] A medium storing a computer program, which, when executed by a processor, implements the steps of the detection method for self-derating operation of photovoltaic equipment as described in the above method embodiments.

[0133] The medium can specifically be any medium capable of storing program code, such as a USB flash drive, external hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0134] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0135] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for detecting self-derating operation of photovoltaic equipment, characterized in that, include: Acquire photovoltaic multidimensional data, which includes meteorological data, electrical data, environmental data, and equipment health data; Based on the dynamic calibration of the single diode physical model using photovoltaic multidimensional data, and combined with the photovoltaic physical characteristics, the deep learning model is trained by transfer learning to obtain the dynamic reference power. Based on the aforementioned photovoltaic multidimensional data, the random forest model is ensemble-learned and optimized. The optimized random forest model is used to classify the load reduction characteristics of photovoltaic equipment, and the load reduction type results are obtained. The specific methods for obtaining the load reduction type results include: Based on the photovoltaic multidimensional data, a training subset is obtained, which includes a training set and a test set; The hyperparameters of the random forest model are optimized using grid search to obtain the hyperparameter configuration; The random forest model is fitted based on the hyperparameter configuration and the training set to obtain an optimized random forest model. The optimized random forest model is evaluated based on the test set to obtain model performance metrics; The performance of different hyperparameters is evaluated based on the model performance metrics and cross-validation to determine the optimal hyperparameter configuration; The random forest model is initialized based on the optimal hyperparameter configuration. During the training process, the features and samples are randomly sampled twice through an ensemble learning method to generate multiple decision trees. Based on the ensemble learning mechanism, the prediction results of multiple decision trees are voted on for load reduction features, and the load reduction type classification result is output. The initial loss value is calculated based on the dynamic reference power and the real-time power in the photovoltaic multidimensional data. The power loss value is then constructed by correcting the initial loss value for multiple factors. The specific methods for obtaining the power loss value include: The initial power loss value is obtained by time integration based on the difference between the dynamic reference power and the real-time power. The power loss value over time was analyzed using a long short-term memory network model to obtain the power change rate; The initial power loss value is dynamically adjusted based on the power change rate to obtain a corrected power loss value. A dynamic electricity price model is obtained by constructing direct economic losses based on the electricity price weights and subsidy coefficients for different time periods. A lifetime loss model is constructed based on the lifetime loss acceleration factor and the corrected power loss value. The power loss value is finally obtained by weighted summation based on the dynamic electricity price model and the lifetime loss model. Based on the power loss value and load reduction type results, and combined with the real-time interaction between the edge computing node and the cloud model, an optimized cloud model is obtained; The photovoltaic equipment is monitored in real time based on the optimized cloud model. When the monitoring results do not meet the preset performance threshold, a real-time alarm is triggered.

2. The detection method for self-derating operation of photovoltaic equipment according to claim 1, characterized in that, Based on dynamic calibration of a single diode physical model using photovoltaic multidimensional data, and combined with photovoltaic physical characteristics, a deep learning model is trained through transfer learning to obtain a dynamic reference power, including: The physical model of a single diode is dynamically calibrated based on real-time temperature and light intensity data from photovoltaic multidimensional data to obtain an optimized physical model of the single diode. Based on the historical power data and weather forecast data in the photovoltaic physical characteristics, the deep learning model is trained by transfer learning to construct a power prediction model. The power prediction model is then used to predict future power changes and obtain the predicted power value. A dynamic reference power is generated by weighted fusion of the optimized single-diode physical model and the predicted power value.

3. A detection device for self-reducing load operation of photovoltaic equipment, characterized in that, include: The acquisition module is used to acquire photovoltaic multidimensional data, which includes meteorological data, electrical data, environmental data, and equipment health data. The calibration module is used to dynamically calibrate the physical model of a single diode based on photovoltaic multidimensional data, and to perform transfer learning training on the deep learning model in combination with photovoltaic physical characteristics to obtain a dynamic reference power. The optimization module is used to optimize the random forest model through ensemble learning based on the photovoltaic multidimensional data. The optimized random forest model is used to classify the load reduction characteristics of photovoltaic equipment to obtain the load reduction type results. The optimization module includes: A partitioning unit is used to partition the photovoltaic multidimensional data to obtain a training subset, wherein the training subset includes a training set and a test set; The optimization subunit is used to optimize the hyperparameters of the random forest model based on grid search to obtain the hyperparameter configuration; A fitting subunit is used to fit the random forest model based on the hyperparameter configuration and the training set to obtain an optimized random forest model; The first evaluation subunit is used to evaluate the optimized random forest model based on the test set and obtain the model performance index. The second subunit is used to evaluate the performance of different hyperparameters based on the model performance metrics and cross-validation, and to determine the optimal hyperparameter configuration. The configuration unit is used to initialize the random forest model based on the optimal hyperparameter configuration. During the training process, it performs double random sampling of features and samples through ensemble learning methods to generate multiple decision trees. The voting unit is used to vote on the deload reduction features of the prediction results of multiple decision trees according to the ensemble learning mechanism, and output the deload reduction type classification result; The calculation module is used to calculate the initial loss value based on the dynamic reference power and the real-time power in the photovoltaic multidimensional data, and to construct the power loss value by performing multi-factor correction on the initial loss value; The calculation module includes: The initial power loss value is obtained by time integration based on the difference between the dynamic reference power and the real-time power. The power loss value over time was analyzed using a long short-term memory network model to obtain the power change rate; The initial power loss value is dynamically adjusted based on the power change rate to obtain a corrected power loss value. A dynamic electricity price model is obtained by constructing direct economic losses based on the electricity price weights and subsidy coefficients for different time periods. A lifetime loss model is constructed based on the lifetime loss acceleration factor and the corrected power loss value. The power loss value is finally obtained by weighted summation based on the dynamic electricity price model and the lifetime loss model. The interaction module is used to obtain an optimized cloud model by combining the power loss value and load reduction type results with the real-time interaction between the edge computing node and the cloud model. The detection module is used to perform real-time detection on photovoltaic equipment based on the optimized cloud model. When the detection result does not meet the preset performance threshold, a real-time alarm is triggered.

4. The detection device for self-reducing load operation of photovoltaic equipment according to claim 3, characterized in that, The calibration module includes: The calibration unit is used to dynamically calibrate the physical model of a single diode based on real-time temperature and light intensity in photovoltaic multidimensional data, so as to obtain an optimized physical model of a single diode. The training unit is used to perform transfer learning training on the deep learning model based on the historical power data and weather forecast data in the photovoltaic physical characteristics, to build a power prediction model, and to predict future power changes through the power prediction model to obtain the predicted power value. The fusion unit is used to perform weighted fusion based on the optimized single-diode physical model and the predicted power value to generate a dynamic reference power.

Images