A distributed photovoltaic power generation real-time monitoring system and method

Abstract

The present invention discloses a distributed photovoltaic power generation real-time monitoring system and method, which relates to the field of distributed photovoltaic power generation system monitoring, including a data acquisition module for collecting and preprocessing the operating data of the distributed photovoltaic power generation system; a data enhancement module for receiving the operating data processed by the data acquisition module; a collaborative learning module for integrating the optimal time series model of each distributed photovoltaic power generation system and constructing a collaborative detection model; a compression deployment module for compressing the collaborative detection model and deploying it to an edge computing device; and a monitoring optimization module for performing real-time monitoring of the distributed photovoltaic power generation system. The present invention monitors the operating status through a real-time monitoring model, quickly detects and responds to faults, and promptly issues an alarm signal to achieve accurate identification and classification of faults; through model compression and edge deployment, the computing resource requirements of the model are reduced, so that it can run efficiently on resource-constrained devices.

Description

Technical Field

[0001] The present invention relates to the field of distributed photovoltaic power generation system monitoring, and in particular to a distributed photovoltaic power generation real-time monitoring system and method. Background Art

[0002] Distributed photovoltaic power generation systems are usually composed of multiple photovoltaic modules, inverters, monitoring equipment, etc., distributed in multiple physical locations to form a wide-area power generation network. The operation of such systems is usually affected by multiple factors such as environmental conditions, component aging, system configuration, etc., and are prone to performance degradation, equipment failure and other problems.

[0003] At present, most of the photovoltaic power generation monitoring systems available on the market adopt a centralized architecture, relying on a central server to collect and analyze all data. However, this centralized architecture has some significant disadvantages. First, since all data needs to be transmitted to the central server for processing, the delay and bandwidth limitation of data transmission become the main bottlenecks when facing large-scale distributed systems, resulting in the system being unable to respond and handle emergency failures in a timely manner. Secondly, the centralized data processing method requires the operation data of all distributed nodes to be concentrated in a central location, which may cause data leakage and privacy protection issues, especially when facing sensitive data. Thirdly, when the centralized architecture gradually expands the system scale, the processing power and data storage capacity of the server become limiting factors, making it difficult to meet the real-time monitoring needs of large-scale distributed photovoltaic systems. Finally, traditional monitoring systems mainly rely on simple threshold judgments or basic statistical analysis methods, which are difficult to accurately identify and predict complex faults, especially when the system operating conditions are complex and changeable.

[0004] Therefore, in view of the shortcomings of the existing technology, there is an urgent need for a distributed photovoltaic power generation real-time monitoring system and method that can overcome the above-mentioned problems, so as to improve the system's operating efficiency, fault detection capability and scalability, and ensure the stable and efficient operation of the photovoltaic power generation system. Summary of the invention

[0005] In view of the problems existing in the existing distributed photovoltaic power generation monitoring system, the present invention is proposed.

[0006] Therefore, the problem to be solved by the present invention is how to provide a real-time monitoring system and method that effectively improves the operating efficiency, fault detection accuracy and system scalability of a distributed photovoltaic power generation system.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In the first aspect, an embodiment of the present invention provides a distributed photovoltaic power generation real-time monitoring system, which includes a data acquisition module for collecting and preprocessing operating data of distributed photovoltaic power generation systems to ensure the quality of data input; a data enhancement module for receiving the operating data processed by the data acquisition module, generating synthetic data through an adversarial network GAN method to expand the data set, and selecting and training an optimal time series model; a collaborative learning module for integrating the optimal time series models of each distributed photovoltaic power generation system, constructing a collaborative detection model, and aggregating and optimizing the parameters of the collaborative detection model under a collaborative learning framework to improve monitoring efficiency; a compression deployment module for compressing the collaborative detection model and deploying it to an edge computing device to achieve efficient monitoring of the collaborative detection model; a monitoring optimization module for performing real-time monitoring of distributed photovoltaic power generation systems, identifying and feedbacking equipment failures, and improving the performance of the collaborative detection model through adaptive optimization technology to adapt to data changes between different devices.

[0009] As a preferred solution of the distributed photovoltaic power generation real-time monitoring system described in the present invention, the operating data of the distributed photovoltaic power generation system includes current, voltage, component temperature, sunshine intensity, inverter output power, ambient temperature, humidity, wind speed and system efficiency; the process of generating synthetic data includes the following steps: constructing an adversarial network model through the operating data of the distributed photovoltaic power generation system; the adversarial network model includes a generator and a discriminator; using the generator of the adversarial network model to generate first synthetic data, and evaluating the similarity between the first synthetic data and the actual data through the discriminator; based on the evaluation result, generating and verifying second synthetic data through adversarial training of the generator and the discriminator; merging the verified second synthetic data with the original data set, expanding the data set, and preprocessing; based on the preprocessed extended data set, constructing and training the optimal time series model using the long short-term memory network LSTM algorithm.

[0010] As a preferred solution of the distributed photovoltaic power generation real-time monitoring system of the present invention, the evaluation formula of the optimal time series model is as follows:

[0011] ;

[0012] in, Score the selection of the optimal time series model, is the total number of data points, is the eigenvalue of the a-th data point, is the mean of the extended data set, is the standard deviation of the data set, and is the weight parameter, is the time characteristic value of the a-th data point, is the time scale factor.

[0013] When the optimal time series model selection score When the optimal time series model is in the range of (0.8,1], the matching degree between the optimal time series model and the photovoltaic power generation data is high, and the dynamic changes of photovoltaic power generation data are predicted and analyzed through the collaborative learning module; when the selection score of the optimal time series model is When the optimal time series model and the photovoltaic power generation data are in the range of (0.6, 0.8], the matching degree between the optimal time series model and the photovoltaic power generation data is moderate, and the parameters of the optimal time series model are optimized; when the selection score of the optimal time series model is When it is in the range of [0,0.6], the optimal time series model and the photovoltaic power generation data have a low match, and the optimal time series model cannot capture the key characteristics of the data, so the model needs to be rebuilt.

[0014] As a preferred solution of the distributed photovoltaic power generation real-time monitoring system described in the present invention, the construction process of the collaborative learning module includes the following steps: setting up a number of collaborative learning nodes for the distributed photovoltaic power generation system, and ensuring the security of data transmission through an encrypted communication protocol; preprocessing the photovoltaic power generation system data of the collaborative learning nodes; constructing and training a collaborative detection model using a convolutional neural network algorithm based on the characteristics of the preprocessed photovoltaic power generation system data; optimizing and adjusting the parameters of the collaborative detection model through the data of different collaborative learning nodes, and compressing it through a compression deployment module; implementing a continuous learning mechanism based on the compressed collaborative detection model, updating and optimizing the collaborative detection model based on newly collected data, and using encryption technology and access control measures to protect the privacy of the data.

[0015] As a preferred solution of the distributed photovoltaic power generation real-time monitoring system of the present invention, the specific formula of the collaborative detection model is as follows:

[0016] ;

[0017] in, is the output of the collaborative detection model, which represents the comprehensive evaluation result of the distributed photovoltaic power generation system operation data. and To adjust the parameters, ReLU is the corrected linear unit activation function, which is used to process the input data of the total number of feature indices of the i-th feature. , weight and bias , is the total number of feature indexes of the input data, is the total number of time-related feature indices, is the feature index of the input data, is the index of the time-related feature, Sigmoid activation function, used to process the intermediate variable of the jth feature index and weight .

[0018] As a preferred solution of the distributed photovoltaic power generation real-time monitoring system described in the present invention, the compression deployment module includes the following steps: using an automated tool to evaluate each layer and parameter of the collaborative detection model to identify the part with the smallest performance contribution; selecting sparsification and quantization technology based on the evaluation results, and compressing the collaborative detection model by removing the neural network connections with small influence and compressing the parameter bit width; adjusting the compressed collaborative detection model to improve the efficiency on photovoltaic power generation data; deploying the compressed and adjusted collaborative detection model to the edge computing device of the photovoltaic power generation system for real-time monitoring and data analysis; evaluating the compression effect by the following formula for the performance change of the collaborative detection model before and after compression:

[0019] ;

[0020] in, To evaluate the time period, is the performance function of the collaborative detection model over time, is the number of features to be evaluated, is the performance metric of the b-th feature before compression, is the performance metric of the b-th feature after compression.

[0021] When the compression effect is scored When the compression effect score is between 85% and 100%, the distributed photovoltaic power generation system is monitored in real time through the collaborative detection model; When the compression effect score is between 65% and 85%, the parameters of the collaborative detection model are readjusted and optimized through the optimal time series model; When it is between 0% and 65%, the collaborative detection model is rebuilt and compressed.

[0022] As a preferred solution of the distributed photovoltaic power generation real-time monitoring system described in the present invention, the monitoring optimization module includes: real-time monitoring of the operating data of the distributed photovoltaic power generation system through a collaborative detection model, and when a fault is detected, an alarm signal is immediately issued, fault data is collected, and the type and cause of the fault are marked; the collaborative detection model is trained based on the fault data to improve the recognition of the fault type, and the trained collaborative detection model is verified and optimized; the optimized collaborative detection model is redeployed to the photovoltaic power generation monitoring system to enhance the fault detection and prevention capabilities.

[0023] In the second aspect, an embodiment of the present invention provides a real-time monitoring method for distributed photovoltaic power generation, which includes: collecting operating data of a distributed photovoltaic power generation system through a data acquisition module, and preprocessing the operating data; a data enhancement module generates synthetic data by adopting an adversarial network GAN method, expanding the operating data set of the distributed photovoltaic power generation system, and selecting an optimal time series model for training; integrating the trained optimal time series model into a collaborative learning framework, and constructing a collaborative detection model using a collaborative learning module; compressing the collaborative detection model through a compression deployment module, applying sparsification and quantization techniques to reduce the complexity of the model, and deploying the compressed model to an edge computing device of a distributed photovoltaic power generation system; a monitoring optimization module monitors the operating data of a distributed photovoltaic power generation system in real time, and when an abnormality or failure of the equipment is detected, an alarm signal is immediately issued to collect and mark the fault data; training and optimizing the collaborative detection model through the fault data to adapt to data changes between different devices, and redeploying the optimized collaborative detection model to the system to enhance fault detection and prevention capabilities.

[0024] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of the distributed photovoltaic power generation real-time monitoring method as described in the first aspect of the present invention are implemented.

[0025] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of the distributed photovoltaic power generation real-time monitoring method as described in the first aspect of the present invention are implemented.

[0026] The beneficial effects of the present invention are as follows: the present invention continuously monitors the operating status of the photovoltaic power generation system through a real-time monitoring model, can quickly detect and respond to faults, issue alarm signals in a timely manner, and automatically collect and annotate fault data to achieve accurate identification and classification of faults; through data enhancement and collaborative learning technology, the model can effectively deal with data imbalance problems and maintain efficient prediction capabilities in a variety of operating environments; through model compression and edge deployment, the computing resource requirements of the model are reduced, so that it can run efficiently on resource-constrained devices, ultimately enhancing the system's fault detection and prevention capabilities and improving the overall stability and operating efficiency of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0028] Figure 1 This is the overall framework diagram of the distributed photovoltaic power generation real-time monitoring system of Example 1.

[0029] Figure 2 This is a flow chart of the distributed photovoltaic power generation real-time monitoring system of Example 2. DETAILED DESCRIPTION

[0030] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0031] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0032] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0033] Example 1

[0034] Reference Figure 1 , which is the first embodiment of the present invention, provides a distributed photovoltaic power generation real-time monitoring system, which is composed of a data acquisition module, a data enhancement module, a collaborative learning module, a compression deployment module and a monitoring optimization module.

[0035] Specifically, the data acquisition module is used to collect and pre-process the operating data of the distributed photovoltaic power generation system to ensure the quality of data input; the data enhancement module is used to receive the operating data processed by the data acquisition module, generate synthetic data through the adversarial network GAN method to expand the data set, select and train the optimal time series model; the collaborative learning module is used to integrate the optimal time series models of each distributed photovoltaic power generation system, construct a collaborative detection model, and aggregate and optimize the parameters of the collaborative detection model under the collaborative learning framework to improve monitoring efficiency.

[0036] It should be noted that the operating data of distributed photovoltaic power generation systems include current, voltage, component temperature, sunshine intensity, inverter output power, ambient temperature, humidity, wind speed and system efficiency. In the real-time monitoring system of distributed photovoltaic power generation, the preprocessing process first cleans the collected operating data, processes missing values ​​and outliers to ensure the integrity and accuracy of the data; removes noise through low-pass filtering; optimizes signal quality; normalizes and standardizes the data to eliminate differences between different dimensions and make the data distribution more uniform; extracts key features through feature selection, combines principal component analysis (PCA) and other technologies to reduce dimensionality and reduce data redundancy; for time series data, applies the lag feature extraction method to capture the time dependency and dynamic changes in the data, and provides high-quality input data for subsequent model training.

[0037] Furthermore, a compression deployment module is used to compress the collaborative detection model and deploy it to an edge computing device to achieve efficient monitoring of the collaborative detection model; a monitoring optimization module is used to perform real-time monitoring of distributed photovoltaic power generation systems, identify and feedback equipment failures, and improve the performance of the collaborative detection model through adaptive optimization technology to adapt to data changes between different devices.

[0038] Furthermore, in the data enhancement module, the process of generating synthetic data includes the following steps: constructing an adversarial network model through the operating data of the distributed photovoltaic power generation system; the adversarial network model includes a generator and a discriminator; using the generator of the adversarial network model to generate first synthetic data, and evaluating the similarity between the first synthetic data and the actual data through the discriminator; based on the evaluation result, generating and verifying the second synthetic data through adversarial training of the generator and the discriminator; merging the verified second synthetic data with the original data set, expanding the data set, and preprocessing; based on the preprocessed extended data set, using the long short-term memory network LSTM algorithm to construct and train the optimal time series model.

[0039] Specifically, the evaluation formula of the optimal time series model is as follows:

[0040] ;

[0041] in, Score the selection of the optimal time series model, is the total number of data points, is the eigenvalue of the a-th data point, is the mean of the extended data set, is the standard deviation of the data set, and is the weight parameter, is the time characteristic value of the a-th data point, is the time scale factor.

[0042] It should be noted that the feature values ​​and time feature values ​​are standardized and weighted summed according to their performance in the data set. Combining the characteristics and time dependence of data points enables a more accurate evaluation of the performance of different time series models on extended datasets. and , the influence of eigenvalues ​​and time eigenvalues ​​on model selection can be flexibly controlled, thereby ensuring that the time series model that best suits the current data set is selected.

[0043] Preferably, when the selection score of the optimal time series model When the optimal time series model is in the range of (0.8,1], the matching degree between the optimal time series model and the photovoltaic power generation data is high, and the dynamic changes of photovoltaic power generation data are predicted and analyzed through the collaborative learning module; when the selection score of the optimal time series model is When the optimal time series model and the photovoltaic power generation data are in the range of (0.6, 0.8], the matching degree between the optimal time series model and the photovoltaic power generation data is moderate, and the parameters of the optimal time series model are optimized; when the selection score of the optimal time series model is When it is in the range of [0,0.6], the optimal time series model and the photovoltaic power generation data have a low match, and the optimal time series model cannot capture the key characteristics of the data, so the model needs to be rebuilt.

[0044] Furthermore, in the collaborative learning module, the construction process includes the following steps: setting up a number of collaborative learning nodes for the distributed photovoltaic power generation system, and ensuring the security of data transmission through an encrypted communication protocol; preprocessing the photovoltaic power generation system data of the collaborative learning nodes; constructing and training a collaborative detection model using a convolutional neural network algorithm based on the characteristics of the preprocessed photovoltaic power generation system data; optimizing and adjusting the parameters of the collaborative detection model through the data of different collaborative learning nodes, and compressing it through a compression deployment module; implementing a continuous learning mechanism based on the compressed collaborative detection model, updating and optimizing the collaborative detection model based on newly collected data, and using encryption technology and access control measures to protect data privacy.

[0045] Furthermore, the specific formula of the collaborative detection model is as follows:

[0046] ;

[0047] in, is the output of the collaborative detection model, which represents the comprehensive evaluation result of the distributed photovoltaic power generation system operation data. and To adjust the parameters, ReLU is the corrected linear unit activation function, which is used to process the input data of the total number of feature indices of the i-th feature. , weight and bias , is the total number of feature indexes of the input data, is the total number of time-related feature indices, is the feature index of the input data, is the index of the time-related feature, Sigmoid activation function, used to process the intermediate variable of the jth feature index and weight .

[0048] It should be noted that the formula of the collaborative detection model generates a comprehensive evaluation result by combining different features and time-related features, which is used to comprehensively evaluate the operating status of the distributed photovoltaic power generation system. Specifically, the formula consists of two main parts: one part processes the main operating characteristics of the system, and the other part processes the time-related features. The first part uses the modified linear unit (ReLU) activation function to perform nonlinear processing on the linear combination of input feature values, so that the model can capture complex feature interactions. This nonlinear processing can effectively deal with common nonlinear phenomena in photovoltaic power generation systems, such as the complex relationship between voltage and current. The second part processes time-related features through the Sigmoid activation function, mapping these features to the range of [0,1], so as to better capture trends and periodic changes in time series. By adding adjustment parameters, the formula can flexibly adjust the weights of these two parts in the final output, so that the model can provide accurate evaluation in different application scenarios. The final output, as a comprehensive evaluation result of the operating status of the photovoltaic power generation system, can be used in multiple aspects such as fault detection, performance monitoring, and system optimization, thereby improving the reliability and operation efficiency of the system.

[0049] Specifically, the compression deployment module includes the following steps: using automated tools to evaluate the layers and parameters of the collaborative detection model and identify the part with the smallest performance contribution; selecting sparsification and quantization techniques based on the evaluation results, and compressing the collaborative detection model by removing neural network connections with small effects and compressing parameter bit widths; adjusting the compressed collaborative detection model to improve the efficiency on photovoltaic power generation data; deploying the compressed and adjusted collaborative detection model to the edge computing device of the photovoltaic power generation system for real-time monitoring and data analysis; evaluating the compression effect of the collaborative detection model based on the performance changes before and after compression using the following formula:

[0050] ;

[0051] in, To evaluate the time period, is the performance function of the collaborative detection model over time, is the number of features to be evaluated, is the performance metric of the b-th feature before compression, is the performance metric of the b-th feature after compression.

[0052] It should be noted that the performance change of the collaborative detection model before and after compression is evaluated, and the compression effect is measured by quantifying the impact of compression on model performance.

[0053] Furthermore, when the compression effect score When the compression effect score is between 85% and 100%, the distributed photovoltaic power generation system is monitored in real time through the collaborative detection model; When the compression effect score is between 65% and 85%, the parameters of the collaborative detection model are readjusted and optimized through the optimal time series model; When it is between 0% and 65%, the collaborative detection model is rebuilt and compressed.

[0054] Furthermore, the monitoring optimization module includes real-time monitoring of the operating data of the distributed photovoltaic power generation system through a collaborative detection model. When a fault is detected, an alarm signal is immediately issued, fault data is collected, and the type and cause of the fault are marked; the collaborative detection model is trained based on the fault data to improve the recognition of the fault type, and the trained collaborative detection model is verified and optimized; the optimized collaborative detection model is redeployed to the photovoltaic power generation monitoring system to enhance fault detection and prevention capabilities.

[0055] Example 2

[0056] Reference Figure 2On the basis of the first embodiment, this embodiment further provides a distributed photovoltaic power generation real-time monitoring method, including the following steps: S1: collecting the operating data of the distributed photovoltaic power generation system through the data acquisition module, and preprocessing the operating data; S2: the data enhancement module generates synthetic data by adopting the adversarial network GAN method, expands the operating data set of the distributed photovoltaic power generation system, and selects the optimal time series model for training; S3: integrating the trained optimal time series model into the collaborative learning framework, and constructing the collaborative detection model using the collaborative learning module; S4: compressing the collaborative detection model through the compression deployment module, applying sparseness and quantization technology to reduce the complexity of the model, and deploying the compressed model to the edge computing device of the distributed photovoltaic power generation system; S5: the monitoring optimization module monitors the operating data of the distributed photovoltaic power generation system in real time, and when an abnormality or failure of the equipment is detected, an alarm signal is immediately issued, and the fault data is collected and marked; S6: training and optimizing the collaborative detection model through the fault data, adapting to the data changes between different devices, and redeploying the optimized collaborative detection model to the system to enhance the fault detection and prevention capabilities.

[0057] This embodiment also provides a computer device, which is suitable for the distributed photovoltaic power generation real-time monitoring system, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the distributed photovoltaic power generation real-time monitoring system proposed in the above embodiment.

[0058] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0059] In summary, the present invention continuously monitors the operating status of the photovoltaic power generation system through a real-time monitoring model, can quickly detect and respond to faults, issue alarm signals in a timely manner, and automatically collect and annotate fault data to achieve accurate identification and classification of faults; through data enhancement and collaborative learning technology, the model can effectively deal with data imbalance problems and maintain efficient prediction capabilities in a variety of operating environments; through model compression and edge deployment, the computing resource requirements of the model are reduced, enabling it to run efficiently on resource-constrained devices, ultimately enhancing the system's fault detection and prevention capabilities and improving the overall stability and operating efficiency of the system.

[0060] Example 3

[0061] Referring to Table 1, which is the third embodiment of the present invention, this embodiment provides a distributed photovoltaic power generation real-time monitoring system. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0062] Specifically, two experimental systems are prepared, one is a photovoltaic power generation monitoring system based on the prior art (system A), and the other is a distributed photovoltaic power generation real-time monitoring system (system B) using the present invention. System A uses traditional data acquisition and a simple time series model for monitoring, while system B integrates a data acquisition module, a data enhancement module, a collaborative learning module, a compression deployment module, and a monitoring optimization module.

[0063] Furthermore, in the experiment, the same operating data, including current, voltage, component temperature, sunshine intensity, inverter output power, ambient temperature, humidity, wind speed and system efficiency, were collected from the photovoltaic power generation devices of the two systems through the data acquisition module. The collected data was preprocessed, such as missing value filling, noise filtering and standardization. Subsequently, in system B, the data enhancement module used the generative adversarial network (GAN) technology to generate additional synthetic data to expand the data set, and the long short-term memory network (LSTM) algorithm was used to build and train the optimal time series model.

[0064] Furthermore, system B uses the collaborative learning module to integrate the optimal time series models of different photovoltaic power generation devices, build a collaborative detection model, and optimize and aggregate model parameters through data sharing between different nodes. The optimized collaborative detection model is compressed through the compression deployment module, using sparse and quantization technology to reduce the model size, and deployed on the edge computing device to ensure the efficiency of real-time monitoring.

[0065] Specifically, as shown in Table 1, it can be seen from the test data that the detection accuracy of system B in various parameters is significantly higher than that of system A. For example, in terms of current detection accuracy, the detection accuracy of system B reached 95%, while that of system A was only 87%, indicating that through data enhancement and collaborative learning technology, system B can more accurately capture and judge the operating status of the photovoltaic power generation system. Similarly, in the detection of other parameters such as voltage, component temperature and sunshine intensity, the detection accuracy of system B is generally about 8-10 percentage points higher than that of system A.

[0066] Table 1 Test data table

[0067]

[0068] Furthermore, the response time of system B is significantly better than that of system A. The average response time of system A is about 15-19 seconds, while system B shortens the response time to 7-10 seconds. This result is due to the compression deployment module in system B. Through sparsification and quantization techniques, the computational complexity is significantly reduced, enabling the model to run efficiently on resource-constrained edge computing devices, thereby speeding up fault detection and early warning. In terms of computing resource utilization, system B also performs better than system A. Although system B integrates more functional modules and complex collaborative detection models, through compression technology, the computing resource utilization of system B is about 5% lower than that of system A on average. This shows that system B can effectively control the consumption of computing resources while ensuring high detection accuracy, reflecting its advantages in efficiency and economy.

[0069] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A distributed photovoltaic power generation real-time monitoring system, characterized in that: include, Data acquisition module, used to collect and pre-process the operating data of distributed photovoltaic power generation systems to ensure the quality of data input; A data enhancement module, used to receive the operating data processed by the data acquisition module, generate synthetic data through the adversarial network GAN method to expand the data set, and select and train the optimal time series model; A collaborative learning module is used to integrate the optimal time series models of each distributed photovoltaic power generation system, build a collaborative detection model, and aggregate and optimize the parameters of the collaborative detection model under the collaborative learning framework to improve monitoring efficiency; A compression deployment module, used to compress the collaborative detection model and deploy it to an edge computing device to achieve efficient monitoring of the collaborative detection model; A monitoring and optimization module is used to perform real-time monitoring of distributed photovoltaic power generation systems, identify and feedback equipment failures, and improve the performance of the collaborative detection model through adaptive optimization technology to adapt to data changes between different devices; The operating data of the distributed photovoltaic power generation system includes current, voltage, component temperature, sunshine intensity, inverter output power, ambient temperature, humidity, wind speed and system efficiency; the process of generating synthetic data includes the following steps: Constructing an adversarial network model through the operation data of the distributed photovoltaic power generation system; the adversarial network model includes a generator and a discriminator; Generate first synthetic data using a generator of the adversarial network model, and evaluate the similarity between the first synthetic data and actual data using a discriminator; Based on the evaluation results, generate and verify the second synthetic data through adversarial training of the generator and the discriminator; Merging the verified second synthetic data with the original data set, expanding the data set, and performing preprocessing; Based on the preprocessed extended data set, the optimal time series model is constructed and trained using the long short-term memory network LSTM algorithm; The evaluation formula of the optimal time series model is as follows: ； in, Score the selection of the optimal time series model, is the total number of data points, is the eigenvalue of the a-th data point, is the mean of the extended data set, is the standard deviation of the data set, and is the weight parameter, is the time characteristic value of the a-th data point, is the time scale factor; When the optimal time series model selection score When it is in the range of (0.8,1], the optimal time series model and the photovoltaic power generation data have a high degree of match, and the dynamic changes of photovoltaic power generation data are predicted and analyzed through the collaborative learning module; When the optimal time series model selection score When it is in the range of (0.6, 0.8], the matching degree between the optimal time series model and the photovoltaic power generation data is moderate, and the parameters of the optimal time series model are optimized; When the optimal time series model selection score When it is in the range of [0,0.6], the optimal time series model and the photovoltaic power generation data have a low match, and the optimal time series model cannot capture the key characteristics of the data, so the model should be rebuilt; The construction process of the collaborative learning module includes the following steps: Set up several collaborative learning nodes for the distributed photovoltaic power generation system and ensure the security of data transmission through encrypted communication protocols; Preprocess the photovoltaic power generation system data of the collaborative learning node; According to the characteristics of the preprocessed photovoltaic power generation system data, a collaborative detection model is constructed and trained using a convolutional neural network algorithm; Optimize and adjust the parameters of the collaborative detection model through the data of different collaborative learning nodes, and compress it through the compression deployment module; According to the compressed collaborative detection model, a continuous learning mechanism is implemented to update and optimize the collaborative detection model based on the newly collected data, and encryption technology and access control measures are used to protect the privacy of the data; The specific formula of the collaborative detection model is as follows: ； in, is the output of the collaborative detection model, which represents the comprehensive evaluation result of the distributed photovoltaic power generation system operation data. and To adjust the parameters, ReLU is the corrected linear unit activation function, which is used to process the input data of the total number of feature indices of the i-th feature. , weight and bias , is the total number of feature indexes of the input data, is the total number of time-related feature indices, is the feature index of the input data, is the index of the time-related feature, Sigmoid activation function, used to process the intermediate variable of the jth feature index and weight ; The compression deployment module includes the following steps: Use automated tools to evaluate the layers and parameters of the co-detection model to identify the parts that contribute the least to performance; Based on the evaluation results, sparsification and quantization techniques are selected to compress the collaborative detection model by removing neural network connections with small influence and compressing parameter bit width; Adjusting the compressed collaborative detection model to improve efficiency on photovoltaic power generation data; Deploy the compressed and adjusted collaborative detection model to the edge computing device of the photovoltaic power generation system for real-time monitoring and data analysis; The performance change of the collaborative detection model before and after compression is evaluated by the following formula: ； in, To evaluate the time period, is the performance function of the collaborative detection model over time, is the number of features to be evaluated, is the performance metric of the b-th feature before compression, is the performance metric of the b-th feature after compression; When the compression effect is scored When it is between 85% and 100%, the distributed photovoltaic power generation system is monitored in real time through the collaborative detection model; When the compression effect is scored When it is between 65% and 85%, the parameters of the collaborative detection model are readjusted and optimized through the optimal time series model; When the compression effect is scored When it is between 0% and 65%, the collaborative detection model is rebuilt and compressed; The monitoring optimization module includes: The collaborative detection model monitors the operating data of the distributed photovoltaic power generation system in real time. When a fault is detected, an alarm signal is immediately issued, fault data is collected, and the type and cause of the fault are marked. Training a collaborative detection model based on the fault data to improve recognition of fault types, and verifying and optimizing the trained collaborative detection model; The optimized collaborative detection model is redeployed to the photovoltaic power generation monitoring system to enhance fault detection and prevention capabilities.

2. A distributed photovoltaic power generation real-time monitoring method, based on the distributed photovoltaic power generation real-time monitoring system according to claim 1, characterized in that: include, Collecting the operating data of the distributed photovoltaic power generation system through the data acquisition module and preprocessing the operating data; The data enhancement module generates synthetic data by adopting the adversarial network GAN method, expands the operating data set of the distributed photovoltaic power generation system, and selects the optimal time series model for training; Integrate the trained optimal time series model into the collaborative learning framework and use the collaborative learning module to build a collaborative detection model; Compressing the collaborative detection model through a compression deployment module, applying sparsification and quantization techniques to reduce the complexity of the model, and deploying the compressed model on an edge computing device of a distributed photovoltaic power generation system; The monitoring and optimization module monitors the operating data of the distributed photovoltaic power generation system in real time. When an abnormality or failure of the equipment is detected, an alarm signal is immediately issued, and the fault data is collected and marked; The fault data is used to train and optimize the collaborative detection model, adapt to data changes between different devices, and redeploy the optimized collaborative detection model to the system to enhance fault detection and prevention capabilities.

3. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the distributed photovoltaic power generation real-time monitoring method according to claim 2 are implemented.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps of the distributed photovoltaic power generation real-time monitoring method according to claim 2 are implemented.

Images