Distributed photovoltaic measurement and control system based on edge calculation

Abstract

The invention relates to the technical field of edge computing, in particular to a distributed photovoltaic measurement and control system based on edge computing. The method comprises the following steps: an edge computing node unit is deployed at a position close to a data source and is used for collecting real-time data from a plurality of distributed photovoltaic nodes, processing the collected real-time data and supporting localized fault diagnosis through an intelligent algorithm; and the resource optimization unit is used for calculating the data processed by the node unit at the edge and optimizing electric energy quality management and reactive compensation measures at a grid-connected point in combination with historical data. According to the system, almost immediate understanding and response to the operation condition of the photovoltaic system are realized, the rapid response capability is helpful for detecting abnormal conditions or the performance reduction trend more timely, then preventive measures are taken to avoid potential problems, the flexibility and adaptability of the reactive compensation control system are improved, and the system is suitable for popularization and application. Stable operation of a power system is ensured, and electric energy quality is improved.

Description

Technical Field

[0001] The present invention relates to the field of edge computing technology, and in particular to a distributed photovoltaic measurement and control system based on edge computing. Background Art

[0002] With the growth of global demand for clean energy and the development of smart grid technology, distributed photovoltaic power generation systems have gradually become an important part of energy supply. However, the large-scale access of distributed photovoltaic power generation systems has brought new challenges to the operation of power systems, such as voltage fluctuations, frequency offsets, and power quality issues. Traditional centralized monitoring and control systems are unable to cope with the complexity and real-time requirements introduced by a large number of distributed power sources. In order to solve these problems and make full use of the advantages of distributed photovoltaic power generation systems, edge computing, as an emerging information processing paradigm, has been introduced into the measurement and control system of distributed photovoltaic power generation systems.

[0003] Edge computing can achieve localized data processing, reduce data transmission delays, improve response speed, and reduce the data load on cloud servers by deploying computing resources at the edge of the network. For distributed photovoltaic power generation systems, this means that the status information of power generation equipment can be obtained and analyzed more quickly, thereby achieving more accurate control and management. In addition, edge computing can also enhance the reliability and security of the system because it reduces dependence on external network connections and can perform critical tasks locally, maintaining basic functions even in the event of network failures.

[0004] However, with the continuous expansion of the scale of photovoltaic systems and the increasing technical requirements, traditional data processing methods can no longer meet the dual needs of immediacy and accuracy. For example, traditional compensation methods are often based on pre-set working points or mode operations, which makes it difficult to achieve the optimal compensation effect under different working conditions. This not only limits the adjustment range of the system, but also reduces the ability to handle emergencies, such as sudden increases in load or short circuit faults. Therefore, a distributed photovoltaic measurement and control system based on edge computing is designed. Summary of the invention

[0005] The purpose of the present invention is to provide a distributed photovoltaic measurement and control system based on edge computing to solve the problem that the traditional compensation method proposed in the above background technology is often based on a pre-set working point or mode operation, and it is difficult to achieve the optimal compensation effect under different working conditions.

[0006] To achieve the above object, the present invention aims to provide a distributed photovoltaic measurement and control system based on edge computing, comprising: An edge computing node unit, which is deployed near a data source and is used to collect real-time data from multiple distributed photovoltaic nodes, process the collected real-time data, and support localized fault diagnosis through intelligent algorithms; wherein the edge computing node unit includes a data aggregation and compression module and a fault diagnosis module; A resource optimization unit, which is used for the data processed by the edge computing node unit and optimizes the power quality management and reactive power compensation measures at the grid connection point in combination with historical data; wherein the resource optimization unit includes a power quality management module, a reactive power compensation control module and an energy storage system integration module.

[0007] As a further improvement of the present technical solution, in the edge computing node unit, the real-time data collected includes at least electrical parameters of voltage, current, power and frequency, at least environmental parameters of light intensity, temperature and humidity, at least meteorological parameters of wind speed, precipitation and cloud coverage, and also includes fault codes, operating time and health indicators of the equipment.

[0008] As a further improvement of the technical solution, in the edge computing node unit: The data aggregation and compression module is used to aggregate and compress the real-time data collected from multiple distributed photovoltaic nodes, and uses a dimensionality reduction algorithm to map high-dimensional data to a lower-dimensional space while retaining the information of the original data; a hybrid compression strategy is used to reassemble the data mapped to the lower-dimensional space into a complete file format, and a method based on a combination of random projection and online learning is introduced to optimize the process of mapping high-dimensional data to a lower-dimensional space using a dimensionality reduction algorithm; The fault diagnosis module is used to continuously monitor the collected data, identify deviations from the normal operating range, and predict future fault points. When a potential fault is found, the affected component is restarted and a warning notification is sent to the staff. For uncertain faults, the fault information is sent to the core node for analysis.

[0009] As a further improvement of the technical solution, in the data aggregation and compression module, a dimensionality reduction algorithm is used to map high-dimensional data to a lower-dimensional space, as follows: S1. First, the input data matrix Centralized processing: ; in, is the matrix after centralization; is the input data matrix; is the sample size; For the samples; ; S2. When the first batch of data is received, the initial covariance matrix is ​​constructed: ; in, is the covariance matrix of the first batch of data; is the number of samples of the first batch of data; This is the matrix after the first batch of data has been centralized; is the transpose operation; Whenever a new batch of data arrives, use the existing covariance matrix And the newly added data To update the current covariance matrix; ; in, is the current covariance matrix; For the forgetting factor; is the covariance matrix of the previous moment; For the newly added data; S3. For each updated covariance matrix , use Lanczos iteration method to perform eigenvalue decomposition and obtain the previous The largest eigenvalue; S4. According to The eigenvectors form the projection matrix , and then the centralized data Projected into a new coordinate system composed of these eigenvectors to form a reduced-dimensional representation : ; in, For dimensionality reduction data; for The projection matrix consists of eigenvectors.

[0010] As a further improvement of the technical solution, in the data aggregation compression module, a method based on random projection and online learning is introduced to optimize the process of mapping high-dimensional data to a lower-dimensional space using a dimensionality reduction algorithm. After optimization, the method is as follows: S5. After centralization, construct a random matrix: ; in, is a random matrix; is the original data dimension, is the dimension of the target after dimensionality reduction; The centralized data Projection to low-dimensional space middle: ; in, is a low-dimensional space; S6. When the first batch of data is received, the initial covariance matrix: ; in, is the covariance matrix of the first batch of data after optimization; The number of samples optimized for the first batch of data; The matrix after optimization and centralization for the first batch of data; Whenever a new batch of data arrives, use the existing covariance matrix And the newly added data Update the current covariance matrix; ; in, is the current optimized covariance matrix; is the learning rate; is the optimized covariance matrix of the previous moment; Representation of newly added data in low-dimensional space; S7. For each updated covariance matrix , using random power iteration to estimate the The largest eigenvalue and its corresponding eigenvector ; According to the cumulative variance contribution rate, select the appropriate number of principal components : ; in, is the optimal number of principal components; For the Large eigenvalues; For user-defined thresholds; according to The eigenvectors form the projection matrix , and then the centralized data Projected into a new coordinate system composed of these eigenvectors, a reduced-dimensional representation is formed: ; in, is the optimized dimensionality reduction data; for The projection matrix consists of eigenvectors.

[0011] As a further improvement of the technical solution, in the data aggregation compression module, a hybrid compression strategy is used to reassemble the data mapped to the lower dimensional space into a complete file format, as follows: S11. Reduce the dimension of the data According to the degree of influence on the quality of reconstructed images, they are classified into key parts and non-key parts; S12, compressing the portion marked as critical using a lossless compression algorithm, and compressing the portion marked as non-critical using a lossy compression technique; S13, merging the above two compression results into a single file format, and recording information about the compression method and the location index of each segment of data in the file header; S14. Adjust compression parameters according to actual conditions to achieve an optimal balance between compression ratio and quality, and verify whether the compressed data meets the expected quality standards and can be restored to an acceptable state after decompression.

[0012] As a further improvement of this technical solution, in the resource optimization unit, The power quality management module is used to evaluate the power quality status at the grid connection point and identify power quality problems. Once a problem is found, the corresponding early warning mechanism is triggered. The power quality management module also continuously tracks the development trend of the problem by configuring an online monitoring device; The reactive power compensation control module is used to automatically adjust the compensation amount according to actual needs, and supports the combination of remote control and local automated decision-making. In the process of automatically adjusting the compensation amount according to actual needs, power quality issues are introduced as input for optimization; The energy storage system integration module is used to dispatch energy storage according to the status of power quality and reactive power compensation.

[0013] As a further improvement of the present technical solution, the power quality problems include voltage deviation, harmonic distortion rate, three-phase imbalance and transient events.

[0014] As a further improvement of the technical solution, the reactive power compensation control module is specifically as follows: Dynamic reactive power compensation: ; in, For time Reactive power compensation amount on is the initial proportional gain coefficient; is the set target voltage; is the actual voltage at the current moment; is the initial integral gain coefficient; For The actual voltage of is the integration variable.

[0015] As a further improvement of the technical solution, in the reactive power compensation control module, in the process of automatically adjusting the compensation amount according to actual needs, the power quality problem is introduced as input for optimization, and the optimization is specifically as follows: Set comprehensive evaluation function , comprehensive evaluation function Different power quality issues are scored based on voltage deviation, harmonic distortion, three-phase unbalance and transient events, specifically: ; in, is the comprehensive evaluation function; For The evaluation score of voltage deviation at the moment; is the weight of the evaluation score of voltage deviation; For The evaluation score of the harmonic distortion rate at the moment; is the weight of the evaluation score of harmonic distortion rate; For The evaluation score of the three-phase imbalance at the moment; is the weight of the evaluation score of the three-phase imbalance; For The evaluation score of the transient event at the moment; is the weight of the evaluation score of transient events; According to the adjustment of proportional coefficient and integral coefficient: ; ; in, is the initial proportional gain coefficient; is the initial integral gain coefficient; is the sensitivity parameter of the proportional gain coefficient; is the sensitivity parameter of the integral gain coefficient; ; in, The optimized time The reactive power compensation amount.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. In this distributed photovoltaic measurement and control system based on edge computing, a method based on a combination of random projection and online learning is introduced for optimization in the process of using a dimensionality reduction algorithm to map high-dimensional data to a lower-dimensional space. The model can be updated immediately after receiving a new batch of data, thereby achieving a near-instant understanding and response to the operating status of the photovoltaic system. This rapid response capability helps to detect abnormal situations or performance degradation trends more promptly, and then take preventive measures to avoid potential problems.

[0017] 2. In the distributed photovoltaic measurement and control system based on edge computing, the power quality problem is introduced as input for optimization in the process of automatically adjusting the compensation amount according to actual needs. This not only improves the flexibility and adaptability of the reactive compensation control system, but also ensures the stable operation of the power system in the face of complex and changeable power grid environments, while improving the power quality. In addition, it also helps to reduce unnecessary energy waste and improve the efficiency and economic benefits of the entire power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is the overall flow chart of the present invention; The meaning of each number in the figure is: 1. Edge computing node unit; 11. Data aggregation and compression module; 12. Fault diagnosis module; 2. Resource optimization unit; 21. Power quality management module; 22. Reactive power compensation control module; 23. Energy storage system integration module. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Example

[0020] See also Figure 1 As shown, a distributed photovoltaic measurement and control system based on edge computing is provided, including an edge computing node unit 1 and a resource optimization unit 2; Wherein, the edge computing node unit 1 is deployed at a location close to the data source, and is used to collect real-time data from multiple distributed photovoltaic nodes, process the collected real-time data, and support localized fault diagnosis through intelligent algorithms; wherein, the edge computing node unit 1 includes a data aggregation and compression module 11 and a fault diagnosis module 12; In the edge computing node unit 1, the collected real-time data includes at least electrical parameters of voltage, current, power and frequency, at least environmental parameters of light intensity, temperature and humidity, at least meteorological parameters of wind speed, precipitation and cloud coverage, and also includes fault codes, operating time and health indicators of the equipment; In the edge computing node unit 1: the data aggregation and compression module 11 is used to aggregate and compress the real-time data collected from multiple distributed photovoltaic nodes, and use a dimensionality reduction algorithm to map high-dimensional data to a lower-dimensional space, while retaining the information of the original data; the data mapped to the lower-dimensional space is recombined into a complete file format using a hybrid compression strategy, and a method based on a combination of random projection and online learning is introduced to optimize the process of mapping high-dimensional data to a lower-dimensional space using a dimensionality reduction algorithm; the fault diagnosis module 12 is used to continuously monitor the collected data, identify deviations from the normal operating range, and predict future fault points. When a potential fault is found, the affected component is restarted and a warning notification is sent to the staff. For uncertain faults, the fault information is sent to the core node for further analysis; The fault diagnosis module 12 can respond to potential problems immediately, automatically restart the affected components to try to restore service, and send warning notifications to staff in a timely manner so that they can take quick action. For those uncertain or complex fault conditions, the fault diagnosis module can also send relevant information to the core node for more in-depth analysis. This not only improves the system's self-healing ability and operating efficiency, reduces downtime and economic losses caused by faults, but also enhances the reliability and safety of the entire system, ensuring that distributed photovoltaic power stations can provide clean energy in a long-term and stable manner. In addition, through early warning and preventive maintenance measures, the service life of the equipment can be effectively extended and long-term operating costs can be reduced.

[0021] In the data aggregation and compression module 11, a dimensionality reduction algorithm is used to map high-dimensional data to a lower-dimensional space. Principal component analysis (PCA) is a widely used technology for data dimensionality reduction. It maps high-dimensional data to a lower-dimensional space through linear transformation while retaining the information of the original data as much as possible. However, the traditional PCA method relies on directly calculating the covariance matrix and solving its eigenvalues ​​and eigenvectors to determine the direction of the principal components. However, in large data sets or streaming data processing scenarios, this method may encounter the problem of low computational efficiency. Therefore, we propose a PCA algorithm based on an incremental update mechanism, which can dynamically adjust the principal components without recalculating the entire covariance matrix, thereby improving computational efficiency, as follows: S1. First, the input data matrix Centralized processing: ; in, is the matrix after centralization; is the input data matrix; is the sample size; For the samples; ; S2. When the first batch of data is received, the initial covariance matrix is ​​constructed: ; in, is the covariance matrix of the first batch of data; is the number of samples of the first batch of data; This is the matrix after the first batch of data has been centralized; is the transpose operation; Whenever a new batch of data arrives, use the existing covariance matrix And the newly added data To update the current covariance matrix; ; in, is the current covariance matrix; is the forgetting factor (usually set to be close to 1 but less than 1), which is used to control the influence of historical data; is the covariance matrix of the previous moment; The covariance matrix describes the strength of the linear relationship between each pair of features in the data set. By constructing and dynamically updating the covariance matrix, we can capture the changing trend of the data over time and remain sensitive to the latest data without completely ignoring past information.

[0022] S3. For each updated covariance matrix , use Lanczos iteration method to perform eigenvalue decomposition and obtain the previous The Lanczos method is one of the most commonly used methods for solving the eigenvalue problem of large-scale sparse matrices. Its main idea is to tridiagonalize the symmetric matrix and perform eigendecomposition on the tridiagonal matrix to effectively approximate the very few eigenvalues ​​of the larger matrix.

[0023] S4. According to The eigenvectors form the projection matrix , and then the centralized data Projected into a new coordinate system composed of these eigenvectors to form a reduced-dimensional representation : ; in, For dimensionality reduction data; for The projection matrix consists of eigenvectors; This PCA algorithm based on the incremental update mechanism not only reduces memory usage and computing time, but also allows the system to process new incoming data in real time without stopping, which is very suitable for applications in large-scale distributed systems. In addition, by introducing the forgetting factor , and can flexibly balance the importance of past and present data, further enhancing the robustness and practicality of the model.

[0024] In the data aggregation and compression module 11, in the process of using the dimensionality reduction algorithm to map high-dimensional data to a lower-dimensional space, a method based on a combination of random projection and online learning is introduced for optimization. After optimization, the specific method is as follows: S5. After centralization, construct a random matrix: ; in, is a random matrix; is the original data dimension, is the dimension of the target after dimensionality reduction. The random matrix can be a normal distribution, a Bernoulli distribution, etc. The centralized data Projection to low-dimensional space middle: ; in, is a low-dimensional space; S6. When the first batch of data is received, the initial covariance matrix: ; in, is the covariance matrix of the first batch of data after optimization; The number of samples optimized for the first batch of data; The matrix after optimization and centralization for the first batch of data; Whenever a new batch of data arrives, use the existing covariance matrix And the newly added data Update the current covariance matrix; ; in, is the current optimized covariance matrix; is the learning rate, which controls the ratio of new and old information; is the optimized covariance matrix of the previous moment; Representation of newly added data in low-dimensional space; S7. For each updated covariance matrix , using random power iteration to estimate the The largest eigenvalue and its corresponding eigenvector ; According to the cumulative variance contribution rate, select the appropriate number of principal components : ; in, is the optimal number of principal components; For the Large eigenvalues; A user-defined threshold (e.g., 95%) that controls the amount of information retained; according to The eigenvectors form the projection matrix , and then the centralized data Projected into a new coordinate system composed of these eigenvectors, a reduced-dimensional representation is formed: ; in, is the optimized dimensionality reduction data; for The projection matrix consists of eigenvectors; In the distributed photovoltaic measurement and control system, the edge computing node unit 1 will continuously generate a large amount of information about the working status of the photovoltaic panels, such as voltage, current, temperature and other parameters. The traditional PCA method requires all historical data to be loaded into the memory for calculation, which not only consumes a lot of time and resources, but is also impractical for rapidly changing data streams. By introducing the IPCA algorithm that combines random projection and online learning, the model can be updated immediately after receiving a new batch of data, thereby achieving a near-instant understanding and response to the operating status of the photovoltaic system. This ability to respond quickly helps to detect abnormal conditions or performance degradation trends more promptly, and then take preventive measures to avoid potential problems.

[0025] Moreover, edge devices usually have certain computing power and storage space limitations, but must be able to efficiently perform complex analysis tasks locally. By dynamically selecting the number of principal components, the degree of dimensionality reduction can be flexibly adjusted according to the specific application scenario, ensuring sufficient information retention without excessively occupying hardware resources. In addition, the use of online learning mechanisms such as mini-batch gradient descent enables edge devices to continuously improve their own models without affecting normal operation, which is crucial for maintaining long-term stable monitoring results.

[0026] In the data aggregation and compression module 11, a hybrid compression strategy is used to reassemble the data mapped to the lower dimensional space into a complete file format. This framework can not only effectively compress the data to save storage space, but also ensure the quality of the data to a certain extent, especially to retain its characteristics as much as possible when restoring the original data, as follows: S11. Reduce the dimension of the data According to the degree of impact on the quality of the reconstructed image, it is classified into key parts and non-key parts; ensuring that the information that is crucial for reconstructing high-quality images is not lost. The key parts usually contain those features that have a greater impact on visual perception, while the non-key parts may be those information that will not significantly affect the overall perception even if lost; S12. Use lossless compression algorithm to compress the parts marked as critical to ensure that this part of the data can be fully restored after decompression without losing any details. Use lossy compression technology to compress the parts marked as non-critical; allow a certain degree of information loss in exchange for a higher compression rate. This method can further reduce the file size while ensuring the integrity of important information. S13, merging the above two compression results into a single file format, and recording the information about the compression method and the location index of each segment of data in the file header to facilitate the subsequent decompression operation; so that each part of the data can be accurately identified and processed during the subsequent decompression, and it is also convenient for management and transmission; S14. Adjust compression parameters according to actual conditions to achieve an optimal balance between compression ratio and quality, and verify whether the compressed data meets the expected quality standards and can be restored to an acceptable state after decompression.

[0027] The hybrid compression strategy not only helps reduce the amount of data and storage costs, but also effectively maintains the key characteristics of the data, which is particularly important when facing large-scale data sets. In this way, resource utilization and service performance can be greatly optimized without affecting the core business logic.

[0028] Resource optimization unit 2, which is used to calculate the data processed by edge computing node unit 1, and optimize the power quality management and reactive power compensation measures at the grid connection point in combination with historical data The resource optimization unit 2 includes a power quality management module 21, a reactive power compensation control module 22 and an energy storage system integration module 23; In Resource Optimization Unit 2, The power quality management module 21 is used to evaluate the power quality status at the grid connection point and identify power quality problems. Once a problem is found, the corresponding early warning mechanism is triggered. The power quality management module 21 also continuously tracks the development trend of the problem by configuring an online monitoring device; Power quality issues include voltage deviation, harmonic distortion, three-phase unbalance and transient events; The reactive power compensation control module 22 is used to automatically adjust the compensation amount according to actual needs, and supports the combination of remote control and local automated decision-making. In the process of automatically adjusting the compensation amount according to actual needs, the power quality problem is introduced as input for optimization; The reactive power compensation control module 22 is specifically as follows: Dynamic reactive power compensation: ; in, For time The reactive power compensation on the time indicates that in order to maintain voltage stability, The amount of reactive power that needs to be injected into or absorbed from the system; is the initial proportional gain coefficient, which determines the current error The degree of direct impact on reactive power compensation. It will make the system more sensitive to voltage deviations and respond to changes more quickly; It is the set target voltage, that is, the voltage level that the power system is expected to maintain, which is usually a fixed value determined by the operating standards or requirements of the power system; It is the actual voltage at the current moment. It is a variable that changes with time and reflects the actual operating status of the power system at any moment. is the initial integral gain coefficient, which affects the cumulative effect of historical errors. The function is to eliminate steady-state errors and ensure that the voltage can still return to the set value after a long period of operation; For The actual voltage of is the integral variable, representing the time from the start time to the current time All time points;

[0029] In summary, this formula uses a PI controller to dynamically adjust the reactive compensation to ensure that the voltage of the power system remains near the desired set point. The proportional part directly reflects the current error, while the integral part takes into account the accumulated errors in the past. The two work together to accurately control the voltage. In the process of automatically adjusting the compensation amount according to actual needs, the power quality problem is introduced as input for optimization. After optimization, the specific results are: Set comprehensive evaluation function , comprehensive evaluation function Different power quality issues are scored based on voltage deviation, harmonic distortion, three-phase unbalance and transient events, specifically: ; in, is the comprehensive evaluation function; For The evaluation score of voltage deviation at the moment; is the weight of the evaluation score of voltage deviation; For The evaluation score of the harmonic distortion rate at the moment; is the weight of the evaluation score of harmonic distortion rate; For The evaluation score of the three-phase imbalance at the moment; is the weight of the evaluation score of the three-phase imbalance; For The evaluation score of the transient event at the moment; The weight of the transient event evaluation score can be used to quantify the impact of different types of power quality issues on the system and adjust the reactive power compensation strategy based on these impacts. The selection of weights reflects the importance of each type of problem, allowing us to prioritize the most critical factors when designing the control system. For example, in some application scenarios, voltage deviation may be more important than other factors, so it can be given a higher weight.

[0030] According to the adjustment of proportional coefficient and integral coefficient: ; ; in, is the initial proportional gain coefficient; is the initial integral gain coefficient; is the sensitivity parameter of the proportional gain coefficient; is the sensitivity parameter of the integral gain coefficient; ; in, The optimized time Reactive power compensation amount on On the one hand, when poor power quality is detected (i.e. When the proportional gain coefficient is larger, and the integral gain factor This helps to avoid new problems caused by overcompensation, such as overcompensation or undercompensation, thereby protecting power equipment from damage; on the other hand, when the power quality is good, the larger and It can enable the control system to respond to voltage changes more quickly, provide the required reactive power support in a timely manner, and ensure that the power factor of the system is at the optimal level.

[0031] Therefore, based on the dynamic adjustment of power quality and This method not only improves the flexibility and adaptability of the reactive power compensation control system, but also ensures the stable operation of the power system while improving the power quality in the face of a complex and changeable power grid environment. In addition, this method also helps to reduce unnecessary energy waste and improve the efficiency and economic benefits of the entire power system.

[0032] The energy storage system integration module 23 is used to dispatch energy storage according to the state of power quality and reactive power compensation; specifically, the energy storage system integration module 23 can coordinate the cooperation between batteries, inverters, battery management systems, energy management systems and other electrical equipment to ensure that the energy storage system can respond quickly when needed, whether it is absorbing excess energy or releasing stored energy. Especially in the face of large fluctuations in renewable energy generation, the energy storage system integration module can help smooth output and ensure the continuity and reliability of power supply. In addition, through intelligent scheduling algorithms, the value of energy storage facilities can be maximized, operating costs can be reduced, and the overall efficiency of the power grid can be improved.

[0033] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and the above embodiments and descriptions are only preferred examples of the present invention, and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.

Claims

1. Distributed photovoltaic measurement and control system based on edge computing, characterized in that: include: An edge computing node unit (1), the edge computing node unit (1) being deployed at a location close to a data source, and being used to collect real-time data from a plurality of distributed photovoltaic nodes, process the collected real-time data, and support localized fault diagnosis through an intelligent algorithm; wherein the edge computing node unit (1) comprises a data aggregation and compression module (11) and a fault diagnosis module (12); A resource optimization unit (2), the resource optimization unit (2) being used for the data processed by the edge computing node unit (1), and optimizing the power quality management and reactive power compensation measures at the grid connection point in combination with historical data; wherein the resource optimization unit (2) comprises a power quality management module (21), a reactive power compensation control module (22) and an energy storage system integration module (23).

2. The distributed photovoltaic measurement and control system based on edge computing according to claim 1 is characterized in that: In the edge computing node unit (1), the collected real-time data includes at least electrical parameters such as voltage, current, power and frequency, at least environmental parameters such as light intensity, temperature and humidity, at least meteorological parameters such as wind speed, precipitation and cloud coverage, and also includes fault codes, operating time and health indicators of the equipment.

3. The distributed photovoltaic measurement and control system based on edge computing according to claim 2 is characterized in that: In the edge computing node unit (1): The data aggregation and compression module (11) is used to aggregate and compress the real-time data collected from multiple distributed photovoltaic nodes, map the high-dimensional data into a lower-dimensional space using a dimensionality reduction algorithm, and retain the information of the original data; The data mapped to the lower dimensional space is reassembled into a complete file format using a hybrid compression strategy, and a method based on random projection and online learning is introduced to optimize the process of mapping high dimensional data to the lower dimensional space using the dimensionality reduction algorithm; The fault diagnosis module (12) is used to continuously monitor the collected data, identify deviations from the normal operating range, and predict future fault points. When a potential fault is found, the affected component is restarted and a warning notification is sent to the staff. For uncertain faults, the fault information is sent to the core node for analysis.

4. The distributed photovoltaic measurement and control system based on edge computing according to claim 3 is characterized in that: In the data aggregation and compression module (11), a dimensionality reduction algorithm is used to map high-dimensional data to a lower-dimensional space, as follows: S1. First, the input data matrix Centralized processing: ； in, is the matrix after centralization; is the input data matrix; is the sample size; For the samples; ; S2. When the first batch of data is received, the initial covariance matrix is ​​constructed: ； in, is the covariance matrix of the first batch of data; is the number of samples of the first batch of data; This is the matrix after the first batch of data has been centralized; is the transpose operation; Whenever a new batch of data arrives, use the existing covariance matrix And the newly added data To update the current covariance matrix; ； in, is the current covariance matrix; For the forgetting factor; is the covariance matrix of the previous moment; For the newly added data; S3. For each updated covariance matrix , use Lanczos iteration method to perform eigenvalue decomposition and obtain the previous The largest eigenvalue; S4. According to The eigenvectors form the projection matrix , and then the centralized data Projected into a new coordinate system composed of these eigenvectors to form a reduced-dimensional representation : ； in, For dimensionality reduction data; for The projection matrix consists of eigenvectors.

5. The distributed photovoltaic measurement and control system based on edge computing according to claim 4 is characterized in that: In the data aggregation and compression module (11), in the process of using the dimensionality reduction algorithm to map high-dimensional data to a lower-dimensional space, a method based on a combination of random projection and online learning is introduced for optimization. After optimization, the specific method is as follows: S5. After centralization, construct a random matrix: ； in, is a random matrix; is the original data dimension, is the dimension of the target after dimensionality reduction; The centralized data Projection to low-dimensional space middle: ； in, is a low-dimensional space; S6. When the first batch of data is received, the initial covariance matrix: ； in, is the covariance matrix of the first batch of data after optimization; The number of samples optimized for the first batch of data; The matrix after optimization and centralization for the first batch of data; Whenever a new batch of data arrives, use the existing covariance matrix And the newly added data Update the current covariance matrix; ； in, is the current optimized covariance matrix; is the learning rate; is the optimized covariance matrix of the previous moment; Representation of newly added data in low-dimensional space; S7. For each updated covariance matrix , using random power iteration to estimate the The largest eigenvalue and its corresponding eigenvector ; According to the cumulative variance contribution rate, select the appropriate number of principal components : ； in, is the optimal number of principal components; For the Large eigenvalues; For user-defined thresholds; according to The eigenvectors form the projection matrix , and then the centralized data Projected into a new coordinate system composed of these eigenvectors, a reduced-dimensional representation is formed: ； in, is the optimized dimension reduction data; for The projection matrix consists of eigenvectors.

6. The distributed photovoltaic measurement and control system based on edge computing according to claim 5 is characterized in that: In the data aggregation compression module (11), a hybrid compression strategy is used to reassemble the data mapped to the lower dimensional space into a complete file format, as follows: S11. Reduce the dimension of the data According to the degree of influence on the quality of reconstructed images, they are classified into key parts and non-key parts; S12, compressing the portion marked as critical using a lossless compression algorithm, and compressing the portion marked as non-critical using a lossy compression technique; S13, merging the above two compression results into a single file format, and recording information about the compression method and the location index of each segment of data in the file header; S14. Adjust compression parameters according to actual conditions to achieve an optimal balance between compression ratio and quality, and verify whether the compressed data meets the expected quality standards and can be restored to an acceptable state after decompression.

7. The distributed photovoltaic measurement and control system based on edge computing according to claim 6 is characterized in that: In the resource optimization unit (2): The power quality management module (21) is used to evaluate the power quality status at the grid connection point and identify power quality problems. Once a problem is found, a corresponding early warning mechanism is triggered. The power quality management module (21) also continuously tracks the development trend of the problem by configuring an online monitoring device. The reactive power compensation control module (22) is used to automatically adjust the compensation amount according to actual needs, and supports the operation in a combination of remote control and local automated decision-making. In the process of automatically adjusting the compensation amount according to actual needs, power quality issues are introduced as input for optimization; The energy storage system integration module (23) is used to dispatch energy storage according to the power quality and the state of reactive power compensation.

8. The distributed photovoltaic measurement and control system based on edge computing according to claim 7 is characterized in that: The power quality issues include voltage deviation, harmonic distortion, three-phase unbalance and transient events.

9. The distributed photovoltaic measurement and control system based on edge computing according to claim 8, characterized in that: The reactive power compensation control module (22) is specifically as follows: Dynamic reactive power compensation: ； in, For time Reactive power compensation amount on is the initial proportional gain coefficient; is the set target voltage; is the actual voltage at the current moment; is the initial integral gain coefficient; For The actual voltage of is the integration variable.

10. The distributed photovoltaic measurement and control system based on edge computing according to claim 9, characterized in that: In the reactive power compensation control module (22), in the process of automatically adjusting the compensation amount according to actual demand, the power quality problem is introduced as input for optimization, and the optimization is specifically as follows: Set comprehensive evaluation function , comprehensive evaluation function Different power quality issues are scored based on voltage deviation, harmonic distortion, three-phase unbalance and transient events, specifically: ； in, is the comprehensive evaluation function; For The evaluation score of voltage deviation at the moment; is the weight of the evaluation score of voltage deviation; For The evaluation score of the harmonic distortion rate at the moment; is the weight of the evaluation score of harmonic distortion rate; For The evaluation score of the three-phase imbalance at the moment; is the weight of the evaluation score of the three-phase imbalance; For The evaluation score of the transient event at the moment; is the weight of the evaluation score of transient events; According to the adjustment of proportional coefficient and integral coefficient: ； ； in, is the initial proportional gain coefficient; is the initial integral gain coefficient; is the sensitivity parameter of the proportional gain coefficient; is the sensitivity parameter of the integral gain coefficient; ； in, The optimized time The reactive power compensation amount.

Images