Transformer parallel operation circulating current suppression method and system based on cloud edge collaborative architecture

Abstract

The invention discloses a transformer parallel operation circulating current suppression method and system based on a cloud edge collaborative architecture, and relates to the technical field of power grid control, and the method comprises the steps: an edge node collects the multi-source dynamic data of a parallel transformer in real time, and constructs a local dynamic impedance model of a single transformer; the cloud receives the multi-modal data uploaded by each edge node, inputs the multi-modal data into an LSTM-SVM hybrid model trained based on a historical database, generates an impedance prediction weight of each edge node, and issues the impedance prediction weight to the corresponding edge node; the edge node receives the impedance prediction weight and injects the impedance prediction weight into a local dynamic impedance model to generate a corrected dynamic impedance matrix; and calculating a circulating current suppression instruction according to the corrected dynamic impedance matrix, and executing real-time regulation and control based on the circulating current suppression instruction. According to the method, the impedance prediction weight is generated through real-time collection of the edge nodes and calculation of the cloud, the local dynamic impedance model is effectively corrected, and then the circulating current suppression instruction is accurately calculated.

Description

Technical Field

[0001] This invention relates to the field of power grid control technology, and in particular to a method and system for suppressing circulating currents during parallel operation of transformers based on a cloud-edge collaborative architecture. Background Technology

[0002] In modern power systems, to ensure power supply reliability, improve power supply flexibility, and meet electricity demand at different times, two or more transformers are often operated in parallel. For example, in areas with significant seasonality in electricity load, some transformers can be shut down during periods of light load to reduce no-load losses, while more transformers can be put into operation during periods of heavy load to meet demand. Statistics show that reasonable parallel operation of transformers can reduce system losses.

[0003] When traditional transformers operate in parallel, they must meet the conditions of identical connection groups, identical turns ratios, and equal short-circuit voltages to ensure no circulating current in the windings under no-load conditions and reasonable load distribution. However, in actual operation, factors such as equipment aging, changes in the operating environment, and dynamic load fluctuations can alter transformer parameters, leading to the generation of circulating currents. Circulating currents not only increase additional losses and reduce operating efficiency but can also cause overheating and shorten service life. While methods exist to suppress circulating currents by optimizing transformer selection and adjusting operating parameters, these methods struggle to accurately address complex and changing operating conditions, resulting in poor accuracy in suppressing circulating currents. Summary of the Invention

[0004] To address the issue of poor accuracy in suppressing circulating current under complex and variable operating conditions in existing technologies, this invention provides a circulating current suppression method and system for transformers operating in parallel based on a cloud-edge collaborative architecture. This method utilizes edge nodes to collect multi-source dynamic data in real time, accurately reflecting the current operating status of the transformer. Combined with the powerful computing capabilities of the cloud, impedance prediction weights are generated to effectively correct the local dynamic impedance model, thereby accurately calculating the circulating current suppression command. The specific technical solution is as follows: In a first aspect, the present invention provides a method for suppressing circulating current in parallel operation of transformers based on a cloud-edge collaborative architecture, comprising: Edge nodes collect multi-source dynamic data of parallel transformers in real time to construct a local dynamic impedance model of a single transformer. The multi-source dynamic data includes electrical quantities, thermodynamic quantities, and time-series load quantities. The cloud receives multimodal data uploaded by each edge node, inputs it into an LSTM-SVM hybrid model trained based on a historical database, generates impedance prediction weights for each edge node, and sends them to the corresponding edge nodes. The edge node receives the impedance prediction weights and injects them into the local dynamic impedance model to generate the corrected dynamic impedance matrix. The circulating current suppression command is calculated based on the corrected dynamic impedance matrix, and real-time control is performed based on the circulating current suppression command.

[0005] Preferably, the edge node collects multi-source dynamic data of the parallel transformers in real time to construct a local dynamic impedance model for a single transformer, including: The three-phase imbalance, winding hot spot temperature rise, and load mutation rate are calculated based on multi-source dynamic data. Based on the three-phase imbalance, winding hot spot temperature rise, and load mutation rate, along with the corresponding initial weight values, a local dynamic impedance model for a single transformer is constructed, expressed as: in, Let be the dynamic impedance matrix of transformer i; This is the reference impedance value; Temperature rise of winding hotspots; The mutation rate of the load; This refers to the three-phase imbalance. This is the temperature drift compensation coefficient; For load mutation gain; This is the imbalance correction factor.

[0006] Preferably, the cloud receives multimodal data uploaded by each edge node, inputs it into an LSTM-SVM hybrid model trained based on a historical database, and generates impedance prediction weights for each edge node, including: After receiving multimodal data uploaded by each edge node, the cloud extracts the corresponding electrical, thermodynamic, and temporal characteristics. The electrical, thermodynamic, and temporal characteristics are input into a historical database to train an LSTM-SVM hybrid model, which outputs the corresponding impedance prediction weights.

[0007] Preferably, the process of receiving multimodal data uploaded by each edge node in the cloud, inputting it into an LSTM-SVM hybrid model trained based on a historical database, and generating impedance prediction weights for each edge node further includes: After receiving multimodal data uploaded by each edge node, the cloud extracts the corresponding electrical, thermodynamic, and temporal characteristics. Insulation degradation factor is calculated based on historical database; The electrical characteristics, thermodynamic characteristics, temporal characteristics, and insulation degradation factors are input into a historical database to train an LSTM-SVM hybrid model, which outputs the corresponding impedance prediction weights.

[0008] Preferably, the corrected dynamic impedance matrix is ​​expressed as: in, The corrected dynamic impedance matrix for transformer i; Weights for impedance prediction; , and They are respectively from The adjusted value.

[0009] Preferably, when the transformer is in steady-state operation, the cloud outputs the impedance prediction weight according to a preset update cycle; when the transformer is in transient operation, the cloud updates the output impedance prediction weight in real time.

[0010] Preferably, the calculation of the circulating current suppression command satisfies: in, This is the corrected full dynamic impedance matrix; This is the reference voltage for the parallel bus. The measured voltage at the parallel connection point; For controlling current commands.

[0011] Secondly, the present invention also provides a transformer parallel operation circulating current suppression system based on a cloud-edge collaborative architecture, which applies the aforementioned method and includes: An edge sensing layer, deployed on each transformer body, includes a multi-source sensor group and a local dynamic modeling unit. The multi-source sensor group is used to collect multi-source dynamic data of the transformer in real time, including electrical quantities, thermodynamic quantities, and time-series load quantities. The local dynamic modeling unit is used to construct a local dynamic impedance model of a single transformer, receive the impedance prediction weights, and inject them into the local dynamic impedance model to generate a corrected dynamic impedance matrix. The cloud analytics layer receives multimodal data uploaded by each edge node, inputs it into an LSTM-SVM hybrid model trained based on a historical database, generates impedance prediction weights for each edge node, and distributes them to the corresponding edge nodes. The collaborative execution layer is used to calculate the circulating current suppression command based on the modified dynamic impedance matrix, and to perform real-time control based on the circulating current suppression command.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention presents a cloud-edge collaborative architecture-based method for suppressing circulating current in parallel operation of transformers. It collects multi-source dynamic data in real time from edge nodes, accurately reflecting the current operating status of the transformer. Combining the powerful computing capabilities of the cloud, it utilizes an LSTM-SVM hybrid model to generate impedance prediction weights, effectively correcting the local dynamic impedance model and thus accurately calculating the circulating current suppression command. Simultaneously, the cloud-edge collaborative architecture fully leverages the advantages of edge computing in rapidly processing data locally and reducing transmission burden, as well as the powerful data analysis and model training capabilities of cloud computing. Edge nodes process local data, uploading only key information to the cloud, reducing data transmission volume and latency. The cloud provides accurate prediction weights to the edge nodes, improving the overall system efficiency. Attached Figure Description

[0013] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0014] Figure 1 This is a flowchart of a method for suppressing circulating current in parallel operation of transformers based on a cloud-edge collaborative architecture, according to the present invention.

[0015] Figure 2 This is a flowchart of an embodiment of a method for suppressing circulating current in parallel operation of transformers based on a cloud-edge collaborative architecture, according to the present invention.

[0016] Figure 3 This is a schematic diagram of a transformer parallel operation circulating current suppression system based on a cloud-edge collaborative architecture according to the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] It should be understood that, when used in this specification, the terms “comprising” and “including” indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0019] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0020] It should also be further understood that the term "and / or" as used in this specification refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes such combinations.

[0021] Please refer to the following examples. Figures 1 to 3 .

[0022] This application provides a method for suppressing circulating current in parallel operation of transformers based on a cloud-edge collaborative architecture, including: Step S1: Edge nodes collect multi-source dynamic data of parallel transformers in real time to construct a local dynamic impedance model for a single transformer. The multi-source dynamic data includes electrical quantities, thermodynamic quantities, and time-series load quantities; including: By installing various sensors on the transformer body, multi-source dynamic data of each transformer operating in parallel are collected in real time. Among them, electrical quantities include real-time electrical parameters such as three-phase voltage, current, power factor, and winding resistance; thermodynamic quantities cover thermal data such as core temperature, winding temperature, ambient temperature, and temperature rise rate; and time-series load quantities are load information such as load current and load power changing over time at different moments.

[0023] The collected data is preprocessed locally at the edge nodes, including filtering and denoising. Then, a local dynamic impedance model of a single transformer is constructed using an impedance modeling algorithm. This model can reflect the real-time changes in transformer impedance under the current operating conditions.

[0024] The edge nodes collect multi-source dynamic data of the parallel transformers in real time to construct a local dynamic impedance model for a single transformer. The multi-source dynamic data includes electrical quantities, thermodynamic quantities, and time-series load quantities. The three-phase imbalance, winding hot spot temperature rise, and load mutation rate are calculated based on multi-source dynamic data. In this embodiment, three-phase current imbalance is used for calculation and analysis. Three-phase current imbalance measures the degree of imbalance of three-phase current and reflects the uniformity of current distribution in a three-phase system. The three-phase imbalance is calculated based on electrical quantity data, and the calculation formula is as follows: In the formula, It is the maximum value among the three-phase currents; This is the average value of the three-phase current; The temperature rise of the hottest spot (highest temperature point) of the winding relative to the ambient temperature (or reference temperature) reflects the influence of current on winding heating. The temperature rise of the hottest spot is calculated based on thermodynamic quantities, and the formula is as follows: In the formula, The heat dissipation coefficient is related to the equipment structure, cooling method, and environmental conditions, and reflects the effect of heat dissipation capacity on suppressing temperature rise. The current is the operating current of the winding. The larger the current, the more significant the heat generation. Furthermore, due to the nonlinearity of the winding thermal resistance, the temperature rise is related to the current to the power of 1.6.

[0025] Load mutation rate measures the drastic change in load between adjacent time points, reflecting the amplitude of dynamic load fluctuations. The load mutation rate is calculated based on time-series load data, using the following formula: In the formula, This represents the current load capacity. This represents the load capacity at the previous moment; This refers to the rated load capacity of the equipment.

[0026] Based on the three-phase imbalance, winding hot spot temperature rise, and load mutation rate, along with the corresponding initial weight values, a local dynamic impedance model for a single transformer is constructed, expressed as: in, Let be the dynamic impedance matrix of transformer i; The reference impedance value is calculated from the parameters on the transformer nameplate. Temperature rise of winding hotspots; The mutation rate of the load; This refers to the three-phase imbalance. This is the temperature drift compensation coefficient; For load mutation gain; This is the imbalance correction factor.

[0027] Step S2: The cloud receives the multimodal data uploaded by each edge node, inputs it into the LSTM-SVM hybrid model trained based on the historical database, generates the impedance prediction weights for each edge node, and sends them to the corresponding edge nodes. Each edge node uploads the processed multi-source dynamic data to the cloud platform via the communication network. The cloud platform trains an LSTM-SVM hybrid model based on a historical database (which stores long-term multimodal data of the transformer and corresponding circulating current states). The LSTM captures the temporal correlation characteristics of the time-series load and dynamic parameters, while the SVM performs regression analysis on the time-series features output by the LSTM to map the relationship between impedance and multi-source data. The cloud platform inputs the received real-time multimodal data into the hybrid model, calculates and generates impedance prediction weights for each edge node, and then distributes these weights to the corresponding edge nodes via the communication link.

[0028] The cloud receives multimodal data uploaded by each edge node, inputs it into an LSTM-SVM hybrid model trained based on a historical database, generates impedance prediction weights for each edge node, and distributes them to the corresponding edge nodes, including: After receiving multimodal data uploaded by each edge node, the cloud extracts the corresponding electrical, thermodynamic, and temporal characteristics. The cloud receives compressed feature packets from edge nodes, decompresses them, and reconstructs the original feature dimensions according to three categories: electrical, thermodynamic, and temporal, to obtain electrical features, thermodynamic features, and temporal features. The electrical, thermodynamic, and temporal features are input into a historical database to train an LSTM-SVM hybrid model, which outputs the corresponding impedance prediction weights and distributes them to the corresponding edge nodes.

[0029] Using electrical, thermodynamic, and temporal characteristics corresponding to three-phase imbalance, winding hot spot temperature rise, and load mutation rate as search keys, similar cases are matched from the historical fault database.

[0030] In the LSTM-SVM hybrid model: Input time series feature sequences in the LSTM time series prediction branch The system analyzes load mutation patterns from historical faults and outputs an impedance drift trend vector for a future period. ; The SVM anomaly classification branch takes as input a coupling matrix of electrical and thermodynamic features and outputs: (a) Normal operating conditions - tag 0; (ii) Circulation Risk Condition - Tag 1 (Trigger Weight Enhancement); (iii) Insulation failure condition - tag 2 (alarm triggered); Dynamically integrate the results from the two branches and perform weighted fusion to generate impedance prediction weights: in, , , These are the correction coefficients for the output of the SVM classifier.

[0031] Output the corresponding impedance prediction weight Distribute to the corresponding edge nodes.

[0032] Step S3: The edge node receives the impedance prediction weights and injects them into the local dynamic impedance model to generate the corrected dynamic impedance matrix. After receiving the impedance prediction weights from the cloud, the edge node injects them as correction coefficients into the local dynamic impedance model constructed in step S1.

[0033] Specifically, by adjusting the correlation weights between impedance parameters and multi-source data in the model, the model's adaptability to real-time operating conditions is optimized, ultimately generating a corrected dynamic impedance matrix. This matrix not only includes the impedance characteristics of a single transformer but also integrates global operating condition information provided by the cloud, enabling it to more accurately reflect the impedance differences among transformers in a parallel system.

[0034] The corrected dynamic impedance matrix is ​​expressed as follows: in, The corrected dynamic impedance matrix for transformer i; Weights for impedance prediction; , and They are respectively from The adjusted value.

[0035] Step S4: Calculate the circulating current suppression command based on the corrected dynamic impedance matrix, and perform real-time control based on the circulating current suppression command.

[0036] The calculation of the circulation suppression command satisfies: in, This is the corrected dynamic impedance matrix; This is the reference voltage for the parallel bus. The measured voltage at the parallel connection point; For controlling current commands.

[0037] Based on the corrected dynamic impedance matrix, the edge node determines the current circulating current value through circulating current calculation. Then, according to the preset circulating current threshold, it derives the required circulating current suppression command in reverse. For example, adjusting the transformer tap position to change the voltage ratio, or compensating for current differences through an active filter. Finally, the edge node sends the suppression command to the transformer's control execution unit, such as a smart circuit breaker or voltage regulator, to achieve real-time control of the parallel transformer's operating status, thereby suppressing the circulating current.

[0038] This embodiment of the transformer parallel operation circulating current suppression method based on a cloud-edge collaborative architecture collects multi-source dynamic data in real time through edge nodes, accurately reflecting the current operating status of the transformer. Combining the powerful computing capabilities of the cloud, an LSTM-SVM hybrid model is used to generate impedance prediction weights, effectively correcting the local dynamic impedance model and thus accurately calculating the circulating current suppression command. The cloud-edge collaborative architecture fully leverages the advantages of edge computing in rapidly processing data locally and reducing transmission burden, as well as the powerful data analysis and model training capabilities of cloud computing. Edge nodes process local data and upload key information to the cloud, reducing data transmission volume and latency. Simultaneously, the cloud provides accurate prediction weights to the edge nodes; the two work together to improve the overall system operating efficiency.

[0039] Specifically, in a preferred embodiment of this application, the process of receiving multimodal data uploaded by each edge node from the cloud, inputting it into an LSTM-SVM hybrid model trained based on a historical database, and generating impedance prediction weights for each edge node further includes: Step S201: After receiving the multimodal data uploaded by each edge node, the cloud extracts the corresponding electrical features, thermodynamic features and temporal features; The cloud receives compressed feature packets from edge nodes, decompresses them, and reconstructs the original feature dimensions according to three categories: electrical, thermodynamic, and temporal, resulting in electrical features, thermodynamic features, and temporal features. Using the three-phase imbalance, winding hot spot temperature rise, and load mutation rate corresponding to the electrical features, thermodynamic features, and temporal features as search keys, similar cases are matched from the historical fault database.

[0040] Step S202: Calculate the insulation degradation factor based on the historical database; Insulation degradation factor injection involves extracting the current equipment aging coefficient from matching cases in the historical database and calculating the insulation degradation factor using the Arrhenius equation. In the formula, Indicates the insulation degradation factor; is a pre-exponential factor, an intrinsic constant related to the chemical properties and molecular structure of insulating materials; The activation energy of a single molecule; Boltzmann's constant; This refers to the temperature of the transformer oil.

[0041] Step S203: Input the electrical characteristics, thermodynamic characteristics, temporal characteristics, and insulation degradation factor into the LSTM-SVM hybrid model based on the historical database, and output the corresponding impedance prediction weights.

[0042] In the LSTM-SVM hybrid model: The LSTM timing prediction branch takes into input timing feature sequences, load mutation modes in historical faults, and insulation degradation factors, and outputs an impedance drift trend vector for a future period of time. The SVM anomaly classification branch takes as input a coupling matrix of electrical and thermodynamic features and outputs: (a) Normal operating conditions - tag 0; (ii) Circulation Risk Condition - Tag 1 (Trigger Weight Enhancement); (iii) Insulation failure condition - tag 2 (alarm triggered); The results from the two branches are dynamically integrated and weighted. In this embodiment, the impedance prediction weights are generated as follows: in, , , These are the correction coefficients for the output of the SVM classifier.

[0043] Output the corresponding impedance prediction weight Distribute to the corresponding edge nodes.

[0044] In this embodiment, the insulation degradation factor quantifies the aging degree of the transformer insulation system and incorporates it into the model input, enabling the LSTM-SVM hybrid model to simultaneously consider short-term dynamic operating conditions and long-term aging effects. This overcomes the limitations of traditional models that rely solely on real-time parameters and ignore the impact of equipment aging on impedance, making it particularly suitable for transformers with long service lives and improving the accuracy of impedance prediction weights. By fusing multiple features with insulation degradation factors as input, LSTM can more accurately capture the long-term correlation between temporal features and insulation degradation, including the temporal coupling between aging rate and load fluctuations. SVM, on the other hand, can more efficiently perform nonlinear mapping on multi-dimensional features, reducing weight prediction bias caused by missing features. The resulting impedance prediction weights not only reflect real-time operating conditions but also incorporate equipment health status information, providing a more targeted basis for correcting the dynamic impedance model of edge nodes.

[0045] Incorporating an insulation degradation factor ensures accurate impedance prediction weights in the model output, allowing the dynamic impedance matrix correction at edge nodes to better reflect the actual transformer conditions, including short-term operating fluctuations and long-term aging trends. This avoids impedance model mismatch caused by weight bias. The corrected matrix more accurately reflects the impedance differences between parallel transformers, ensuring the accuracy of circulating current calculations and suppression commands, and reducing ineffective regulation.

[0046] Specifically, in a preferred embodiment of this application, when the transformer is in a steady-state operation, the cloud outputs the impedance prediction weight according to a preset update cycle; when the transformer is in a transient operation, the cloud updates the output impedance prediction weight in real time.

[0047] In practice, when the transformer is in steady-state operation, the cloud updates the output impedance prediction weights every 5 minutes; when the transformer is in transient operation, the cloud iterates and updates the output impedance prediction weights in real time with a step size of 1ms.

[0048] In this embodiment, edge nodes collect data in real time at a high frequency, coupled with a transient condition update strategy that iterates in real time with a 1ms step size, to quickly capture instantaneous changes in the transformer's operating status. When transient situations such as short-circuit faults occur, a response can be made in a short time, adjusting operating parameters in a timely manner to prevent the fault from escalating. Whether it is refreshing the matrix every 5 minutes under steady-state conditions to track slowly changing operating parameters, or rapidly iterating in real time to respond to sudden situations under transient conditions, this embodiment can adapt well. It can cope with complex operating conditions such as different types of load mutations and large fluctuations in ambient temperature and humidity, enhancing the adaptability of the power system to various operating conditions.

[0049] Secondly, embodiments of this application also provide a transformer parallel operation circulating current suppression system based on a cloud-edge collaborative architecture, which applies the aforementioned method and includes: An edge sensing layer, deployed on each transformer body, includes a multi-source sensor group and a local dynamic modeling unit. The multi-source sensor group is used to collect multi-source dynamic data of the transformer in real time, including electrical quantities, thermodynamic quantities, and time-series load quantities. The local dynamic modeling unit is used to construct a local dynamic impedance model of a single transformer, receive the impedance prediction weights, and inject them into the local dynamic impedance model to generate a corrected dynamic impedance matrix. The cloud analytics layer receives multimodal data uploaded by each edge node, inputs it into an LSTM-SVM hybrid model trained based on a historical database, generates impedance prediction weights for each edge node, and distributes them to the corresponding edge nodes. The collaborative execution layer is used to calculate the circulating current suppression command based on the modified dynamic impedance matrix, and to perform real-time control based on the circulating current suppression command.

[0050] The system achieves precise suppression of circulating current during parallel operation of transformers through a three-layer architecture. The collaborative process of each layer is as follows: A multi-source sensor array is deployed on each transformer to collect electrical, thermodynamic, and time-series load data in real time, and preprocesses the data locally. A local dynamic modeling unit constructs a local dynamic impedance model for each transformer based on the collected real-time data and the transformer's inherent parameters. Edge nodes upload the preprocessed multimodal data to the cloud. The cloud calls an LSTM-SVM hybrid model trained on a historical database, combining extracted electrical, thermodynamic, and time-series features to generate impedance prediction weights for each edge node through model analysis, and then distributes these weights to the corresponding edge nodes. After receiving the impedance prediction weights from the cloud, the edge nodes inject them into their local dynamic impedance models to generate a corrected dynamic impedance matrix. The collaborative execution layer calculates the current circulating current value based on the corrected matrix using a circulating current calculation model, compares it with a preset threshold to derive suppression commands, and drives the execution unit to adjust in real time, ultimately achieving circulating current suppression.

[0051] The functional explanations of each unit in this embodiment are the same as those of a transformer parallel operation optimization control method, and the technical effects are the same, so they will not be repeated here.

[0052] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical functional division. In actual implementation, there may be other division methods, such as multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored.

[0053] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0054] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0055] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the specification of the present invention.

Claims

1. A method for suppressing circulating current in parallel operation of transformers based on a cloud-edge collaborative architecture, characterized in that, include: Edge nodes collect multi-source dynamic data of parallel transformers in real time to construct a local dynamic impedance model of a single transformer. The multi-source dynamic data includes electrical quantities, thermodynamic quantities, and time-series load quantities. The cloud receives multimodal data uploaded by each edge node, inputs it into an LSTM-SVM hybrid model trained based on a historical database, generates impedance prediction weights for each edge node, and sends them to the corresponding edge nodes. The edge node receives the impedance prediction weights and injects them into the local dynamic impedance model to generate the corrected dynamic impedance matrix. The circulating current suppression command is calculated based on the corrected dynamic impedance matrix, and real-time control is performed based on the circulating current suppression command.

2. The method for suppressing circulating current in parallel operation of transformers based on a cloud-edge collaborative architecture according to claim 1, characterized in that, The edge nodes collect multi-source dynamic data of the parallel transformers in real time, and construct a local dynamic impedance model for a single transformer, including: The three-phase imbalance, winding hot spot temperature rise, and load mutation rate are calculated based on multi-source dynamic data. Based on the three-phase imbalance, winding hot spot temperature rise, and load mutation rate, along with the corresponding initial weight values, a local dynamic impedance model for a single transformer is constructed, expressed as: in, Let be the dynamic impedance matrix of transformer i; This is the reference impedance value; Temperature rise of winding hotspots; The mutation rate of the load; This refers to the three-phase imbalance. This is the temperature drift compensation coefficient; For load mutation gain; This is the imbalance correction factor.

3. The method for suppressing circulating current in parallel operation of transformers based on a cloud-edge collaborative architecture according to claim 1, characterized in that, The cloud receives multimodal data uploaded by each edge node, inputs it into an LSTM-SVM hybrid model trained based on a historical database, and generates impedance prediction weights for each edge node, including: After receiving multimodal data uploaded by each edge node, the cloud extracts the corresponding electrical, thermodynamic, and temporal characteristics. The electrical, thermodynamic, and temporal characteristics are input into a historical database to train an LSTM-SVM hybrid model, which outputs the corresponding impedance prediction weights.

4. The method for suppressing circulating current in parallel operation of transformers based on a cloud-edge collaborative architecture according to claim 1, characterized in that, The cloud-based system receives multimodal data uploaded from each edge node, inputs it into an LSTM-SVM hybrid model trained based on a historical database, and generates impedance prediction weights for each edge node, which also includes: After receiving multimodal data uploaded by each edge node, the cloud extracts the corresponding electrical, thermodynamic, and temporal characteristics. Insulation degradation factor is calculated based on historical database; The electrical characteristics, thermodynamic characteristics, temporal characteristics, and insulation degradation factors are input into a historical database to train an LSTM-SVM hybrid model, which outputs the corresponding impedance prediction weights.

5. A method for suppressing circulating current in parallel operation of transformers based on a cloud-edge collaborative architecture, as described in claim 3 or 4, characterized in that, The corrected dynamic impedance matrix is ​​expressed as follows: in, The corrected dynamic impedance matrix for transformer i; Weights for impedance prediction; , and They are respectively from The adjusted value.

6. The method for suppressing circulating current in parallel operation of transformers based on a cloud-edge collaborative architecture according to claim 1, characterized in that, When the transformer is in steady-state operation, the cloud outputs the impedance prediction weight according to the preset update cycle; when the transformer is in transient operation, the cloud updates the output impedance prediction weight in real time.

7. The method for suppressing circulating current in parallel operation of transformers based on a cloud-edge collaborative architecture according to claim 1, characterized in that, The calculation of the circulation suppression command satisfies: in, This is the corrected full dynamic impedance matrix; This is the reference voltage for the parallel bus. The measured voltage at the parallel connection point; For controlling current commands.

8. A circulating current suppression system for transformers operating in parallel based on a cloud-edge collaborative architecture, characterized in that, The method described by any one of claims 1 to 7 includes: An edge sensing layer, deployed on each transformer body, includes a multi-source sensor group and a local dynamic modeling unit. The multi-source sensor group is used to collect multi-source dynamic data of the transformer in real time, including electrical quantities, thermodynamic quantities, and time-series load quantities. The local dynamic modeling unit is used to construct a local dynamic impedance model of a single transformer, receive the impedance prediction weights, and inject them into the local dynamic impedance model to generate a corrected dynamic impedance matrix. The cloud analytics layer receives multimodal data uploaded by each edge node, inputs it into an LSTM-SVM hybrid model trained based on a historical database, generates impedance prediction weights for each edge node, and distributes them to the corresponding edge nodes. The collaborative execution layer is used to calculate the circulating current suppression command based on the modified dynamic impedance matrix, and to perform real-time control based on the circulating current suppression command.

Images