A multi-dry transformer group sub-coordination and cluster allocation scheduling system

Abstract

The application discloses a multi-dry-type transformer group coordination and cluster allocation scheduling system, belongs to the technical field of power systems, and realizes safe parallel connection of transformers of different groups by recognizing actual connection groups according to working condition data through a group adaptation module, utilizing phase matching and impedance dynamic balancing technology. On the basis, a "cluster division-load prediction-multi-objective optimization scheduling" coordination architecture is constructed, the time and space balance of loads in the group is realized based on cluster division, a comprehensive optimization model is established by taking the minimum system total loss as the target in combination with future load prediction, the particle swarm algorithm is adopted to solve the model, and safety constraints such as group compatibility, capacity, voltage and temperature are integrated, global optimal allocation is realized, technical difficulties such as difficulty in parallel connection of transformers of different groups, increased circulating current caused by parameter drift and lack of coordinated optimization in cluster scheduling are effectively solved, and the safety, economy and intelligent level of transformer parallel operation are improved.

Description

Technical Field

[0001] This invention relates to the field of power system technology, and more specifically, to a multi-unit dry-type transformer group coordination and cluster load dispatching system. Background Technology

[0002] In power distribution networks, the parallel operation of multiple dry-type transformers is widely used in scenarios requiring flexible capacity expansion or redundant power supply. In recent years, relevant technical solutions have been proposed for the clustered management of multiple transformers. For example, Chinese invention patent CN113077020A discloses a transformer cluster management method and system, which uses a hierarchical clustering algorithm to group transformers with similar ledger parameters and operating condition data into the same transformer cluster. Then, by comparing the status monitoring data of transformers in the same cluster, the differences between the equipment in the cluster can be understood, and it can be quickly determined whether a transformer is in an abnormal state. However, the above technologies mainly focus on the static division and anomaly detection of transformer clusters. In the engineering practice of parallel operation of transformers, the following technical problems still cannot be effectively solved:

[0003] First, it is difficult for dry-type transformers with different connection groups to operate in parallel: the parallel operation of transformers requires conditions such as equal turns ratio, the same connection group, and the same per-unit short-circuit impedance. However, in actual engineering, multiple dry-type transformers with different connection groups often coexist in the same substation. If they are directly connected in parallel, the phase difference caused by the different connection groups will generate a huge circulating current between the two transformers, instantly burning out the windings and seriously threatening equipment safety and power supply reliability.

[0004] Second, dynamic parameter drift leads to increased circulating current in parallel operation: Even if the transformers initially meet the parallel conditions, during long-term operation, due to factors such as equipment aging and temperature changes, the short-circuit impedance, turns ratio and other parameters of each transformer will dynamically drift, resulting in a gradual increase in circulating current and uneven load distribution during parallel operation. Transformers with larger impedances will be overloaded, while those with smaller impedances will be lightly loaded, which seriously affects the lifespan of the equipment and the operating efficiency of the system.

[0005] Third, cluster scheduling lacks an integrated mechanism for group coordination and load allocation optimization: although traditional solutions such as CN113077020A can divide transformer clusters and detect anomalies, they do not solve the problem of coordinated operation control of transformers within the cluster, and lack the ability to have a global perception of the overall operating status of multiple transformers and dynamic load allocation optimization capabilities.

[0006] Therefore, there is an urgent need to propose a multi-dry-type transformer group coordination and cluster load scheduling system that can solve the above problems. Summary of the Invention

[0007] The purpose of this invention is to address the aforementioned technical deficiencies by providing a system for coordinated grouping and clustered load scheduling of multiple dry-type transformers.

[0008] The objective of this invention can be achieved through the following technical solution: a multi-unit dry-type transformer group coordination and cluster load dispatching system, comprising:

[0009] The operation sensing and storage module is used to collect the operating status data and store the ledger parameters of each dry-type transformer. The operating status data includes operating condition data, thermal status data, load demand data and switch status data.

[0010] The group adaptation module is used to identify the actual connection group of each dry-type transformer based on the operating condition data, and realize the parallel and coordinated operation of dry-type transformers with different connection groups through phase matching and impedance balancing, and synchronously output the group compatibility constraints.

[0011] The cluster partitioning module is used to receive operating condition data, ledger parameters, and thermal status data, partition dry-type transformers with similar characteristics into clusters, perform anomaly detection and dynamic cluster updates, and output the management group partitioning results.

[0012] The load forecasting module is used to perform multi-time-scale load forecasting based on load demand data and historical load data, and output load semantic features.

[0013] The cluster load allocation collaborative optimization scheduling module is used to construct a comprehensive optimization model based on operating status data, load semantic characteristics, group compatibility constraints, and management group division results. It uses a rolling optimization mechanism to solve the model in each scheduling cycle and generate scheduling control instructions.

[0014] Furthermore, the operating condition data includes three-phase voltage, three-phase current, active power, reactive power, harmonic content, and three-phase imbalance; thermal condition data includes winding temperature; switch condition data includes circuit breaker opening and closing status; load demand data includes instantaneous values ​​of total active load and total reactive load; and ledger parameters include transformer capacity, impedance voltage, connection group, commissioning time, no-load loss, short-circuit loss, and rated capacity.

[0015] Furthermore, the group adaptation module includes:

[0016] The group identification and phase matching submodule is used to identify the actual connection group of each dry-type transformer based on the phase difference data of the three-phase voltage, and adjust the output voltage of different connection groups to be consistent with the phase through the intelligent power electronic interface device configured on the low-voltage side of each dry-type transformer, so that the dry-type transformers of different connection groups can be connected in parallel on the common bus.

[0017] The impedance detection and equalization submodule is used to detect the equivalent impedance of each dry-type transformer in real time based on the response data of three-phase voltage and three-phase current, calculate the equivalent impedance deviation between each transformer, and when it exceeds the preset threshold, adjust the output characteristics of the intelligent power electronic interface device to make the equivalent impedance of each transformer tend to be balanced, and suppress the circulating current in parallel operation to within the preset ratio of the rated current.

[0018] The constraint generation submodule is used to generate group compatibility constraints based on the phase matching completion status and impedance equalization results, including the flag phases that allow parallel connection and the impedance deviation thresholds between each transformer.

[0019] Furthermore, the process of clustering dry-type transformers with similar characteristics includes:

[0020] Feature parameters are extracted from operating condition data, ledger parameters, and thermal state data and combined to form feature vectors. Correlation analysis is used to remove redundant features to obtain the optimal feature set. Feature similarity between dry-type transformers is calculated. Dry-type transformers with feature similarity higher than the preset similarity threshold are assigned to the same transformer cluster. The transformer cluster number and the list of transformer identifiers contained in each cluster are used as the management group division result.

[0021] Furthermore, the process of anomaly detection and dynamic cluster updates includes:

[0022] The system monitors the operational status data of the same transformer cluster in real time, calculates the cluster average value of each parameter of the operational status data in each monitoring period, and determines that the dry-type transformer is in an abnormal state and triggers an early warning when the operational status data of a certain dry-type transformer deviates from the cluster average value by more than the set deviation warning threshold. When a dry-type transformer in a transformer cluster exits due to an abnormal state or a new dry-type transformer is put into operation, the cluster division is re-triggered to perform dynamic cluster updates.

[0023] Furthermore, the load forecasting module performs multi-timescale load forecasting as follows:

[0024] Historical load demand data is obtained as historical load data. Discrete wavelet transform is used to decompose the historical load data into multi-scale data to obtain multi-time granularity load data sequences. Recurrent neural networks are used to extract time-series features from the multi-time granularity load data sequences to obtain load density prediction data. Predicted load curves are constructed, and a large language model is used to perform semantic analysis on the changing trend of load density prediction data to extract load semantic features, including peak load, duration, and occurrence period.

[0025] Furthermore, when constructing the comprehensive optimization model, the optimization objectives are to minimize the total system loss, maximize the load balance, and optimize the voltage qualification rate. The optimization variables are the circuit breaker opening and closing status and the active power output distribution ratio and reactive power output distribution ratio of each dry-type transformer. The constraints are capacity constraints, voltage constraints, unbalance constraints, temperature constraints, and group compatibility constraints.

[0026] The model's inputs include active power, reactive power, three-phase voltage, three-phase current, circuit breaker status, management group division results, no-load loss, short-circuit loss, rated capacity, predicted load curve, load semantic characteristics, and group compatibility constraints; the outputs include: switching commands for each circuit breaker, active power output allocation ratio for each dry-type transformer, reactive power output allocation ratio, and on-load tap changer adjustment commands.

[0027] Furthermore, in the comprehensive optimization model:

[0028] The total system loss is the sum of the no-load loss of each transformer and the square of the load rate multiplied by the short-circuit loss. The load rate is obtained by dividing the active power by the rated capacity. The load balance is calculated by summing the load rate variances according to the management group. The voltage qualification rate is the sum of the squares of the differences between the three-phase voltage and the reference value.

[0029] The capacity constraint is that the load rate of each transformer does not exceed the maximum allowable load rate corresponding to its rated capacity; the voltage constraint is that the three-phase voltage amplitude on the low-voltage side of each transformer is maintained within the upper and lower limits specified by the power grid; the unbalance constraint is that the three-phase unbalance does not exceed the preset threshold; the temperature constraint is that the winding temperature does not exceed the maximum allowable temperature of its insulation class; the group compatibility constraint is that dry-type transformers operating in parallel on the same busbar must meet the requirements of phase matching and equivalent impedance deviation not exceeding the impedance deviation threshold.

[0030] Compared with the prior art, the advantages of this invention are:

[0031] 1. This invention identifies the actual connection group of each dry-type transformer based on operating condition data. Through phase matching and impedance dynamic balancing technology, it not only enables dry-type transformers with different connection groups to operate safely in parallel, effectively eliminating the risk of burnout caused by different connection groups, but also tracks and compensates for parameter drift caused by equipment aging or temperature changes in real time, and stably suppresses the parallel circulating current within the preset safe ratio of the rated current, significantly improving the compatibility, stability and long-term reliability of multiple dry-type transformers operating in parallel.

[0032] 2. Based on group adaptation, this invention achieves three core beneficial effects by constructing a collaborative architecture of "cluster partitioning → load prediction → multi-objective optimized scheduling":

[0033] First, it is based on the cluster partitioning results for intra-group load spatiotemporal balancing, combined with proactive pre-scheduling driven by load semantic features, which effectively avoids the defects of traditional "one-size-fits-all" balancing.

[0034] Second, with the goal of minimizing the total system loss, a comprehensive optimization model is established by integrating no-load loss, load rate and short-circuit loss, and the particle swarm algorithm is used to solve the problem in a rolling manner, so that the transformer cluster can operate in the optimal operating range for a long time.

[0035] Third, the comprehensive optimization model incorporates group compatibility constraints (phase matching and impedance deviation meeting standards) as well as multiple safety boundaries such as capacity, voltage, and temperature, ensuring safe parallel operation of transformers in different groups while achieving global optimal load distribution. It also has the ability to adaptively update faults, significantly improving the safety, economy, and intelligence of parallel operation of multiple dry-type transformers. Attached Figure Description

[0036] Figure 1 This is a block diagram illustrating the system module principle of the present invention;

[0037] Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0038] 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 embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0039] Example 1: This invention discloses a multi-unit dry-type transformer group coordination and cluster load dispatching system. Please refer to [link / reference]. Figures 1-2 It includes a runtime perception and storage module, a group adaptation module, a cluster partitioning module, a load prediction module, and a cluster load allocation collaborative optimization scheduling module.

[0040] The operation sensing and storage module is used to collect the operating status data of each dry-type transformer and store the ledger parameters and historical load data. The operating status data includes operating condition data, thermal status data, load demand data and switch status data. Through outlier removal and sliding filter preprocessing, random errors and gross errors of each sensor are effectively eliminated to ensure data quality.

[0041] Specifically, the operating condition data includes three-phase voltage, three-phase current, active power, reactive power, harmonic content, and three-phase imbalance; the thermal condition data includes winding temperature; the switch condition data includes circuit breaker opening and closing status; the load demand data includes instantaneous values ​​of total active load and total reactive load; and the ledger parameters include transformer capacity, impedance voltage, connection group, commissioning time, no-load loss, short-circuit loss, and rated capacity.

[0042] The group adaptation module is connected to the operation sensing and storage module and the intelligent power electronic interface device of each dry-type transformer, respectively. It is used to identify the actual connection group of each dry-type transformer according to the operating condition data, and realize the parallel and coordinated operation of dry-type transformers with different connection groups through phase matching and impedance equalization processing, and synchronously output the group compatibility constraints.

[0043] The group adaptation module specifically includes:

[0044] The group identification and phase matching submodule is used to identify the actual connection group of each dry-type transformer based on the phase difference data of the three-phase voltage, and adjust the output voltage of different connection groups to be consistent with the phase through the intelligent power electronic interface device configured on the low-voltage side of each dry-type transformer, so that the dry-type transformers of different connection groups can be connected in parallel on the common bus.

[0045] The impedance detection and equalization submodule is used to detect the equivalent impedance of each dry-type transformer in real time based on the response data of three-phase voltage and three-phase current, calculate the equivalent impedance deviation between each transformer, and when the equivalent impedance deviation exceeds the preset threshold, adjust the output characteristics of the intelligent power electronic interface device to make the equivalent impedance of each transformer tend to be balanced, and suppress the circulating current during parallel operation to within the preset ratio of the rated current.

[0046] The constraint generation submodule is used to generate group compatibility constraints based on the phase matching completion status and impedance equalization results, including the flag phase that allows parallel connection and the impedance deviation threshold between each transformer.

[0047] By using phase matching and impedance dynamic balancing technology, dry-type transformers of different connection groups can be safely connected in parallel, eliminating the risk of burnout caused by different groups; real-time tracking and compensation for parameter drift caused by equipment aging or temperature changes, and stable suppression of parallel circulating current within the preset safe ratio of rated current, significantly improving the compatibility, stability and long-term reliability of parallel operation.

[0048] The cluster partitioning module, connected to the operation perception and storage module, is used to receive operating condition data, ledger parameters, thermal status data, partition dry-type transformers with similar characteristics into clusters, output the management group partitioning results, and perform cluster dry-type transformer anomaly detection and cluster dynamic updates.

[0049] The process of grouping dry-type transformers with similar characteristics into clusters includes:

[0050] Receive operating condition data, ledger parameters, and thermal status data. Extract three-phase voltage, three-phase current, active power, harmonic content, and three-phase imbalance from the operating condition data. Extract transformer capacity, impedance voltage, connection group, and commissioning time from the ledger parameters. Extract winding temperature from the thermal status data.

[0051] The load rate of each dry-type transformer is calculated based on the active power and the preset rated capacity. The voltage deviation of each dry-type transformer is calculated based on the three-phase voltage. The load rate, voltage deviation and the extracted feature parameters are combined to form a feature vector. The correlation analysis method is used to delete redundant features to obtain the optimal feature set. The feature similarity between each dry-type transformer is calculated based on Euclidean distance. Dry-type transformers with feature similarity higher than the preset similarity threshold are assigned to the same transformer cluster. The transformer cluster number and the list of transformer identifiers contained in each cluster are used as the management group division result.

[0052] The process of anomaly detection and dynamic cluster updates includes:

[0053] Anomaly Detection: Real-time monitoring of the operating status data of the same transformer cluster, calculation of the cluster average value of each parameter of the operating status data in each monitoring period, when the operating status data of a certain dry-type transformer deviates from the cluster average value by more than the set deviation warning threshold, the operating status data exceeding the set deviation warning threshold is marked as abnormal operating status data, the dry-type transformer is determined to be in an abnormal state and an early warning is triggered. By continuously monitoring the deviation from the cluster average value, early signs of performance degradation can be identified (such as a slow rise in winding temperature and a gradual increase in harmonic content). After the early warning is triggered, maintenance personnel can arrange targeted maintenance to avoid unplanned power outages caused by sudden faults. Dry-type transformers in abnormal states can be repaired or removed. On the one hand, this prevents abnormal transformers from polluting the cluster average value and ensures the real-time maintenance of the consistency of characteristics within the cluster. On the other hand, it provides a safety constraint boundary for the subsequent cluster load distribution collaborative optimization scheduling module.

[0054] Cluster dynamic update: When a dry-type transformer in a transformer cluster exits due to an abnormal state or a new dry-type transformer is put into operation, the feature extraction and fusion processing is re-triggered to obtain the current optimal feature set. Based on the current optimal feature set, a new cluster configuration scheme is formed to adapt to the changes in cluster members caused by the commissioning and decommissioning of dry-type transformers. This ensures that the cluster partitioning result is always consistent with the physical equipment status, provides an accurate intra-group equilibrium calculation benchmark for the comprehensive optimization model built in the cluster load allocation collaborative optimization scheduling module, and supports the adaptive closed loop of rolling optimization.

[0055] The cluster partitioning module groups transformers with similar electrical characteristics into the same management group, providing a reasonable grouping calculation benchmark for the load balance index. This requires that the load rates of transformers within a group be similar, while allowing differences between different groups, thus avoiding the unreasonable allocation caused by the "one-size-fits-all" balancing method in traditional approaches.

[0056] The load forecasting module, connected to the operation sensing and storage module, is used to perform multi-time-scale load forecasting based on load demand data and historical load data, and outputs the forecasted load curve and load semantic features. The main processes include:

[0057] Historical load data is obtained by acquiring historical load demand data. Historical load data from the past 7 days with a sampling interval of 15 minutes is obtained. Discrete wavelet transform is used to decompose the historical load data into multi-scale components, which are high-frequency detail components (short-term fluctuations) and low-frequency approximate components (trend and periodic components), forming a multi-time-granularity load data sequence. Recurrent neural network is used to extract time-series features from the multi-time-granularity load data sequence to obtain load density prediction data and construct a predicted load curve.

[0058] A large language model is used to perform semantic analysis on the changing trend of load density prediction data and extract load semantic features, including peak load, duration, and occurrence time. The large language model here is a large-scale natural language processing model based on deep learning. It usually adopts the Transformer architecture and learns the syntax, semantics and context of language by pre-training on massive text data. It has the ability to understand, generate, reason and summarize text.

[0059] Based on cluster partitioning, the load forecasting module captures semantic features such as the peak value, duration, and time of occurrence of future loads in advance, enabling the scheduling module to predict load change trends and proactively add transformers before peak hours and perform rotational maintenance during off-peak hours, thus achieving a leap from "passive response" to "proactive pre-scheduling".

[0060] The cluster load allocation collaborative optimization scheduling module is connected to all the above modules. It is used to construct a comprehensive optimization model based on the running status data, load semantic features, group compatibility constraints, and management group division results. It also uses a rolling optimization mechanism to solve the model in each scheduling cycle and generate scheduling control instructions.

[0061] When constructing a comprehensive optimization model, the cluster load allocation and collaborative optimization scheduling module takes the minimum total system loss, the highest load balance, and the optimal voltage qualification rate as the optimization objectives, the circuit breaker opening and closing status and the active power output allocation ratio and reactive power output allocation ratio of each dry-type transformer as optimization variables, and capacity constraints, voltage constraints, unbalance constraints, temperature constraints and group compatibility constraints as constraints.

[0062] The model's inputs include active power, reactive power, three-phase voltage, three-phase current, circuit breaker status, management group division results, no-load loss, short-circuit loss, rated capacity, predicted load curve, load semantic characteristics, and group compatibility constraints; the outputs include: switching commands for each circuit breaker, active power output allocation ratio for each dry-type transformer, reactive power output allocation ratio, and on-load tap changer adjustment commands.

[0063] The comprehensive optimization model is updated using a rolling optimization mechanism. Specifically, at the beginning of each scheduling cycle, the current running state data is used as the initial state, and the particle swarm optimization algorithm is used to solve the comprehensive optimization model to generate scheduling control instructions. At the beginning of the next cycle, the model is solved again with the latest data, forming a rolling optimization closed loop.

[0064] The total system loss is the sum of the no-load loss of each transformer and the square of the load rate multiplied by the short-circuit loss. The load rate is obtained by dividing the active power by the rated capacity. The load balance is calculated by summing the load rate variances according to the management group. The voltage qualification rate is the sum of the squares of the differences between the three-phase voltage and the reference value.

[0065] The active power output allocation ratio is defined as: the ratio of the active power allocated to each dry-type transformer to the predicted total active power load at that moment in the predicted load curve, provided that the total load demand is met; the value ranges from 0 to 1 and is a continuous real number. The reactive power output allocation ratio is defined as: the ratio of the reactive power allocated to each dry-type transformer to the predicted total reactive power load; the value ranges from 0 to 1 and is a continuous real number. The circuit breaker's open / closed state is a binary variable, with a value of 0 indicating open and a value of 1 indicating closed.

[0066] The capacity constraint is that the load rate of each transformer does not exceed the maximum allowable load rate corresponding to its rated capacity; the voltage constraint is that the three-phase voltage amplitude on the low-voltage side of each transformer is maintained within the upper and lower limits specified by the power grid; the unbalance constraint is that the three-phase unbalance does not exceed the preset threshold; the temperature constraint is that the winding temperature does not exceed the maximum allowable temperature of its insulation class; the group compatibility constraint is that dry-type transformers operating in parallel on the same busbar must meet the requirements of phase matching and equivalent impedance deviation not exceeding the impedance deviation threshold.

[0067] It should be added that the article involves comparisons of various thresholds. Thresholds, preset values, preset ranges, etc., are set for result comparison and analysis to determine good or bad. The magnitude of these thresholds is determined by a combination of large-scale model analysis of sample data and human experience. They can also be appropriately adjusted based on seasonal or common-sense influencing factors.

[0068] In summary, the group adaptation module identifies the actual connection group based on the operating condition data, and uses phase matching and impedance dynamic balancing technology to enable dry-type transformers of different connection groups to operate safely in parallel. It compensates for parameter drift in real time and suppresses circulating current within the safe ratio of rated current, significantly improving the compatibility, stability and long-term reliability of parallel operation.

[0069] Based on this, a collaborative architecture of "cluster partitioning → load forecasting → multi-objective optimization scheduling" is constructed. Based on cluster partitioning, the spatiotemporal balance of load within the group is achieved. Combined with proactive pre-scheduling driven by load semantic features, a comprehensive optimization model is established with the goal of minimizing the total system loss. The particle swarm optimization algorithm is used for rolling solution, and group compatibility constraints and multiple safety boundaries such as capacity, voltage, and temperature are incorporated to ensure safe parallel operation while achieving global optimal load distribution. This effectively solves technical problems such as the difficulty of paralleling transformers of different groups, the increase in circulating current due to parameter drift, and the lack of collaborative optimization in cluster scheduling. It significantly improves the safety, economy, and intelligence level of parallel operation of multiple dry-type transformers.

[0070] The above description is merely a preferred embodiment of the present invention; however, the scope of protection of the present invention is not limited thereto; any equivalent substitutions or modifications made by those skilled in the art within the technical scope disclosed in the present invention, based on the technical solution and its improved concept, should be covered within the scope of protection of the present invention.

Claims

1. A multi-unit dry-type transformer group coordination and cluster load dispatching system, characterized in that: include: The operation sensing and storage module is used to collect the operation status data of each dry-type transformer and store the ledger parameters and historical load data. The operation status data includes operating condition data, thermal status data, load demand data and switch status data. The group adaptation module is used to identify the actual connection group of each dry-type transformer based on the operating condition data, and through phase matching and impedance balancing, enable the parallel coordinated operation of dry-type transformers with different connection groups, and output the group compatibility constraints. The cluster partitioning module is used to partition dry-type transformers with similar characteristics into clusters based on operating condition data, ledger parameters, and thermal status data, and to perform anomaly detection and dynamic cluster updates, outputting the management group partitioning results. The load forecasting module is used to perform multi-time-scale load forecasting based on load demand data and historical load data, construct forecast load curves, and output load semantic features. The cluster load allocation collaborative optimization scheduling module is used to construct a comprehensive optimization model based on operating status data, load semantic characteristics, group compatibility constraints, and management group division results. It uses a rolling optimization mechanism to solve the model in each scheduling cycle and generate scheduling control instructions.

2. The multi-unit dry-type transformer group coordination and cluster load dispatching system according to claim 1, characterized in that: The operating condition data includes three-phase voltage, three-phase current, active power, reactive power, harmonic content, and three-phase imbalance; thermal condition data includes winding temperature; switch condition data includes circuit breaker opening and closing status; load demand data includes instantaneous values ​​of total active load and total reactive load; and ledger parameters include transformer capacity, impedance voltage, connection group, commissioning time, no-load loss, short-circuit loss, and rated capacity.

3. The multi-unit dry-type transformer group coordination and cluster load dispatching system according to claim 2, characterized in that: The group adaptation module includes: The group identification and phase matching submodule is used to identify the actual connection group of each dry-type transformer based on the phase difference data of the three-phase voltage, and adjust the output voltage of different connection groups to be consistent with the phase through the intelligent power electronic interface device configured on the low-voltage side of each dry-type transformer, so that the dry-type transformers of different connection groups can be connected in parallel on the common bus. The impedance detection and equalization submodule is used to detect the equivalent impedance of each dry-type transformer in real time based on the response data of three-phase voltage and three-phase current. When the calculated equivalent impedance deviation exceeds the impedance deviation threshold, the output characteristics of the intelligent power electronic interface device are adjusted to make the equivalent impedance of each transformer tend to be balanced, and the circulating current during parallel operation is suppressed within the preset ratio of the rated current. The constraint generation submodule is used to generate group compatibility constraints based on the phase matching completion status and impedance equalization results, including the flag phases that allow parallel connection and the impedance deviation thresholds between each transformer.

4. The multi-unit dry-type transformer group coordination and cluster load dispatching system according to claim 3, characterized in that: The process of grouping dry-type transformers with similar characteristics into clusters includes: Feature parameters are extracted from operating condition data, ledger parameters, and thermal state data and combined to form feature vectors. Correlation analysis is used to remove redundant features to obtain the optimal feature set. Feature similarity between dry-type transformers is calculated. Dry-type transformers with feature similarity higher than the preset similarity threshold are assigned to the same transformer cluster. The transformer cluster number and the list of transformer identifiers contained in each cluster are used as the management group division result.

5. The multi-unit dry-type transformer group coordination and cluster load dispatching system according to claim 4, characterized in that: The process of anomaly detection and dynamic cluster updates includes: The system monitors the operational status data of the same transformer cluster in real time, calculates the cluster average value of each parameter of the operational status data in each monitoring period, and determines that the dry-type transformer is in an abnormal state and triggers an early warning when the operational status data of a certain dry-type transformer deviates from the cluster average value by more than the set deviation warning threshold. When a dry-type transformer in a transformer cluster exits due to an abnormal state or a new dry-type transformer is put into operation, the cluster division is re-triggered to perform dynamic cluster updates.

6. The multi-unit dry-type transformer group coordination and cluster load dispatching system according to claim 5, characterized in that: The process of load forecasting by the load forecasting module for multi-timescale load forecasting includes: Historical load demand data is obtained as historical load data. Discrete wavelet transform is used to decompose the historical load data into multi-scale data to obtain multi-time granularity load data sequences. Recurrent neural networks are used to extract time-series features from the multi-time granularity load data sequences to obtain load density prediction data. Predicted load curves are constructed, and a large language model is used to perform semantic analysis on the changing trend of load density prediction data to extract load semantic features, including peak load, duration, and occurrence period.

7. A multi-unit dry-type transformer group coordination and cluster load dispatching system according to claim 6, characterized in that: When constructing the comprehensive optimization model, the optimization objectives are to minimize the total system loss, maximize the load balance, and optimize the voltage qualification rate. The optimization variables are the circuit breaker opening and closing status and the active power output distribution ratio and reactive power output distribution ratio of each dry-type transformer. The constraints are capacity constraints, voltage constraints, unbalance constraints, temperature constraints, and group compatibility constraints. The model's inputs include active power, reactive power, three-phase voltage, three-phase current, circuit breaker status, management group division results, no-load loss, short-circuit loss, rated capacity, load semantic characteristics, and group compatibility constraints; the outputs include: switching commands for each circuit breaker, active power output allocation ratio for each dry-type transformer, reactive power output allocation ratio, and on-load tap changer adjustment commands.

8. A multi-unit dry-type transformer group coordination and cluster load dispatching system according to claim 7, characterized in that: In the comprehensive optimization model: The total system loss is the sum of the no-load loss of each transformer and the square of the load rate multiplied by the short-circuit loss. The load rate is obtained by dividing the active power by the rated capacity. The load balance is calculated by summing the load rate variances according to the management group. The voltage qualification rate is the sum of the squares of the differences between the three-phase voltage and the reference value. The capacity constraint is that the load rate of each transformer does not exceed the maximum allowable load rate corresponding to its rated capacity; the voltage constraint is that the three-phase voltage amplitude on the low-voltage side of each transformer is maintained within the upper and lower limits specified by the power grid; the unbalance constraint is that the three-phase unbalance does not exceed the preset threshold; the temperature constraint is that the winding temperature does not exceed the maximum allowable temperature of its insulation class; the group compatibility constraint is that dry-type transformers operating in parallel on the same busbar must meet the requirements of phase matching and equivalent impedance deviation not exceeding the impedance deviation threshold.

Images