A reactive power optimization control system for distribution stations

Abstract

The present invention relates to the field of distribution station control technology, and in particular to a reactive power optimization control system for distribution stations. It includes a distribution station data acquisition module, a reactive power prediction module, an optimization decision module, a distributed collaborative control module, and a monitoring interaction module. The distribution station data acquisition module collects parameters such as voltage and current and ensures data integrity; the reactive power prediction module generates a prediction report based on graph neural network modeling; the optimization decision module uses a differential evolution algorithm to generate an optimal compensation plan; the distributed collaborative control module realizes collaboration between devices through fractional calculus; and the monitoring interaction module realizes equipment diagnosis, aging prediction, and dynamic feedback. The present invention realizes the high efficiency, reliability, and intelligence of reactive power optimization control of distribution stations through advanced algorithms and collaborative control technologies, thereby improving the operating efficiency and reliability of distribution stations.

Description

Technical Field

[0001] The invention relates to the technical field of power distribution station control, in particular to a reactive power optimization control system of a power distribution station. Background Art

[0002] With the continuous development of power systems, distribution stations play an important role in power systems. The main function of distribution stations is to receive electric energy from the upper power grid and convert it into a voltage level that meets user needs, while realizing the distribution and dispatch of electric energy. In modern power grids, how to improve the transmission efficiency of electric energy, reduce line losses, and ensure voltage quality has become the core goal of distribution system optimization.

[0003] Reactive power is an important parameter that cannot be ignored in the power system. It has a direct impact on the voltage stability, line loss and service life of the power grid. However, there are still the following problems: traditional reactive power control methods usually rely on fixed compensation equipment (such as capacitor banks) or simple manual adjustments, which often cannot meet the needs of modern distribution stations for dynamic and precise control; in existing technologies, reactive compensation technology has gradually been applied to distribution systems, but it often relies on a single control target; in addition, reactive compensation equipment lacks an efficient collaborative control mechanism, making it difficult to achieve real-time autonomous decision-making, affecting the system response efficiency. Summary of the invention

[0004] To solve the above problems, the present invention provides a reactive power optimization control system for a distribution station, which solves the problem of how to achieve dynamic and accurate reactive power control in the distribution station, while taking into account multi-objective optimization and efficient coordinated control of reactive compensation equipment, thereby improving the efficiency, reliability and intelligence of reactive power optimization control of the distribution station.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] A reactive power optimization control system for a distribution station, comprising a distribution station data acquisition module, a reactive power prediction module, an optimization decision module, a distributed collaborative control module and a monitoring interaction module which are sequentially connected in communication;

[0007] The distribution station data acquisition module is used to collect parameter data of each node in the distribution station; the parameter data includes voltage, current, reactive power, equipment status and environmental parameters;

[0008] The reactive power prediction module is used to model the dynamic changes of reactive power based on the parameter data, using a method combining a graph neural network and an attention mechanism, capture the complex dependencies of time series and space, and generate a reactive power prediction report;

[0009] The optimization decision module is used to generate an optimal reactive power compensation scheme under a multi-objective optimization framework using an adaptive differential evolution algorithm based on the reactive power prediction report and power grid operation constraints;

[0010] The distributed collaborative control module is used to realize autonomous decision-making and collaborative control among reactive power compensation devices through a collaborative control algorithm based on fractional-order calculus;

[0011] The monitoring interaction module is used to combine intelligent algorithms with expert system rule bases to monitor the operating status of distribution stations, diagnose potential anomalies and predict equipment aging trends, while providing dynamic feedback through an interactive interface.

[0012] Furthermore, the operation process of the reactive power prediction module includes the following steps:

[0013] Mapping the parameter data into a graph structure to construct a dynamic weight graph model;

[0014] The structural characteristics of the dynamic weight graph model are processed by using a graph convolutional network, the neighborhood node information is aggregated layer by layer, the spatial dependency relationship between nodes is extracted, and a feature vector representing the local reactive power distribution of each node is generated;

[0015] Based on the gated recurrent unit, the extracted spatial features are modeled in time series to capture the timing patterns in the dynamic changes of reactive power, and the temporal attention mechanism is used to generate global spatiotemporal feature representation;

[0016] The global spatiotemporal characteristics are input into the reactive power prediction model, and the reactive power demand changes of each node in the future period are output in combination with the historical trend and real-time characteristics, and a reactive power prediction report is generated, including the reactive power change range and confidence interval.

[0017] Furthermore, the nodes in the dynamic weight graph model represent devices or grid nodes, and the edge weights are dynamically adjusted according to the electrical distance and power flow distribution.

[0018] Furthermore, the formula of the reactive power prediction model is as follows:

[0019]

[0020] in, represents the reactive power demand at the predicted time t; represents the global output weight of node i; represents the feature transformation function of node i; represents the temporal attention weight of node i at time step k; Represents the time series feature vector of point i at time step k, including historical voltage, historical current, historical reactive power, power flow characteristics and electrical distance; Represents the dynamic weight between nodes, combining electrical distance and real-time power difference; and represents the input feature vector of nodes i and j at time step k; A nonlinear transformation that represents the interaction of spatial features between nodes; It represents the fusion function of time and space features; N represents the total number of nodes; T represents the length of the time series; Represents the norm of the weight matrix.

[0021] Furthermore, the operation process of the optimization decision module includes the following steps:

[0022] Based on the reactive power forecast report, a multi-objective optimization model is constructed. The optimization objectives include reducing network losses, improving voltage stability, minimizing reactive power compensation costs, and setting grid operation constraints, such as voltage range, equipment capacity restrictions, and safe operation constraints.

[0023] The adaptive differential evolution algorithm is used to generate the initial population, and each individual represents a reactive power compensation scheme, including equipment regulation status and power allocation strategy.

[0024] In the framework of multi-objective optimization, through mutation, crossover and selection operations, the population individuals are gradually improved; combined with fitness evaluation, the high-quality solutions that meet the optimization objectives and constraints are selected;

[0025] When the iteration termination conditions are met, the optimal reactive power compensation scheme is output, including the equipment adjustment strategy and the predicted power grid performance indicators;

[0026] Combined with the distributed collaborative control module, the generated optimal solution is dynamically adjusted.

[0027] Furthermore, the operation process of the distributed collaborative control module includes the following steps:

[0028] According to the optimal reactive power compensation scheme generated by the optimization decision module, the global optimization target is decomposed into specific regional control targets, and the regional collaborative control diagram structure is constructed based on equipment distribution and electrical topology;

[0029] Based on the regional collaborative control graph structure, combined with device status information, response time, electrical distance and dynamic load conditions, the control priority of each device is calculated and task nodes are allocated;

[0030] A collaborative control algorithm based on fractional calculus is used to dynamically adjust the collaborative weights between nodes according to the actual operating status and power flow distribution of the equipment, generating the optimal interaction mode between reactive compensation devices.

[0031] Based on adaptive control rules, each device independently optimizes its own output strategy and updates collaborative goals based on neighborhood information. At the same time, the device execution results are monitored in real time and verified to see whether they meet global constraints.

[0032] Furthermore, the formula of the cooperative control algorithm is as follows:

[0033]

[0034] in, It represents the reactive power output control quantity of device i at time t; represents the adaptive adjustment weight of node i; represents the collaborative interaction weight of node i; represents the weight of global constraint feedback; represents the normalized offset adjustment coefficient; represents the adaptive control rule function of node i; Represents the set of neighboring nodes of node i; represents the dynamic cooperation weight between nodes i and j, which is determined by the electrical distance and power flow distribution; represents the optimization of the spatial distribution of reactive power; represents the characteristic response function of node k; represents the norm of the state vector of node i; represents the dynamic cooperation weight between nodes i and k.

[0035] Furthermore, the distribution station data acquisition module also integrates a multi-node data storage and sharing mechanism based on blockchain technology, while supporting collaborative data access for devices at different levels.

[0036] Furthermore, the monitoring interaction module adopts a fault adaptive diagnosis algorithm based on dynamic transfer learning to achieve accurate health status assessment and maintenance plan generation of distribution station equipment under different operating environments.

[0037] The beneficial effects of the present invention are:

[0038] The data acquisition module of the distribution station of the present invention can obtain multi-dimensional data including voltage, current, reactive power, equipment status and environmental parameters, provide comprehensive and accurate operating status information for the system, and improve the basic data quality of optimization control. By combining the graph neural network and the attention mechanism, the reactive power prediction module can capture the complex time series and spatial dependency of the dynamic change of reactive power, effectively improve the accuracy and reliability of the prediction, and provide a scientific basis for subsequent decision-making. The optimization decision module using the adaptive differential evolution algorithm comprehensively considers the power grid operation constraints under the multi-objective optimization framework, and can quickly generate the optimal reactive power compensation scheme, taking into account both economy and power grid stability. The distributed collaborative control module is based on the collaborative control algorithm of fractional calculus, which enables the reactive compensation equipment to have autonomous decision-making capabilities, and realizes collaborative control between devices, thereby improving the response speed and overall coordination performance of the system. The monitoring interaction module combines the intelligent algorithm and the expert system rule base to realize the real-time monitoring of the operating status of the distribution station, the rapid diagnosis of potential anomalies, and the accurate prediction of the aging trend of equipment, thereby improving the safety and maintenance efficiency of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a module schematic diagram of a reactive power optimization control system of a distribution station according to the present invention.

[0040] Figure 2 It is a flowchart of the operation process of the reactive power prediction module provided by an embodiment of the present invention.

[0041] Figure 3 It is a flow chart of the operation process of the distributed collaborative control module provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0042] See also Figure 1-3 As shown, the present invention relates to a reactive power optimization control system for a distribution station.

[0043] Example

[0044] A reactive power optimization control system for a distribution station, comprising a distribution station data acquisition module, a reactive power prediction module, an optimization decision module, a distributed collaborative control module and a monitoring interaction module which are sequentially connected in communication;

[0045] The distribution station data acquisition module is used to collect parameter data of each node in the distribution station; the parameter data includes voltage, current, reactive power, equipment status and environmental parameters; the distribution station data acquisition module also integrates a multi-node data storage and sharing mechanism based on blockchain technology, and supports collaborative data access of devices at different levels.

[0046] It should be noted that the system uses high-precision voltage and current sensors to ensure the accuracy of collected data, and supports measurement ranges from low voltage to high voltage (such as 10kV, 35kV); it is equipped with environmental sensors (such as temperature and humidity, air pressure, and dust sensors) to adapt to different distribution station scenarios. Built-in industrial-grade processors and data interfaces support the collection of multiple signals (such as Modbus, IEC61850 protocols); integrated edge computing modules can pre-process the collected data locally (such as filtering and abnormal elimination). Supports multiple communication methods, including fiber optic communication, 5G, and LoRa wireless communication, to meet the needs of different scenarios; equipped with anti-interference shielding function to ensure stable signal transmission. Adopt an independent power supply system, support both AC and DC power supply modes; equipped with surge protection devices and backup power supplies to ensure continuous operation of the system in emergency situations.

[0047] The collected data is stored in multiple nodes through blockchain, and each node has a complete copy of the data to ensure the integrity and reliability of the data; smart contracts are used to control data access rights, such as different levels of equipment or user access rights (for example, senior administrators can access all data, while ordinary users can only view part of the data). AES and RSA dual encryption technology is used to protect data privacy during data transmission and storage; it is equipped with anti-tampering function, and once the data is illegally modified, an alarm will be triggered.

[0048] The reactive power prediction module is used to model the dynamic changes of reactive power based on the parameter data, using a method combining a graph neural network and an attention mechanism, capture the complex dependencies of time series and space, and generate a reactive power prediction report;

[0049] The operation process of the reactive power prediction module includes the following steps:

[0050] Mapping the parameter data into a graph structure to construct a dynamic weight graph model; the nodes in the dynamic weight graph model represent devices or grid nodes, and the edge weights are dynamically adjusted according to electrical distance and power flow distribution;

[0051] Specifically, according to the distribution station grid topology, each grid node (such as transformers, switches, and reactive compensation equipment) is defined as a node in the graph structure. Node attributes include parameters such as voltage, current, and reactive power, as well as device type and operating status. The electrical distance between grid nodes is calculated using the impedance formula (such as using the reactance and admittance between nodes). Combined with the flow calculation results, the edge weights are dynamically adjusted according to the strength of power transmission, and larger power flows correspond to higher weights. The edge weights are dynamically updated to reflect the real-time interaction characteristics between nodes. Combining the characteristics of the above nodes and edges, a graph model with dynamic weights is generated to reflect the operating status of the distribution station in real time.

[0052] The structural characteristics of the dynamic weight graph model are processed by using a graph convolutional network, the neighborhood node information is aggregated layer by layer, the spatial dependency relationship between nodes is extracted, and a feature vector representing the local reactive power distribution of each node is generated;

[0053] Specifically, initial features are assigned to each node, including its real-time electrical parameters (such as reactive power, voltage), device status, and environmental parameters. The graph convolutional network gradually aggregates the neighbor information of each node to the current node through layer-by-layer processing. In this process, the spatial dependencies between nodes are captured, such as how the reactive power change of a node is affected by the adjacent nodes. After multiple layers of graph convolution, each node generates a feature vector that represents its reactive power distribution characteristics in the local space. These feature vectors not only contain the attributes of the node itself, but also incorporate the information of its neighboring nodes.

[0054] Based on the gated recurrent unit, the extracted spatial features are modeled in time series to capture the timing patterns in the dynamic changes of reactive power, and the temporal attention mechanism is used to generate global spatiotemporal feature representation;

[0055] Specifically, the extracted node spatial features are organized into time series in chronological order to reflect the dynamic changes of reactive power. The gated recurrent unit (GRU) model is used to process the time series to explore the dynamic change patterns and trends of reactive power. GRU can effectively handle long-term dependencies, such as identifying day-and-night periodic changes or sudden fluctuations in reactive power. Based on time series modeling, the temporal attention mechanism is applied to dynamically evaluate the importance of each time point. The features of key time points are given higher weights, so as to generate global spatiotemporal features more accurately.

[0056] The global spatiotemporal characteristics are input into the reactive power prediction model, and the reactive power demand changes of each node in the future period are output in combination with the historical trend and real-time characteristics, and a reactive power prediction report is generated, including the reactive power change range and confidence interval.

[0057] Specifically, spatial features and temporal features are combined to generate a global spatiotemporal feature representation of each node, including its local relationships, historical trends, and dynamic changes. A multi-layer neural network is used to process the global spatiotemporal features to generate reactive power prediction results for each node in the future period. The prediction results include the specific numerical changes of reactive power, possible fluctuation ranges, and operating deviations. The reactive power demand forecast value for each node in the future period is output. At the same time, the reactive power change range and confidence interval are generated to provide a reference for subsequent optimization.

[0058] Furthermore, the formula of the reactive power prediction model is as follows:

[0059]

[0060] in, represents the reactive power demand at the predicted time t; represents the global output weight of node i; represents the feature transformation function of node i; represents the temporal attention weight of node i at time step k; Represents the time series feature vector of point i at time step k, including historical voltage, historical current, historical reactive power, power flow characteristics and electrical distance; Represents the dynamic weight between nodes, combining electrical distance and real-time power difference; and represents the input feature vector of nodes i and j at time step k; A nonlinear transformation that represents the interaction of spatial features between nodes; It represents the fusion function of time and space features; N represents the total number of nodes; T represents the length of the time series; Represents the norm of the weight matrix.

[0061] The optimization decision module is used to generate an optimal reactive power compensation scheme under a multi-objective optimization framework using an adaptive differential evolution algorithm based on the reactive power prediction report and power grid operation constraints;

[0062] The operation process of the optimization decision module includes the following steps:

[0063] Based on the reactive power forecast report, a multi-objective optimization model is constructed. The optimization objectives include reducing network losses, improving voltage stability, minimizing reactive power compensation costs, and setting grid operation constraints, such as voltage range, equipment capacity restrictions, and safe operation constraints.

[0064] Specifically, the reactive power forecast report is input, including the reactive power demand change, change range and confidence interval of each node in the future period. Important forecast information, such as high-fluctuation nodes and weak grid areas, is extracted to provide initial conditions for the optimization model.

[0065] Define the optimization goals as follows:

[0066] Reduce network losses: The goal is to minimize active power losses in the grid and ensure efficient use of energy.

[0067] Improve voltage stability: Through reactive power compensation, the voltage at the grid nodes is kept within the specified range to avoid voltage fluctuations or collapse.

[0068] Minimize reactive power compensation costs: Comprehensively consider the start-up and shutdown costs and power regulation costs of compensation equipment (such as capacitors and transformers) to optimize economic efficiency.

[0069] Set the running constraints as follows:

[0070] Voltage range: Ensure that the voltage of each node is within the specified safe operating range (such as 0.95-1.05 times the rated value).

[0071] Equipment capacity limitation: The maximum adjustment range of the compensation equipment is limited according to its rated capacity and operating status.

[0072] Safe operation constraints: Ensure that network power flows meet thermal stability and short-circuit protection requirements.

[0073] Combining the above objectives and constraints, a multi-objective optimization model is formed to guide subsequent optimization calculations.

[0074] The adaptive differential evolution algorithm is used to generate the initial population, and each individual represents a reactive power compensation scheme, including equipment regulation status and power allocation strategy.

[0075] In the framework of multi-objective optimization, through mutation, crossover and selection operations, the population individuals are gradually improved; combined with fitness evaluation, the high-quality solutions that meet the optimization objectives and constraints are selected;

[0076] Specifically, mutation operations are performed on individuals in the population to generate new candidate solutions: some values ​​are randomly selected in the individual's encoding vector for adjustment, such as changing the device power allocation value or adjusting the start / stop status; the mutation amplitude is dynamically adjusted according to the iterative process to ensure a balance between exploration ability and convergence speed.

[0077] Perform a crossover operation on individuals in the population, exchange part of the encoded information of two individuals, and generate a new compensation scheme: for example, transfer the device state from one individual to another while retaining part of the power allocation strategy. The crossover probability is dynamically adjusted to increase the diversity of solutions.

[0078] Select the next generation of individuals based on fitness values: Adopt a selection strategy based on fitness sorting, and give priority to individuals that meet the constraints and have better optimization objectives. At the same time, retain some poor solutions to maintain population diversity and avoid falling into local optimality. Recalculate the fitness values ​​of the newly generated population individuals. According to the multi-objective optimization framework, evaluate the performance of individuals on multiple objectives, and use the Pareto frontier to screen high-quality solutions.

[0079] When the iteration termination conditions are met, the optimal reactive power compensation scheme is output, including the equipment adjustment strategy and the predicted power grid performance indicators;

[0080] Specifically, determine whether the optimization process meets the termination conditions, such as reaching the maximum number of iterations or the objective function converging to the preset threshold. Output the optimal reactive power compensation scheme, including: the adjustment status of each compensation device (start / stop status and target compensation power); predicted grid performance indicators (such as network loss, voltage deviation and compensation cost); present the scheme in graphical and tabular form for easy user understanding and verification. Display the trade-off results of multi-objective optimization, such as the relationship curve between loss reduction and compensation cost. Provide a list of optional schemes, and users can choose schemes with different optimization priorities according to actual needs.

[0081] Combined with the distributed collaborative control module, the generated optimal solution is dynamically adjusted.

[0082] The distributed collaborative control module is used to realize autonomous decision-making and collaborative control among reactive power compensation devices through a collaborative control algorithm based on fractional-order calculus;

[0083] The operation process of the distributed collaborative control module includes the following steps:

[0084] According to the optimal reactive power compensation scheme generated by the optimization decision module, the global optimization target is decomposed into specific regional control targets, and the regional collaborative control diagram structure is constructed based on equipment distribution and electrical topology;

[0085] Specifically, the optimal reactive power compensation scheme generated by the receiving optimization decision module includes the target reactive power distribution of the entire network, the start and stop status of the equipment, and the power allocation strategy. According to the electrical topology of the distribution network, the power grid is divided into several regions, each of which contains a group of closely connected devices (such as transformers, capacitors, and switches). The regional division is based on electrical distance, power flow distribution, and equipment location to ensure that the equipment in each region has high synergy. The global optimization goal is decomposed into local goals for each region, such as voltage stability, reactive power balance, and loss minimization within the region. When decomposing the goals, the boundary conditions between regions are considered to ensure that the realization of regional goals maximizes the contribution of the global goal. The devices and nodes in each region are represented as a graph structure, where the nodes represent the devices and the edges represent the collaborative relationship between the devices. According to the distribution of the equipment, the electrical connection relationship, and the power flow distribution, the edge weights are dynamically assigned to construct a regional collaborative control graph.

[0086] Based on the regional collaborative control graph structure, combined with device status information, response time, electrical distance and dynamic load conditions, the control priority of each device is calculated and task nodes are allocated;

[0087] Specifically, the device operating status (such as current power output, start / stop status, health score) and environmental conditions (such as load level, temperature and humidity) are obtained in real time. The control priority of each device is calculated based on the following factors:

[0088] Device status information: Normally operating devices have a higher priority, while aging or faulty devices have a lower priority.

[0089] Response time: Devices with fast response times are given higher priority to quickly adjust the system.

[0090] Electrical distance: Equipment close to key nodes has higher priority, such as reactive power compensation equipment close to the main transformer.

[0091] Dynamic load conditions: In areas with large load fluctuations, equipment with higher flexibility is preferred.

[0092] A ranking algorithm is used to assign a control priority list to each device.

[0093] Assign control tasks to devices based on priority, such as reactive power adjustment range, voltage regulation target, etc. Ensure that high-priority devices take on more critical tasks while maintaining an even distribution of overall tasks.

[0094] A collaborative control algorithm based on fractional calculus is used to dynamically adjust the collaborative weights between nodes according to the actual operating status and power flow distribution of the equipment, generating the optimal interaction mode between reactive compensation devices.

[0095] Specifically, in the regional collaborative control diagram, the collaboration weights between nodes are adjusted in real time based on the equipment status and flow distribution.

[0096] The following factors are considered in weight adjustment:

[0097] Operation status: Healthy devices receive higher collaboration weights;

[0098] Electrical distance: The weight of equipment close to the load center or key node increases;

[0099] Power flow distribution: Edges with larger power flows receive higher weights to ensure compensation efficiency.

[0100] The fractional-order calculus method is used to describe the dynamic cooperation relationship between devices and establish a balance between global and local control objectives. The fractional-order model has higher dynamic response accuracy and can capture the nonlinear characteristics of the device operation state. Based on the fractional-order control model, the cooperation strategy between devices is calculated, such as the sharing ratio of power compensation and the priority of device startup. The optimal interaction mode between devices in the area is output to ensure the realization of regional goals while minimizing interference with adjacent areas.

[0101] Based on adaptive control rules, each device independently optimizes its own output strategy and updates collaborative goals based on neighborhood information. At the same time, the device execution results are monitored in real time and verified to see whether they meet global constraints.

[0102] Specifically, each device independently optimizes its own reactive power output strategy based on the assigned tasks and real-time operating status. The optimization process takes into account the physical constraints of the device (such as capacity limitations), operating costs, and safety requirements. Each device receives status information of neighboring devices (such as output power, device health) in real time and dynamically adjusts the collaborative goals.

[0103] For example, when a neighboring device stops operating due to overload, other devices automatically adjust their output strategies to make up for the shortfall. Adaptive control rules are used to dynamically adjust control parameters (such as response thresholds and compensation ranges) based on the grid operation status. This ensures that the control rules can flexibly adapt to load fluctuations and topology changes in the grid.

[0104] Furthermore, the formula of the cooperative control algorithm is as follows:

[0105]

[0106] in, It represents the reactive power output control quantity of device i at time t; represents the adaptive adjustment weight of node i, reflecting the impact of the equipment operation status on reactive power compensation; represents the cooperative interaction weight of node i, which affects the collaborative optimization of reactive power distribution among neighboring nodes; Represents the weight of global constraint feedback to ensure reactive power adjustment meets grid operation conditions; represents the normalized offset adjustment coefficient; represents the adaptive control rule function of node i; Represents the set of neighboring nodes of node i; represents the dynamic cooperation weight between nodes i and j, which is determined by the electrical distance and power flow distribution; represents the optimization of the spatial distribution of reactive power; represents the characteristic response function of node k; represents the norm of the state vector of node i; represents the dynamic cooperation weight between nodes i and k.

[0107] The monitoring interaction module is used to combine intelligent algorithms with expert system rule bases to monitor the operating status of distribution stations, diagnose potential anomalies and predict equipment aging trends, and provide dynamic feedback through an interactive interface; the monitoring interaction module adopts a fault adaptive diagnosis algorithm based on dynamic transfer learning to achieve accurate health status assessment and maintenance plan generation of distribution station equipment under different operating environments.

[0108] Specifically, the fault adaptive diagnosis model based on transfer learning combines historical fault data and real-time collected data to quickly generate diagnosis models for different environments; automatically update fault model parameters to adapt to the dynamic changes in the operating environment of the distribution station. The integrated expert system rule base contains typical fault modes and corresponding solutions (such as cable insulation aging, poor contact of switchgear, etc.); through the knowledge reasoning engine, the cause of the abnormality is analyzed in real time to generate actionable maintenance suggestions. A multi-level alarm mechanism is implemented, which is divided into three levels: prompt, warning and emergency according to the severity of the fault; in the event of a serious fault, an audible and visual alarm is triggered and the on-duty personnel are notified by SMS, email, etc.

[0109] Combined with the real-time operation data and historical maintenance records of the equipment, a machine learning algorithm is used to generate the equipment health score; the health score range (such as 0-100 points) and color distinction (such as green, yellow, red) intuitively reflect the health status of the equipment. Based on time series analysis models (such as ARIMA or LSTM), the aging trend and remaining service life of the equipment are predicted; trend charts and specific aging rate indicators are provided. Based on the health status score and trend analysis, potential failure risks are identified in advance; and targeted maintenance plans are generated for key equipment (such as transformers and switches).

[0110] Generate customized maintenance plans based on equipment health scores and fault warnings, including maintenance priorities and schedules; automatically list required spare parts and maintenance tools, and push them to the warehouse management system. Dynamically assign maintenance tasks to corresponding staff based on maintenance plans and on-site workloads; provide work order management functions to track the progress of maintenance tasks. Automatically record the details of each maintenance operation (such as replacement parts, fault handling time), and update the equipment's operation files; support historical record query to provide a basis for subsequent equipment management.

[0111] In summary, the present invention adopts high-precision voltage and current sensors, supports full-range measurement from low voltage to high voltage, and integrates environmental sensors (such as temperature, humidity, air pressure, etc.) to adapt to various scenario requirements; local data preprocessing (such as anomaly elimination and filtering) is realized through the edge computing module, which greatly improves the accuracy and real-time performance of the data; blockchain technology is combined to realize distributed data storage and sharing, and ensure the integrity and immutability of the data.

[0112] The method of combining graph neural network and attention mechanism in the present invention accurately models the dynamic changes of reactive power. By constructing a dynamic weight graph model, the complex spatiotemporal dependencies between power grid nodes are captured, and the operating status of the distribution station is reflected in real time; combined with the time attention mechanism, global spatiotemporal features are generated to improve the accuracy and dynamic adaptability of reactive power prediction, providing a reliable prediction basis for optimization control. Through the adaptive differential evolution algorithm, the optimal reactive power compensation scheme is generated under the multi-objective optimization framework. The optimization objectives comprehensively cover key requirements such as reducing network losses, improving voltage stability, and minimizing compensation costs, and comprehensively consider the power grid operation constraints (such as voltage range, equipment capacity limitations) to ensure that the compensation scheme is economical and efficient; the output multi-objective optimization trade-off results provide users with a variety of choices to meet the needs of different operating scenarios.

[0113] The present invention adopts a collaborative control algorithm based on fractional calculus to achieve autonomous decision-making and dynamic collaboration between devices in the region; comprehensively calculates control priorities through device status information, electrical distance and load conditions to ensure efficient use of resources; dynamically adjusts the collaboration weights between nodes, optimizes the interaction mode between devices, ensures the achievement of regional goals while reducing the impact on other regions, and improves the overall operating efficiency of the system. The fault adaptive diagnosis algorithm using dynamic transfer learning can quickly adapt to changes in the operating environment of the distribution station and achieve accurate health status assessment and fault diagnosis of equipment; combined with the expert system rule base, it can quickly analyze potential anomalies and generate maintenance suggestions to enhance the intelligent maintenance capabilities of the system; intuitively presents the operating status of the equipment through a multi-level alarm mechanism and health score, and provides aging trend prediction and remaining life assessment, providing a scientific basis for equipment management and maintenance plans.

[0114] The above implementation modes are merely descriptions of the preferred implementation modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary engineering and technical personnel in the field shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A reactive power optimization control system for a distribution station, characterized in that: It includes a distribution station data acquisition module, a reactive power prediction module, an optimization decision module, a distributed collaborative control module and a monitoring interaction module which are sequentially connected in communication; The distribution station data acquisition module is used to collect parameter data of each node in the distribution station; the parameter data includes voltage, current, reactive power, equipment status and environmental parameters; The reactive power prediction module is used to model the dynamic changes of reactive power based on the parameter data, using a method combining a graph neural network and an attention mechanism, capture the complex dependencies of time series and space, and generate a reactive power prediction report; The optimization decision module is used to generate an optimal reactive power compensation scheme under a multi-objective optimization framework using an adaptive differential evolution algorithm based on the reactive power prediction report and power grid operation constraints; The distributed collaborative control module is used to realize autonomous decision-making and collaborative control among reactive power compensation devices through a collaborative control algorithm based on fractional-order calculus; The monitoring interaction module is used to combine intelligent algorithms with expert system rule bases to monitor the operating status of distribution stations, diagnose potential anomalies and predict equipment aging trends, while providing dynamic feedback through an interactive interface; The operation process of the reactive power prediction module includes the following steps: Mapping the parameter data into a graph structure to construct a dynamic weight graph model; The structural characteristics of the dynamic weight graph model are processed by using a graph convolutional network, the neighborhood node information is aggregated layer by layer, the spatial dependency relationship between nodes is extracted, and a feature vector representing the local reactive power distribution of each node is generated; Based on the gated recurrent unit, the extracted spatial features are modeled in time series to capture the timing patterns in the dynamic changes of reactive power, and the temporal attention mechanism is used to generate global spatiotemporal feature representation; The global spatiotemporal characteristics are input into the reactive power prediction model, and the reactive power demand changes of each node in the future period are output in combination with the historical trend and real-time characteristics, and a reactive power prediction report is generated, including the reactive power change range and confidence interval.

2. A reactive power optimization control system for a distribution station according to claim 1, characterized in that: The nodes in the dynamic weight graph model represent devices or grid nodes, and the edge weights are dynamically adjusted according to the electrical distance and power flow distribution.

3. The reactive power optimization control system of a distribution station according to claim 1, characterized in that: The formula of the reactive power prediction model is as follows: ； in, represents the reactive power demand at the predicted time t; represents the global output weight of node i; represents the feature transformation function of node i; represents the temporal attention weight of node i at time step k; Represents the time series feature vector of point i at time step k, including historical voltage, historical current, historical reactive power, power flow characteristics and electrical distance; Represents the dynamic weight between nodes, combining electrical distance and real-time power difference; and represents the input feature vector of nodes i and j at time step k; A nonlinear transformation that represents the interaction of spatial features between nodes; It represents the fusion function of time and space features; N represents the total number of nodes; T represents the length of the time series; Represents the norm of the weight matrix.

4. The reactive power optimization control system of a distribution station according to claim 1, characterized in that: The operation process of the optimization decision module includes the following steps: Based on the reactive power forecast report, a multi-objective optimization model is constructed. The optimization objectives include reducing network losses, improving voltage stability, minimizing reactive power compensation costs, and setting grid operation constraints, including voltage range, equipment capacity restrictions, and safe operation constraints. The adaptive differential evolution algorithm is used to generate the initial population, and each individual represents a reactive power compensation scheme, including equipment regulation status and power allocation strategy. In the framework of multi-objective optimization, through mutation, crossover and selection operations, the population individuals are gradually improved; combined with fitness evaluation, the high-quality solutions that meet the optimization objectives and constraints are selected; When the iteration termination conditions are met, the optimal reactive power compensation scheme is output, including the equipment adjustment strategy and the predicted power grid performance indicators; Combined with the distributed collaborative control module, the generated optimal solution is dynamically adjusted.

5. The reactive power optimization control system of a distribution station according to claim 1, characterized in that: The operation process of the distributed collaborative control module includes the following steps: According to the optimal reactive power compensation scheme generated by the optimization decision module, the global optimization target is decomposed into specific regional control targets, and the regional collaborative control diagram structure is constructed based on equipment distribution and electrical topology; Based on the regional collaborative control graph structure, combined with device status information, response time, electrical distance and dynamic load conditions, the control priority of each device is calculated and task nodes are allocated; A collaborative control algorithm based on fractional calculus is used to dynamically adjust the collaborative weights between nodes according to the actual operating status and power flow distribution of the equipment, generating the optimal interaction mode between reactive compensation devices. Based on adaptive control rules, each device independently optimizes its own output strategy and updates collaborative goals based on neighborhood information. At the same time, the device execution results are monitored in real time and verified to see whether they meet global constraints.

6. A reactive power optimization control system for a power distribution station according to claim 5, characterized in that: The formula of the cooperative control algorithm is as follows: ； in, It represents the reactive power output control quantity of device i at time t; represents the adaptive adjustment weight of node i; represents the collaborative interaction weight of node i; represents the weight of global constraint feedback; represents the normalized offset adjustment coefficient; represents the adaptive control rule function of node i; Represents the set of neighboring nodes of node i; represents the dynamic cooperation weight between nodes i and j, which is determined by the electrical distance and power flow distribution; represents the optimization of the spatial distribution of reactive power; represents the characteristic response function of node k; represents the norm of the state vector of node i; represents the dynamic cooperation weight between nodes i and k.

7. The reactive power optimization control system of a power distribution station according to claim 1, characterized in that: The distribution station data acquisition module also integrates a multi-node data storage and sharing mechanism based on blockchain technology, and supports collaborative data access for devices at different levels.

8. The reactive power optimization control system of a distribution station according to claim 1, characterized in that: The monitoring interaction module adopts a fault adaptive diagnosis algorithm based on dynamic transfer learning to achieve accurate health status assessment and maintenance plan generation of distribution station equipment under different operating environments.

Images