Method and system for cooperative control of electrical equipment based on distributed control system

Abstract

The application relates to the technical field of industrial control network communication, and discloses an electrical equipment cooperative control method and system based on a distributed control system. The method comprises the following steps: constructing a time limit topology drift attribute graph based on equipment action time limits, linkage dependencies and network states; forming an action window tensor according to target equipment, action stages, allowable time limits and protection priorities, and aggregating the action window tensor into a scheduling batch; jointly optimizing the scheduling batch by using the attribute graph, determining the mapping relationship between a sending window and a forwarding path to generate a final control routing table; checking instructions and calculating time delay compensation according to the action time limits and the routing table, and then issuing; updating network attributes and dependency relationships according to equipment execution feedback, and then evaluating sequence violation risks and generating a cooperative control report. The application effectively improves the instruction execution real-time performance and cooperative reliability of the distributed control system.

Description

Technical Field

[0001] This application relates to the field of industrial control network communication technology, and in particular to a method and system for collaborative control of electrical equipment based on a distributed control system. Background Technology

[0002] As power systems become increasingly automated, electrical equipment such as circuit breakers, disconnectors, and capacitor banks in substations are commonly deployed in a distributed manner, accessing the control system via heterogeneous communication networks such as industrial Ethernet and fieldbus. While this deployment enhances system flexibility and scalability, the different electrical devices are located in different physical locations. Transmitting control commands across network segments requires passing through multiple switches and protocol conversion nodes. Furthermore, the discrepancies in the local clocks of each node make it difficult to ensure that the arrival times of the same batch of control commands at different target devices are consistent. Random jitter in transmission delays caused by network load fluctuations can also lead to misalignments in the originally designed order of command arrival during actual execution.

[0003] When there are inter-device operational requirements, the aforementioned timing misalignment can lead to serious safety risks. Taking busbar switching operations as an example, the circuit breaker must be disconnected before the isolating switch is opened; the entire process must strictly adhere to the sequence specified in electrical safety regulations. If network jitter causes the isolating switch opening command to arrive and be executed before the circuit breaker opening command, a dangerous situation of opening the isolating switch under load will occur, causing equipment damage or even personal injury. Similarly, in scenarios such as capacitor bank switching and on-load tap changing of transformers, each device has a clearly defined execution dead zone and dependencies. The response time from receiving the command to completing the actual action can range from tens to hundreds of milliseconds, while network jitter can also reach tens of milliseconds. The combination of these two factors can easily exceed the safe timing window, endangering the stable operation of the power grid.

[0004] Existing control methods typically rely on preset fixed delay compensation values ​​or simple serial command issuance strategies to ensure timing, lacking the ability to dynamically and comprehensively perceive network conditions, device execution characteristics, and inter-device dependencies. When network load changes, link quality fluctuates, or device characteristics drift, static compensation parameters cannot accurately offset the uncertainty of transmission delay, making it difficult to ensure both timing safety and control efficiency. Therefore, providing a collaborative control method for electrical equipment based on a distributed control system to uniformly manage device action time limits, inter-device dependencies, network jitter, and clock skew in heterogeneous communication networks, dynamically adjust command sending timing and transmission paths, and continuously correct the scheduling model through execution feedback to achieve a balance between strict timing coordination and efficient control is an urgent problem to be solved. Summary of the Invention

[0005] This application provides a method and system for coordinated control of electrical equipment based on a distributed control system, which is used to improve the real-time performance of instruction execution and the reliability of coordination in a distributed control system.

[0006] In a first aspect, this application provides a method for coordinated control of electrical equipment based on a distributed control system, comprising: S1. Obtain the device action time limit table, linkage dependency table and network status information in the target control network; S2. Map the device action time limit table and linkage dependency table to the node features and logical dependency edge features of the graph structure, respectively. Based on the network state information, map them to the physical communication edge features of the graph structure. Combine the node features, logical dependency edge features and physical communication edge features to generate a time limit topology drift attribute graph. S3. Obtain the target electrical equipment and action stage corresponding to the control command to be executed, combine the maximum allowable arrival time limit and protection priority to construct a four-dimensional action window tensor, and aggregate it according to the preset scheduling cycle to form a candidate scheduling batch. S4. Based on the time-limited topology drift attribute graph, perform joint optimization on the candidate scheduling batch, determine the mapping relationship between the candidate sending window and the cross-network segment forwarding path, and generate the final control routing table; S5. Verify control commands based on the equipment action time limit table and the final control routing table, calculate delay compensation parameters, write and send control commands to the target electrical equipment; S6. Obtain the execution confirmation information returned by the target electrical equipment, and update the time limit topology drift attribute graph and linkage dependency table; S7. Based on the updated time-limited topology drift attribute graph, assess the risk of sequential violation, determine and issue subsequent control instructions, and generate a collaborative control assessment report.

[0007] Secondly, this application provides a collaborative control system for electrical equipment based on a distributed control system, comprising: The data acquisition module is used to obtain the device action time limit table, linkage dependency table and network status information in the target control network; The topology modeling module is used to map the device action time limit table and linkage dependency table into the node features and logical dependency edge features of the graph structure, respectively, and to map them into the physical communication edge features of the graph structure based on network state information. The module combines the node features, logical dependency edge features and physical communication edge features to generate a time limit topology drift attribute graph. The window batch construction module is used to obtain the target electrical equipment and action stage corresponding to the control command to be executed, combine the maximum allowable arrival time limit and protection priority to construct a four-dimensional action window tensor, and aggregate it according to the preset scheduling cycle to form a candidate scheduling batch. The joint routing module is used to jointly optimize candidate scheduling batches based on the time-limited topology drift attribute graph, determine the mapping relationship between candidate sending windows and cross-network segment forwarding paths, and generate the final control routing table. The verification and distribution module is used to verify control commands based on the device action time limit table and the final control routing table, calculate delay compensation parameters, write and distribute control commands to the target electrical equipment; The status update module is used to obtain the execution confirmation information returned by the target electrical equipment and update the time limit topology drift attribute graph and linkage dependency table; The assessment and follow-up control module is used to assess the risk of sequential violations based on the updated time-limited topology drift attribute graph, determine and issue subsequent control commands, and generate a collaborative control assessment report.

[0008] Compared with the prior art, the beneficial effects of this application are at least as follows: 1. By constructing a time-limited topology drift attribute graph, the device action time limit, linkage dependency, network jitter, and clock deviation are integrated into a graph structure model, which solves the problem of the separation of information between the device side and the network side in existing methods, and provides a complete perception basis for subsequent scheduling decisions.

[0009] 2. By encoding control commands into multi-dimensional action window tensors and jointly solving the sending window and forwarding path on the graph, the coordinated optimization of timing constraints and path selection is achieved, avoiding misalignment of linkage sequence caused by unreasonable sending timing or path.

[0010] 3. By collecting link congestion and node buffer status in real time and dynamically reallocating path weights, combined with writing the compensation amount for execution dead zone calculation into the command issuance, network latency jitter is actively compensated, ensuring that commands arrive on time and the linkage sequence is correct.

[0011] 4. By continuously correcting the graph model and dependency table parameters through the execution confirmation information returned by the receiving device, a closed-loop feedback mechanism is formed, enabling the scheduling strategy to dynamically adapt to network changes and device characteristic drift. Based on the assessment of the risk of sequence violation, the distribution mode can be flexibly switched, balancing the safety and efficiency of control. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart of the electrical equipment cooperative control method based on a distributed control system according to this application; Figure 2The iterative convergence curve of the multi-objective joint optimization objective function in this application is shown. Figure 3 This is a comparison chart of the deadline default rates under different network load conditions for this application; Figure 4 This is a schematic diagram of the structure of the electrical equipment collaborative control system based on a distributed control system according to this application. Detailed Implementation

[0014] This application provides a method and system for coordinated control of electrical equipment based on a distributed control system. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0015] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the electrical equipment cooperative control method based on a distributed control system in this application includes: Step S1: Obtain the device action time limit table, linkage dependency table and network status information in the target control network.

[0016] In one specific embodiment, the process of performing step S1 may specifically include the following steps: Obtain the target control network, which includes the first communication network and the second communication network. Extract the equipment action time limit table, which contains the control cycle, maximum allowable arrival time limit, execution dead zone and protection priority of each electrical device, and the linkage dependency table, which contains the linkage pre- and post-linkage dependencies, from the target control network. Store the equipment action time limit table and the linkage dependency table in the hybrid database. The hybrid database adopts a storage mode that combines time-series database and relational database. The switching path of the first communication network is collected using the link layer discovery protocol, and the bus segment of the second communication network is collected through the bus master station topology scanning function. The hop count of the protocol conversion node in the target control network, the historical jitter percentile of each communication link, and the local clock deviation of each node are collected synchronously. The switching path, bus segment, hop count of protocol conversion nodes, historical jitter percentile value, and local clock deviation are integrated into the network status information of the target control network.

[0017] Specifically, the control cycle, maximum permissible arrival time, execution dead zone, and protection priority of each electrical device are extracted from the target control network, and these parameters are written into the device action time limit table. The control cycle represents the minimum time interval for a device to receive a control command; this parameter is set to 4ms to 20ms for protection devices and 100ms to 500ms for remote control devices. The maximum permissible arrival time represents the maximum allowed time for a control command to travel from the master station to the target device; this parameter is set to 50ms for switch opening and closing commands. The execution dead zone represents the minimum time window that a device must experience from receiving a command to the actual action of the mechanical actuator; the execution dead zone for vacuum circuit breaker opening is set to 40ms. Protection priorities are divided into four levels: emergency protection, fast control, routine regulation, and background maintenance. The linkage dependency table records the action constraints between devices; each row records the dependency type, dependency direction, minimum interval time, and maximum interval time between a pair of devices. For example, the pre-dependency of a line protection device's closing action is that the bus is in the closed position and there is no differential blocking signal; the post-dependency is that the reclosing is completed within a 400ms delay trigger. The equipment action time limit table and the linkage dependency table are stored in a hybrid database, which uses a storage mode combining time-series and relational databases. The time-series database stores continuous measurement data that evolves over time, including control cycles, execution dead zones, and historical jitter quantiles. The data is indexed by timestamps and compressed in time-series format, with a sampling resolution set to 1ms, aligned with the shortest control cycle of process layer messages. The data retention strategy is configured to use a sliding window covering the most recent 168 hours, covering the complete equipment maintenance cycle and network load fluctuation cycle. The relational database stores topology and dependency data with association constraints, using a primary-foreign key association mechanism to ensure consistency between equipment identifiers and dependencies. The two databases establish a mapping relationship through unique equipment identifiers. Dynamic sample records in the time-series database reference the equipment primary key in the relational database through foreign keys, enabling joint queries of static attributes and dynamic states.

[0018] The link-layer discovery protocol (LLDP) is used to collect the switching paths of the first communication network. LLDP operates at the data link layer. The master station periodically sends LLDP data frames to each switch in the first communication network. The frame structure includes a local identifier, port identifier, time-to-live (TTL) parameter, and optional organization-specific information fields. The sending period is set to 5 seconds, balancing the real-time nature of link status updates with network bandwidth usage. After receiving LLDP frames from neighboring devices, switches store the neighbor information in the LLDP neighbor table in their local management information base. The master station reads the LLDP neighbor table entries in each switch's management information base using the Simple Network Management Protocol (SMMP) to obtain port-level adjacency relationships. The algorithm reconstructs the complete switching path from the master station to each process-layer intelligent electronic device based on a graph traversal algorithm. This algorithm starts at the master station, ends at the target intelligent electronic device, uses switch port adjacency relationships as edges, and port latency as edge weights. It employs a depth-first search strategy to enumerate all feasible paths, with a path search depth limited to 8 hops. This depth covers all possible paths from the master station to the end intelligent electronic device via no more than 6 switches in the core-aggregation-access three-level architecture of the substation process-layer network, preventing cyclic expansion during graph traversal. The bus segment of the second communication network is acquired through the bus master station's topology scanning function. The second communication network adopts a process fieldbus or controller area network (Controller Area Network) bus architecture. The bus master station periodically sends topology scan request frames, containing the master station address, scan sequence number, and cyclic redundancy check (CRC) code. The scan period is set to 10 seconds. Upon receiving the scan request, each slave device connected to the bus returns a response frame in its physical connection order. The response frame contains the slave device address, device type code, and registration status word. The bus master station establishes a bus segment mapping table based on response timing and device addresses, with a timeout period set to 50ms. If a slave station does not respond within the timeout period, it is marked as offline. The system synchronously collects the hop count of protocol conversion nodes in the target control network, the historical jitter quantile of each communication link, and the local clock deviation of each node. The hop count of protocol conversion nodes is calculated using a path tracing algorithm. The boundary gateway device between the industrial Ethernet and fieldbus is retrieved from the reconstructed exchange path and bus segment mapping table, and the number of protocol conversion nodes traversed from the master station to the target device is counted. The historical jitter quantile is obtained by statistically calculating the transmission delay samples of the most recent 1000 messages. This sample size covers approximately 10 to 15 minutes of continuous communication, achieving a balance between statistical significance and storage overhead, and meeting the sample density requirements for 95th percentile calculation. The master station records the sending timestamp of each control command and the actual arrival timestamp returned by the device; the difference between the two is stored as a transmission delay sample in the timing database. At the end of each statistical period, the accumulated time delay samples within that period are sorted in ascending order, and the 900th, 950th, and 990th sample points are extracted as the 90th, 95th, and 99th percentile jitter values, respectively.Node local clock offset is measured using a precision time protocol. The master station, acting as the clock source boundary clock, sends delay request messages to each node and receives delay response messages. The clock offset is obtained by calculating half the difference between the round-trip time and the dwell time, with a measurement interval of 1 second. The filtering algorithm employs an exponentially weighted moving average method with a smoothing factor of 0.8. This smoothing factor causes the weight of historical samples to decay exponentially, with the current sample accounting for 80% of the weight and historical accumulation accounting for 20%. This method provides rapid convergence for step clock jumps while maintaining sufficient suppression of random noise.

[0019] The hop count, historical jitter percentile, and local clock offset of the switching path, bus segment, and protocol conversion node are integrated into the network status information of the target control network. The integration process uses a structured data encapsulation format, with the target device identifier as the index key, to establish a three-layer data record containing a physical path layer, a link state layer, and a clock synchronization layer. The physical path layer stores the ordered sequence of switching paths corresponding to the device and their mapping relationship to bus segments; the link state layer stores the historical jitter percentile and link congestion flags for each hop on the path; and the clock synchronization layer stores the local clock offset and hop count of protocol conversion nodes for each node on the path. The integration algorithm traverses each device in the device action time limit table, querying its corresponding switching path, bus segment, hop count, jitter percentile, and clock offset through the device identifier, mapping the above heterogeneous data into a unified network status description vector. This vector is maintained in memory using a hash table structure, with the key being the hash value of the device identifier and the value being a serialized byte stream of the above three layers of data records, supporting random access based on device identifier and fuzzy matching based on path features. When any underlying acquisition parameter changes, such as an update of the historical jitter quantile value of a link or a port status switch of a switch, an incremental update mechanism is triggered. This mechanism replaces only the changed data fields while keeping the remaining fields unchanged, ensuring the real-time performance and consistency of network status information. This integration process connects the path description on the industrial Ethernet side with the topology description on the fieldbus side through protocol conversion node hop counts, forming an end-to-end network status description from the master station to the end electrical equipment.

[0020] Step S2: Map the device action time limit table and linkage dependency table to the node features and logical dependency edge features of the graph structure, respectively. Based on the network state information, map them to the physical communication edge features of the graph structure. Combine the node features, logical dependency edge features and physical communication edge features to generate a time limit topology drift attribute graph.

[0021] In one specific embodiment, the process of performing step S2 may specifically include the following steps: The device action time limit table and linkage dependency table are obtained from the hybrid database. Using the attribute graph node feature encoding method, the device action time limit table and linkage dependency table are mapped to the node features and logical dependency edge features of the graph structure, respectively. The switching path, bus segment, hop count of protocol conversion node, historical jitter quantile value and node local clock deviation in the network status information of the target control network are mapped to the physical communication edge features of the graph structure. By using a dual-layer storage fusion method based on attribute graphs, combining node features, physical communication edge features, and logical dependency edge features, an adjacency structure layer and an attribute tensor layer are constructed to generate a time-limited topology drift attribute graph.

[0022] Specifically, the device action time limit table and linkage dependency table are read from the relational storage partition of the hybrid database, while the control cycle and execution dead zone samples of each device within the most recent 168 hours are read from the time-series storage partition. Using the attribute graph node feature encoding method, the device action time limit table and linkage dependency table are mapped to node features and logical dependency edge features of a graph structure, respectively. Node feature encoding uses a fixed-length numerical vector to represent each electrical device. The vector dimension is set to 5 dimensions, determined based on five fields in the device action time limit table: control cycle, maximum allowable arrival time limit, lower limit of execution dead zone, upper limit of execution dead zone, and protection priority. The first dimension stores the normalized value of the control cycle after minimum and maximum scaling, with a lower limit of 0 milliseconds and an upper limit of 500 milliseconds. This upper limit covers the maximum control cycle of all types of electrical equipment in the substation, specifically implemented by dividing the control cycle by 500. The second dimension stores the original value of the maximum allowable arrival time limit. The third and fourth dimensions store the lower and upper limits of the execution dead zone, respectively. The fifth dimension stores the protection priority level encoding: emergency protection level is mapped to the value 1, fast control level to the value 2, normal adjustment level to the value 3, and background maintenance level to the value 4. The linkage dependency table is mapped to logical dependency edge features. Logical dependency edges connect device nodes with action constraints in the graph as directed edges. The feature vector dimension of the edge is set to 4 dimensions, determined by four fields in the dependency table: dependency type, dependency direction, minimum sequence interval, and maximum sequence interval. The first dimension records the dependency type identifier: rigid sequence dependency is assigned a value of 1, flexible temporal dependency is assigned a value of 2, and state-triggered dependency is assigned a value of 3. The second dimension records the dependency direction: forward dependency is assigned a value of 1, and backward dependency is assigned a value of -1. The third and fourth dimensions record the minimum and maximum sequence intervals, respectively. For example, there is a rigid sequence dependency between the opening of the bus tie circuit breaker and the opening of the bus tie isolating switch. The feature vector of this logical dependency edge has a value of 1 in the first dimension, 1 in the second dimension, 300 in the third dimension, and 1500 in the fourth dimension.

[0023] The switching paths, bus segments, hop counts of protocol conversion nodes, historical jitter quantiles, and node local clock offsets in the network status information of the target control network are mapped to physical communication edge features in a graph structure. Physical communication edges connect directly adjacent nodes on the communication line, including those between the master station and the core switch, between the core switch and the aggregation switch, between the aggregation switch and the protocol conversion gateway, between the protocol conversion gateway and the process layer switch, between the process layer switch and intelligent electronic devices, and between the bus master station and the bus slave station. The physical communication edge feature vector is set to 6 dimensions, determined based on six fields in the network status information: link type, hop count of protocol conversion nodes, historical jitter 90th, 95th, and 99th quantiles, and node local clock offset. The first dimension records the link type encoding: 1 for industrial Ethernet switching links, 2 for fieldbus segment links, and 3 for internal forwarding links within the protocol conversion gateway. The second dimension records the cumulative hop count of protocol conversion nodes in the end-to-end path to which the physical communication edge belongs. This hop count is accumulated hop by hop along the switching path and bus segment from the master station, increasing by 1 for each protocol conversion gateway passed through. The third, fourth, and fifth dimensions record the historical jitter 90th, 95th, and 99th percentile values ​​of the communication link, respectively, in microseconds. These three percentile values ​​are extracted from the time-series database by link identifier, with the extraction window covering the most recent 1000 latency samples. The sixth dimension records the local clock offset of the downstream nodes of the physical communication edge, in microseconds. For example, for a physical communication edge from the aggregation switch to the protocol conversion gateway, the first dimension has a value of 3, the second dimension has a value of 1, the third dimension records 8, the fourth dimension records 15, the fifth dimension records 32, and the sixth dimension records -11. This set of data indicates that the hop involves protocol conversion and the jitter is significantly higher than that of a pure switching link.

[0024] A dual-layer storage fusion method for attribute graphs is used, combining node features, physical communication edge features, and logical dependency edge features to construct an adjacency structure layer and an attribute tensor layer, generating a time-limited topology drift attribute graph. The adjacency structure layer uses a compressed sparse row format to store the topological connections of the graph, allocating a row pointer array of length equal to the number of nodes plus one, and a column index array of length equal to the total number of edges. The i-th element in the row pointer array records the starting offset of the adjacent edge of the i-th node in the column index array, and the i-th plus one element records the ending offset; the difference between the two is the out-degree of the i-th node. The column index array sequentially records the target node number pointed to by each edge. The attribute tensor layer uses a separate storage mode. The node attribute tensor shape is N x 5, where N is the total number of nodes, and the element data type is 32-bit floating-point. The physical communication edge attribute tensor shape is M1 x 6, where M1 is the total number of physical communication edges. The logical dependency edge attribute tensor shape is M2 x 4, where M2 is the total number of logical dependency edges. The edge identifier generation algorithm uses a left-shifted 16-bit shifted source node number followed by a bitwise OR operation with the target node number to ensure each edge has a unique identifier. During dual-layer storage fusion, a mapping table is established between edge identifiers and attribute tensor row indices. This mapping table uses an open-addressable hash structure, with the hash function being the hash table capacity modulo the edge identifier. The collision resolution strategy employs linear probing with a probing step size of 1 and a load factor of 0.75. This load factor balances the probability of hash collisions with memory usage. When the number of inserted elements exceeds the product of the capacity and the load factor, resizing is triggered, with a resizing factor of 2. After the time-limited topology drift attribute graph is generated, the adjacency structure layer supports adjacency queries based on node identifiers and path enumeration based on depth-first search, while the attribute tensor layer supports feature retrieval based on edge identifiers and batch attribute calculation based on tensor operations. The two layers establish a one-to-one correspondence through edge identifiers.

[0025] This process integrates device action time limits, linkage dependencies, and network status into a unified graph structure model, solving the problem of information fragmentation between the device side and the network side. It provides a complete perception foundation for subsequent scheduling decisions, enabling joint optimization of the timing of control command transmission and transmission path within the same data structure.

[0026] Step S3: Obtain the target electrical equipment and action stage corresponding to the control command to be executed, and construct a four-dimensional action window tensor by combining the maximum allowable arrival time limit and protection priority, and aggregate it into a candidate scheduling batch according to the preset scheduling cycle.

[0027] In one specific embodiment, the process of performing step S3 may specifically include the following steps: Obtain the control commands to be executed generated by the master station, and parse the target electrical equipment and action stage corresponding to the control commands to be executed; Extract the maximum permissible arrival time and protection priority of the target electrical equipment from the equipment action time limit table; Using tensor encoding, the target electrical equipment, action stage, maximum permissible arrival time limit corresponding to the allowable arrival window, and the priority level corresponding to the protection priority are encoded into a four-dimensional action window tensor. According to the preset scheduling cycle, all four-dimensional action window tensors generated in the same cycle are categorized and aggregated according to the network segment to which the device belongs and the action type to generate candidate scheduling batches.

[0028] Specifically, the control commands to be executed generated by the master station enter the processing flow in the form of messages. The message header contains a target electrical equipment identifier field and an action type field. The parsing process employs a field splitting strategy, extracting the target electrical equipment identifier from the message header. This identifier uses substation coding rules and is composed of the bay number and equipment type code. Action stage information is extracted from the message action field. Action stages are divided into four categories: preparation stage, trigger stage, hold stage, and reset stage. This classification is based on the differences in command sensitivity in the electrical equipment control sequence. The trigger stage has the highest sensitivity to delay, while the preparation and reset stages have decreasing sensitivity in that order. After parsing, the target electrical equipment identifier and action stage code are passed to the query stage.

[0029] The maximum permissible arrival time and protection priority of the target electrical equipment are extracted from the equipment action time limit table. The query process uses the parsed target electrical equipment identifier as the search key, performs an index lookup in the relational storage partition of the hybrid database, and returns the maximum permissible arrival time and protection priority field values ​​for that equipment. The maximum permissible arrival time represents the maximum allowed time span from when the control command is issued from the master station to when it reaches the target electrical equipment. This parameter is pre-configured in the equipment action time limit table according to the equipment type and voltage level; for switch opening and closing commands, this parameter is set to 50ms. The protection priority represents the preemption capability of the control command during the scheduling process, and is divided into four levels: emergency protection level, fast control level, routine regulation level, and background maintenance level. For example, the maximum permissible arrival time of the protection device for the 301 bay line is 30ms, and the protection priority is emergency protection level.

[0030] Using tensor encoding, the target electrical equipment, action phase, allowable arrival window corresponding to the maximum permissible arrival time limit, and priority level corresponding to the protection priority are encoded into a four-dimensional action window tensor. The calculation logic of the allowable arrival window is jointly determined based on the maximum permissible arrival time limit and the execution dead zone parameter. The lower bound of the allowable arrival window is set to the control command generation time plus the estimated minimum transmission delay under the current network conditions, and the upper bound of the allowable arrival window is set to the control command generation time plus the maximum permissible arrival time limit minus half of the lower bound of the execution dead zone. This calculation logic ensures that even when network jitter reaches a statistically bad state, the control command can still arrive at the target electrical equipment before the execution dead zone begins. The first dimension of the four-dimensional action window tensor encodes the target electrical equipment identifier, using hash encoding to map the string identifier to a numerical vector. The hash function uses the BKDR algorithm, and the seed value is set to 31, which strikes a balance between the string hash collision rate and computational complexity. The second dimension encodes the action phase: the preparation phase is mapped to the numerical value 1, the trigger phase to the numerical value 2, the hold phase to the numerical value 3, and the return phase to the numerical value 4. The third dimension encodes the arrival window, consisting of a lower and upper bound. In the four-dimensional tensor, this is expanded into two consecutive numerical positions, storing the offsets of the lower and upper bounds relative to the command generation time. The fourth dimension encodes the priority level: emergency protection level is mapped to value 1, fast control level to value 2, normal adjustment level to value 3, and background maintenance level to value 4. After tensor encoding, each control command is represented as a 1×4 tensor row vector, where the third dimension actually occupies two consecutive numerical positions. Therefore, the storage length of a single four-dimensional action window tensor corresponds to five consecutive numerical positions. In-row elements are stored consecutively in the order of the first, second, and third dimension lower bounds, third dimension upper bounds, and fourth dimension. For example, in the trigger phase closing command of the 301 bay line protection device, the first dimension of its four-dimensional action window tensor is the hash code value of the 301 identifier, the second dimension is 2, the third dimension lower bound is 2.5ms, the third dimension upper bound is 28ms, and the fourth dimension is 1.

[0031] According to a preset scheduling cycle, all four-dimensional action window tensors generated within the same cycle are categorized and aggregated based on the network segment to which the device belongs and the action type to generate candidate scheduling batches. The preset scheduling cycle is set to 4ms, which aligns with the shortest control cycle of protection devices in the process layer network, ensuring that the frequency of scheduling batch generation matches the network transmission capacity. The classification and aggregation algorithm first divides the four-dimensional action window tensor into a first communication network set and a second communication network set based on the network segment to which the device belongs. The network segment determination is based on the network field encoding field in the target electrical equipment identifier. A 1 in the first bit of this field indicates that the device is connected to the first communication network, and a 1 in the second bit indicates that the device is connected to the second communication network. Then, based on the action type, each network segment set is further divided into a circuit breaker subset, a regulating subset, and a reset subset. The action type determination is based on the encoding value of the second dimension of the four-dimensional action window tensor. The triggering stage and the reset stage are assigned to the circuit breaker subset, the holding stage is assigned to the regulating subset, and the preparation stage is assigned to the verification subset. During the aggregation process, four-dimensional action window tensors belonging to the same network segment and the same action type are merged into a scheduling group within the candidate scheduling batch. Each scheduling group is then sorted in ascending order according to the priority level of the fourth dimension, with smaller priority levels indicating higher priority. This arrangement forms a tensor matrix. The candidate scheduling batch data structure is stored in dictionary form, with the key being a combination of the network segment identifier and the action type identifier. The network segment identifier occupies the high 8 bits, and the action type identifier occupies the low 8 bits, resulting in a 16-bit integer code. For example, within a certain scheduling cycle, the master station simultaneously generates a trigger-stage closing command for circuit breaker 301, a trigger-stage opening command for disconnector 302, and a holding-stage upshift command for on-load tap changer of the main transformer. Circuit breakers 301 and 302 are both connected to the first communication network and both involve trigger-stage opening and closing actions, thus belonging to the same scheduling group. The on-load tap changer of the main transformer is connected to the second communication network and involves a regulating action, thus belonging to a separate scheduling group. Within the previous scheduling group, the priority level is sorted according to the fourth dimension. 301, with a priority level of 1, is ranked first in the group, and 302, with a priority level of 2, is ranked second in the group.

[0032] Step S4: Perform joint optimization on the candidate scheduling batch based on the time-limited topology drift attribute graph, determine the mapping relationship between the candidate sending window and the cross-network segment forwarding path, and generate the final control routing table.

[0033] In one specific embodiment, the process of performing step S4 may specifically include the following steps: Based on the expected transmission delay and logical dependency relationship of candidate scheduling batches in the time-limited topology drift attribute graph, calculate the deadline default rate, the number of relative order reversals, and the drift confidence decay value. Based on the deadline default rate, the number of relative order reversals, and the drift confidence decay value, a multi-objective weighted joint optimization objective function is constructed. Using a multi-objective weighted joint optimization objective function as the solution objective, a graph path constraint solution algorithm is used to solve the candidate scheduling batch on the time-limited topology drift attribute graph, and output the mapping relationship between the candidate transmission window and the cross-segment forwarding path corresponding to each control command in the candidate scheduling batch; Collect the link congestion status, protocol conversion node buffer occupancy status, electrical equipment protection status, and action cascading depth on the cross-network segment forwarding path. Combine this with a dynamic weight reallocation algorithm to correct the mapping relationship between candidate sending windows and cross-network segment forwarding paths. Use the corrected mapping relationship as the final control routing table.

[0034] Specifically, based on the expected transmission delay and logical dependencies of the candidate scheduling batch in the time-limited topology drift attribute graph, the deadline default rate, relative order reversal count, and drift reliability decay value are calculated. The expected transmission delay is calculated using a path accumulation mechanism. For each control command corresponding to a cross-segment forwarding path in the candidate scheduling batch, the basic transmission delay of each physical communication edge on the path is accumulated, and a hop count penalty coefficient for protocol conversion nodes is introduced. Assume the path contains... The first physical communication edge, The basic transmission delay of the strip is The protocol conversion node indicator parameter is (If the edge passes through a protocol conversion gateway, then) (Take 1 otherwise, take 0) Estimated transmission delay The specific implementation is as follows: The penalty coefficient of 0.5 for protocol conversion nodes is based on the fact that protocol conversion gateways involve protocol parsing, packet reassembly, and buffer queuing. Actual testing shows that the processing latency is on average 50% to 80% higher than that of pure switching links. A conservative estimate of 0.5 is used to allow for timing margins. The deadline default rate represents the probability of control commands failing when they exceed the upper bound of the allowed arrival window. Let the upper bound of the allowed arrival window be... The path transmission delay follows a normal distribution with a mean of 1 / 2. Take the above-mentioned expected transmission delay Standard deviation of distribution Pick ,in For the first The 95th percentile of the historical jitter of the strip will Substitute into the standard normal cumulative distribution function Deadline default rate The specific implementation is as follows: The relative order reversal count is performed for instruction pairs with rigid sequential dependencies. This involves traversing all logical dependency edges in the candidate scheduling batch, assuming the expected arrival time of the preceding instruction is calculated based on the midpoint of the candidate sending window and the expected transmission delay. The expected arrival time of the subsequent instruction is The minimum sequential interval of the logically dependent edge feature records is ,like This counts as one order reversal. After completing the traversal, the total number of times the condition is met is accumulated. The drift confidence decay value characterizes the degree of deviation between historical jitter data and the current real-time network state. The moving standard deviation of the most recent 1000 latency samples is extracted for each physical communication edge. Calculate the ratio ,like Then the attenuation value of that edge ,otherwise The overall path drift confidence decay value is taken as the arithmetic mean of the decay values ​​of each edge on the path.

[0035] Based on the deadline default rate, the number of relative order reversals, and the drift confidence decay value, a multi-objective weighted joint optimization objective function is constructed. The objective function adopts a weighted summation form, with weight coefficients... This set of coefficients is determined based on the priority relationship among three factors: the direct impact of deadline default on instruction validity, the risk of order reversal jeopardizing equipment safety, and the reflection of model confidence by drift reliability. A normalization factor maps the three dimensional indices to the same numerical range, thus normalizing the deadline default rate. The relative order reversal count is normalized by scaling with the maximum value of all candidate scheduling batches within the batch. Scaling is performed using the total number of logical dependency edges within the batch, and the drift confidence decay value is normalized. The scaling factor is the maximum value obtained by weighting the path hop count. The objective function is a multi-objective weighted joint optimization. The specific implementation is as follows: The solution direction is minimization, that is, seeking to minimize the... The combination of the sending window and path with the smallest value.

[0036] Using a multi-objective weighted joint optimization objective function as the solution objective, a graph path constraint solution algorithm is employed to solve for candidate scheduling batches on a time-limited topology drift attribute graph. The algorithm outputs the mapping relationship between candidate transmission windows and cross-segment forwarding paths corresponding to each control instruction in the candidate scheduling batch. The solution algorithm combines constraint satisfaction with heuristic search. In the initialization phase, a preliminary range of candidate transmission windows is generated for each control instruction, and the lower bound of the transmission window is determined. Set to allow access to the lower boundary of the window Subtract the maximum transmission delay of the path Send window upper boundary Set to allow access to the upper bound of the window Subtract the minimum transmission delay of the path All feasible paths from the master station to the target electrical equipment are enumerated on the time-limited topology drift attribute graph, with a path search depth limit of 8 hops, consistent with the path enumeration constraint of the adjacency structure layer. For each feasible path, its expected transmission delay and transmission delay variance are calculated. The total path delay is obtained by summing the expected transmission delays of each physical communication edge on the path, and the standard deviation of the total path delay is obtained by taking the square root of the sum of the squares of the 95th percentile values ​​of the historical jitter of each edge. Based on this, the probability that the arrival time of the control command under this path exceeds the allowable arrival window, and the probability that the order of the linkage dependent command pair under this path will be reversed are calculated. The joint optimization of the candidate sending window and the path is achieved through iterative adjustment. In each iteration, the sending window boundary is shrunk or expanded according to the current objective function value. The shrinkage step size is set to 0.1 milliseconds, and the expansion step size is set to 0.2 milliseconds. This asymmetric step size setting allows the algorithm to prioritize tightening the timing to improve efficiency, and appropriately relaxes it when the constraints are not satisfied to ensure feasibility. The iteration termination condition is set to the rate of change of the objective function value being less than 0.01 for three consecutive iterations, or reaching the maximum number of iterations of 100.

[0037] The system collects data on link congestion status, protocol conversion node buffer occupancy status, electrical equipment protection status, and action cascading depth along cross-segment forwarding paths. Combined with a dynamic weight reallocation algorithm, it corrects the mapping relationship between candidate sending windows and cross-segment forwarding paths, using the corrected mapping relationship as the final control routing table. Link congestion status is obtained by reading the input and output queue lengths of each switch port. When the queue length exceeds 70% of the port buffer capacity, the link is considered congested, and the path weight is multiplied by a penalty factor. , ,in This refers to queue occupancy. The cache occupancy status of the protocol conversion node is obtained by querying the ratio of the currently used bytes in the protocol conversion gateway's message buffer to its total capacity; this is the cache occupancy rate. Additional penalties are imposed when the percentage exceeds 90%; penalty factor. e is a natural constant; this negative exponential form means that the higher the cache occupancy, the smaller the penalty factor. hour ≈0.41, when hour The protection status of electrical equipment is read in real time from the equipment action time limit table. When the target electrical equipment is in the protection lockout state, the weights of all paths leading to that equipment are directly set to 0, and the path is removed from the candidate set. The action cascading depth is obtained by traversing the linkage dependency table and calculating the length of the subsequent dependency chain starting from the target electrical equipment. Cascade depth At that time, the path delay jitter sensitivity coefficient This coefficient is obtained by taking the reciprocal of the original amplification factor of 1.8, which attenuates the weight of paths with large cascade depths. After dynamic weight redistribution, the optimal path is selected for each control command based on the weight ranking. The weight correction formula is as follows: Where Yold is the initial path weight, obtained by normalizing the inverse of the objective function output by the graph path constraint solving algorithm. , , All values ​​are less than 1, indicating the final weight of the unfavorable path. < The candidate sending window is adjusted accordingly to match the latency characteristics of the new path, and the lower and upper bounds of the adjusted sending window and the selected path identifier are written into the final control routing table.

[0038] refer to Figure 2 This figure shows the iterative convergence curves of the objective function in a multi-objective joint optimization algorithm, illustrating the decrease in the objective function value at different iteration numbers. The figure contains three curves, corresponding to three strategies: optimizing only the deadline default rate, optimizing only the order reversal, and joint optimization. The comparison demonstrates the advantages of the joint optimization strategy in terms of convergence speed and final optimization effect.

[0039] refer to Figure 3 This figure compares the deadline default rates under different network load conditions, showing the trend of deadline default rates for the three scheduling methods as network load increases. The figure compares the static routing method, the dynamic priority method, and the time-limited topology drift-aware scheduling method proposed in this application, intuitively demonstrating the effectiveness of this method in suppressing the risk of instruction failure under high load conditions.

[0040] Step S5: Verify the control command based on the device action time limit table and the final control routing table, calculate the delay compensation parameters, write and send the control command to the target electrical equipment.

[0041] In one specific embodiment, the process of performing step S5 may specifically include the following steps: Extract the execution dead zone of the target electrical equipment from the equipment action time limit table. Based on the instruction sending order and execution dead zone in the final control routing table, verify whether the control instructions meet the preset rigid sequence rules. Add an executable flag to the control instructions that pass the verification. Based on the executable flag and the path transmission delay in the final control routing table, calculate the delay compensation parameters and write the corresponding control instructions. The delay compensation parameters include the local delay compensation amount and the advance triggering amount. Control commands are sent to the target electrical equipment via the indicated path through the sending window corresponding to the final control routing table.

[0042] Specifically, the execution dead zone of the target electrical equipment is extracted from the equipment action time limit table. The execution dead zone represents the minimum time window that the electrical equipment must go through from receiving a control command to the actual action of the mechanical actuator. The extraction process uses the target electrical equipment identifier as the search key, performs an index query in the relational storage partition of the hybrid database, and returns the lower and upper limits of the execution dead zone field values. The lower limit of the execution dead zone represents the shortest duration required from the start of the internal electromagnetic mechanism of the equipment to the change of the contact state, and the upper limit of the execution dead zone represents the longest duration after considering mechanical wear and temperature drift. Based on the command sending order and execution dead zone in the final control routing table, it is verified whether the control commands meet the rigid priority rule. The final control routing table records the lower and upper limits of the sending window for each control command in the candidate scheduling batch. The command sending order is arranged in ascending order according to the lower limit of the sending window, and the control command with the smaller lower limit value is sent first. The rigid priority rule comes from the linkage dependency table, which records the pre-dependency and post-dependency relationships between devices, as well as the minimum priority interval time. The verification algorithm iterates through all instruction pairs with rigid sequential dependencies in the candidate scheduling batch. Let W be the lower bound of the sending window of the preceding instruction in the final control routing table, and X be the path transmission delay; then the expected arrival time of the preceding instruction is W+X. Let Y be the lower bound of the sending window of the following instruction, and Z be the path transmission delay; then the expected arrival time of the following instruction is Y+Z. Let H be the minimum sequential interval recorded in the dependency table, and I be the lower limit of the execution dead zone; then the actual executable time of the following instruction is no earlier than the expected arrival time of the preceding instruction plus H plus I. The specific implementation of the verification condition is as follows: If this inequality holds, then the rigid priority rule is satisfied; if it does not hold, then the time gap is calculated. When F is greater than 0, a local delay compensation amount needs to be applied to the subsequent instruction. The compensation amount is set to F, or the sending window of the subsequent instruction is shifted backward by F time. An executable flag is added to the control instructions that pass the verification. The executable flag is a 1-bit Boolean field embedded in the metadata area of ​​the control instruction header. The field length is 1 byte; a value of 1 indicates that the verification has passed and the instruction can be sent, while a value of 0 indicates that the verification has failed and the instruction is temporarily suspended. Control instructions that fail the verification enter the waiting queue. After the execution confirmation information of the preceding dependent instruction is returned, the verification process is retried. Control instructions that pass the verification enter the compensation amount calculation stage.

[0043] Based on the executable flag and the path transmission delay in the final control routing table, delay compensation parameters are calculated. These parameters include local delay compensation and early triggering. Local delay compensation is used when a subsequent command arrives too early, causing the target electrical equipment to delay its action after receiving the command, waiting for the preceding dependency conditions to be met. Early triggering is used when the path transmission delay fluctuates significantly but the command needs to arrive precisely, causing the master station to issue the control command before the lower bound of the transmission window, utilizing the minimum path delay characteristic to ensure the command arrives at the target electrical equipment at the desired time. Let the average path transmission delay be... The standard deviation is Allow access to the lower boundary of the window. Then the amount of early triggering The specific implementation is as follows: In this formula The basis for this is that 99.7% of the samples under a normal distribution fall within the range of mean ± 3 standard deviations, and 3 standard deviations is taken as the extreme jitter margin. The specific implementation of the local delay compensation amount P is as follows: When P is greater than 0, P is written to the delay field of the control command; when... When it is greater than 0, Write the advance trigger field of the control command. If the same command requires both local delay compensation and advance triggering, the local delay compensation amount is applied first, and the advance trigger amount is set to 0 to avoid compensation logic conflicts.

[0044] The control commands are sent to the target electrical equipment according to the sending window corresponding to the final control routing table, via the indicated path. The lower and upper boundaries of the sending window are read from the final control routing table. The master station scheduler maintains a sending queue, which is sorted by time according to the lower boundary of the sending window. When the lower boundary of the sending window is reached, the scheduler encapsulates the control command carrying the executable flag and delay compensation parameters into a message conforming to the first communication network protocol. The message header is marked with the target electrical equipment identifier, action type, local delay compensation amount, advance trigger amount, and executable flag. After the message is sent from the master station, it is forwarded hop by hop along the path sequence indicated by the final control routing table, passing through the core switch, aggregation switch, protocol conversion gateway, and process layer switch, and finally reaching the target electrical equipment. After parsing the message, if the executable flag is 1 and the local delay compensation amount is greater than 0, the target electrical equipment starts a delay countdown in the internal timer. After the countdown ends, the mechanical actuator is triggered to act. If the advance trigger amount is greater than 0, the equipment ignores this parameter, and the master station executes the advance amount at the sending time. During message forwarding, each intermediate node looks up its local forwarding table based on the target address in the message header. The protocol conversion gateway performs format conversion from industrial Ethernet message to fieldbus message, keeping the payload unchanged during the conversion process and re-encapsulating the bus address, function code, and checksum field.

[0045] Step S6: Obtain the execution confirmation information returned by the target electrical equipment, and update the time limit topology drift attribute graph and linkage dependency table.

[0046] In one specific embodiment, the process of performing step S6 may specifically include the following steps: Receive the execution confirmation information returned by the target electrical equipment corresponding to the issued control command, combine it with the expected transmission delay in the time-limited topology drift attribute graph, calculate the deviation value between the actual transmission delay and the expected transmission delay, and use the deviation value as a new jitter sample. The execution confirmation information includes: action completion timestamp and execution result code. Based on the new jitter samples and execution confirmation information, the historical jitter quantiles and node local clock deviations of each communication link in the time-limited topology drift attribute map are updated to obtain the updated time-limited topology drift attribute map. Based on the action completion timestamp in the execution confirmation information, calculate the actual linkage interval between devices and update the time window parameters of the pre- and post-linkage dependencies in the linkage dependency table.

[0047] Specifically, the system receives execution confirmation information from the target electrical equipment corresponding to the issued control command. This confirmation information is generated by the target electrical equipment after the action is completed and transmitted back to the master station along the original path. This information includes two fields: an action completion timestamp and an execution result code. The action completion timestamp records the timing value of the target electrical equipment's local clock at the instant the mechanical actuator completes its state switch. The execution result code uses enumeration encoding: a value of 0 indicates successful action completion, a value of 1 indicates the action was rejected due to protection locking, and a value of 2 indicates the action was discarded because it exceeded the allowed arrival window. Upon receiving the execution confirmation information, the master station parses the action completion timestamp and execution result code and correlates them with the original issued control command. The correlation matching is based on the consistency between the sequence number field in the control command message and the response sequence number field in the execution confirmation information. The master station records the sending timestamp when the control command is issued and calculates the actual transmission delay by combining it with the actual arrival timestamp in the execution confirmation information. The actual transmission delay equals the actual arrival timestamp minus the sending timestamp. The time-limited topology drift attribute graph stores the expected transmission delay of the cross-segment forwarding path corresponding to the control command. This expected transmission delay is calculated by accumulating the paths during the joint pathfinding phase. The deviation between the actual transmission delay and the expected transmission delay is calculated; the deviation is equal to the actual transmission delay minus the expected transmission delay. This deviation represents the combined impact of network jitter and path state drift during this transmission process and is stored as a new jitter sample in the time-series database. The jitter sample is stored in a triplet format, containing a path identifier, a timestamp, and a deviation value. The path identifier is generated by combining the source node number and the target node number, the timestamp is the time the master station sends the data, and the deviation value is the result of the above calculation.

[0048] Based on new jitter samples and execution confirmation information, the historical jitter quantiles and node local clock offsets of each communication link in the time-limited topology drift attribute graph are updated. The historical jitter quantiles are updated using a sliding window statistical method. The most recent 1000 jitter samples for each communication link are extracted from the time-series database to form a sliding window sample set. This sample set is sorted in ascending order of numerical value, and the 900th, 950th, and 990th sample points are extracted as the 90th, 95th, and 99th quantiles, respectively. When the sample set has fewer than 1000 samples, the rank is calculated proportionally to the existing sample count, and non-integer ranks are handled using linear interpolation. The updated quantiles are written into the 3rd, 4th, and 5th dimensions of the edge feature vector of the corresponding physical communication edge in the time-limited topology drift attribute graph. The node local clock offset is updated based on a systematic difference analysis between the actual arrival timestamp and the sending timestamp. Let the sending timestamp be A, the actual arrival timestamp be B, and the expected transmission delay extracted from the time-limited topology drift attribute graph be... If, after deducting the expected transmission delay, there is a systematic difference D between the device's local clock and the master station's clock, then the specific implementation of D is as follows: The difference value is averaged after multiple measurements and then updated to the node characteristic clock deviation field of the corresponding node in the time-limited topology drift attribute graph. The filtering process uses an exponentially weighted moving average method. Let the current systematic difference measurement value be D, the previous clock deviation estimate be Q, and the smoothing factor S be set to 0.2. Then the updated clock deviation... The specific implementation is as follows: The smoothing factor S is set to 0.2 because clock drift is a slowly varying process. A smaller smoothing factor gives historical estimates 80% weight and current measurements 20% weight, suppressing random noise in a single measurement and preventing drastic fluctuations in clock bias estimation due to occasional jitter. The updated historical jitter quantiles and the node's local clock bias together constitute the updated time-limited topology drift attribute map.

[0049] Based on the action completion timestamp in the execution confirmation information, the actual linkage interval between devices is calculated, and the time window parameters of the pre- and post-linkage dependencies in the linkage dependency table are updated. The calculation of the actual linkage interval applies to device pairs with pre- and post-linkage dependencies. Let the action completion timestamp of the pre-linkage device be T, and the action completion timestamp of the post-linkage device be U, then the specific implementation of the actual linkage interval V is as follows: The actual linkage interval reflects the true timing relationship of mechanical actions between devices, including the combined effects of network transmission latency, device execution dead zones, and mechanical response time. The actual linkage interval is compared with the minimum sequential interval recorded in the linkage dependency table. If the actual linkage interval is less than the minimum sequential interval, a risk of sequence violation is identified. In this case, the lower limit of the time window parameter for that dependency relationship in the linkage dependency table is adjusted to the actual linkage interval, while the upper limit remains unchanged. If the actual linkage interval is greater than the maximum sequential interval, the upper limit of the time window parameter is adjusted to the actual linkage interval, while the lower limit remains unchanged. If the actual linkage interval is between the minimum and maximum sequential intervals, the time window parameter is fine-tuned. The fine-tuning amount uses a weighted average of the actual linkage interval and the midpoint of the historical time window, with weighting coefficients set to 0.3 and 0.7. This weighting is based on the fact that historical time window parameters reflect long-term statistical patterns and occupy a dominant weight, while the current actual linkage interval reflects recent operating conditions and occupies a secondary weight, avoiding drastic drift of the time window parameter due to a single abnormal operation. After the time window parameter is updated, it is written to the linkage dependency table. All subsequent control commands involving this device pair are rigidly validated based on the updated time window parameter.

[0050] Step S7: Based on the updated time-limited topology drift attribute graph, assess the risk of sequential violation, determine and issue subsequent control instructions, and generate a collaborative control assessment report.

[0051] In one specific embodiment, the process of performing step S7 may specifically include the following steps: Based on the updated time-limited topology drift attribute graph, calculate the order violation risk value and determine whether the order violation risk value exceeds the preset risk threshold. If so, a protection priority degradation sequence is constructed based on the protection priority in the equipment action time limit table, and subsequent control commands are reordered according to the protection priority degradation sequence before being issued. If not, continue to issue subsequent control commands according to the path and sending window policy of the currently effective final control routing table; Execute subsequent control commands, receive execution confirmation information returned by the target electrical equipment, trigger exception handling process based on the execution result code in the execution confirmation information, and statistically analyze the collaborative control performance indicators within a preset period and generate a collaborative control evaluation report.

[0052] Specifically, based on the updated time-limited topology drift attribute map, a sequence violation risk value is calculated. This risk value represents the probability of a relative sequence reversal in subsequent control commands. The calculation process uses the Monte Carlo simulation method. For each pair of control commands with rigid sequential dependencies in the candidate scheduling batch, the expected transmission delays of the preceding and following command paths are extracted from the updated time-limited topology drift attribute map as the mean. The 95th percentile of the historical jitter of each communication link on each path is extracted, and the square root of the sum of squares is calculated as the standard deviation. Normal distributions of the transmission delays of the preceding and following paths are constructed respectively. In each simulation, the path transmission delays of the preceding and following commands are extracted from their corresponding normal distributions. The simulated arrival time of the following command is subtracted from the simulated arrival time of the preceding command. If the difference is less than the minimum sequential interval recorded in the linkage dependency table, a sequence reversal is determined. The above simulation process is repeated 1000 times. This number is determined according to the statistical confidence requirement. 1000 simulations can control the estimation error within 3%. The ratio of the number of sequence reversals to the total number of simulations is the sequence violation risk value. The preset risk threshold is set at five percent, which is determined based on the statistical probability of acceptable maloperation of power system relay protection, achieving a balance between control safety and dispatch efficiency.

[0053] The system determines whether the risk value of a sequential violation exceeds a preset risk threshold. If so, it constructs a protection priority degradation sequence based on the protection priority in the equipment action time limit table. Protection priorities are divided into four levels: emergency protection, fast control, routine regulation, and background maintenance. When constructing the degradation sequence, all subsequent control commands in the candidate scheduling batch are sorted from highest to lowest protection priority, with emergency protection at the beginning, fast control next, routine regulation next, and background maintenance at the end. After sorting, subsequent control commands are sent serially according to this protection priority degradation sequence. After each control command is sent, the master station waits for the execution confirmation information corresponding to that command to return. Only after confirming the completion of the action or clarifying the execution status is the next command sent. This serial sending mode reduces network concurrency load and minimizes timing interleaving caused by multi-path parallel transmission, thereby suppressing sequential violations. For example, in the fault isolation operation after the main transformer trips, the candidate scheduling batch includes the main transformer high-voltage side isolating switch opening command, the neutral point grounding switch opening command, and the backup power supply automatic connection command. When the sequential violation risk value exceeds the threshold, the command to open the isolating switch on the high-voltage side of the main transformer belongs to the emergency protection level and is placed at the forefront of the protection priority degradation sequence. The remaining commands are arranged in sequence and issued one by one.

[0054] If the sequence violation risk value does not exceed the preset risk threshold, subsequent control commands will continue to be issued according to the path and transmission window policy of the currently effective final control routing table. Parallel scheduling mode is maintained, with multiple control commands in the candidate scheduling batch transmitted simultaneously through their respective allocated paths. Each control command is independently forwarded on its corresponding path according to the lower and upper bounds of the transmission window recorded in the final control routing table, fully utilizing the parallel transmission capabilities of heterogeneous communication networks and shortening the overall control cycle.

[0055] After executing subsequent control commands, the system receives execution confirmation information from the target electrical equipment. This confirmation information includes an action completion timestamp and an execution result code. The execution result code uses enumeration encoding: 0 indicates successful action completion, 1 indicates the action was rejected due to protection interlocking, and 2 indicates the action was discarded because it exceeded the allowed arrival window. An exception handling process is triggered based on the execution result code: When the execution result code is 1, the master station queries the current protection status of the target electrical equipment to determine if the protection interlocking condition has been eliminated. If the interlocking condition still exists, subsequent control commands related to the equipment are paused, and verification and re-issuance are retried after the interlocking condition is eliminated. If the interlocking condition has been eliminated, the control command is regenerated and enters the scheduling process. When the execution result code is 2, the master station analyzes whether the actual transmission delay of the path corresponding to the control command exceeds expectations. If the path delay is abnormal, the path is marked as unavailable in the time-limited topology drift attribute graph, and an alternative path is selected to resend the command. The system also statistically analyzes the collaborative control performance indicators within a preset period, set to 1 hour, which covers the complete equipment inspection interval and network load fluctuation cycle. The performance indicators for coordinated control include the deadline default rate, the number of sequence reversals, the number of path switching events, the number of protection priority degradation triggers, the average path delay, and the path delay variance. The deadline default rate is the proportion of control commands that failed due to exceeding the allowable arrival window within the statistical period, out of the total number of commands issued. The number of sequence reversals is the number of command pairs with relative sequence reversals within the statistical period. The number of path switching events is the number of times the system switched to an alternate path due to original path failure or congestion within the statistical period. The number of protection priority degradation triggers is the number of times a degradation sequence was initiated due to sequence violation risk exceeding the threshold within the statistical period. A coordinated control evaluation report is generated, incorporating the above performance indicators. The report is archived at an hourly granularity to provide data support for subsequent operation and maintenance and parameter tuning. For example, if the closing command of a certain outgoing switch fails due to a communication message verification error, the master station determines that the failure is an intermittent communication failure based on the execution result code and immediately resends the command through the same path. If it fails after three consecutive resends, it switches to the alternate path and resends the command.

[0056] It is understood that the executing entity of this application can be an electrical equipment collaborative control system based on a distributed control system, or it can be a terminal or a server; the specific implementation is not limited here. This application's embodiments use a server as an example for illustration.

[0057] The above describes the electrical equipment cooperative control method based on a distributed control system in the embodiments of this application. The following describes the electrical equipment cooperative control system based on a distributed control system in the embodiments of this application. Please refer to [link / reference]. Figure 4 One embodiment of the electrical equipment collaborative control system based on a distributed control system in this application includes: The data acquisition module is used to obtain the device action time limit table, linkage dependency table and network status information in the target control network; The topology modeling module is used to map the device action time limit table and linkage dependency table into the node features and logical dependency edge features of the graph structure, respectively, and to map them into the physical communication edge features of the graph structure based on network state information. The module combines the node features, logical dependency edge features and physical communication edge features to generate a time limit topology drift attribute graph. The window batch construction module is used to obtain the target electrical equipment and action stage corresponding to the control command to be executed, combine the maximum allowable arrival time limit and protection priority to construct a four-dimensional action window tensor, and aggregate it according to the preset scheduling cycle to form a candidate scheduling batch. The joint routing module is used to jointly optimize candidate scheduling batches based on the time-limited topology drift attribute graph, determine the mapping relationship between candidate sending windows and cross-network segment forwarding paths, and generate the final control routing table. The verification and distribution module is used to verify control commands based on the device action time limit table and the final control routing table, calculate delay compensation parameters, write and distribute control commands to the target electrical equipment; The status update module is used to obtain the execution confirmation information returned by the target electrical equipment and update the time limit topology drift attribute graph and linkage dependency table; The assessment and follow-up control module is used to assess the risk of sequential violations based on the updated time-limited topology drift attribute graph, determine and issue subsequent control commands, and generate a collaborative control assessment report.

[0058] Through the collaborative efforts of the aforementioned components, the system uses a time-limited topology drift attribute graph as a unified data representation structure, covering the entire process of information acquisition, model building, scheduling decision-making, command execution, and status feedback. The data acquisition module obtains device action time limit tables, linkage dependency tables, and network status information from heterogeneous networks. The topology modeling module integrates the execution characteristics of the device side and the transmission characteristics of the network side into a single graph model, solving the problem of separate processing and difficulty in joint optimization of the two types of information in existing methods. The window batch construction module encodes control commands into multi-dimensional action window tensors carrying allowable arrival windows and protection levels, and aggregates them according to the scheduling cycle. The joint pathfinding module performs multi-objective solutions on the graph model with deadline default rate, relative order reversal count, and drift reliability decay value as objectives, simultaneously determining the sending window and cross-network segment forwarding path for each command, avoiding linkage order conflicts that may be introduced by step-by-step decision-making on sending timing and transmission paths. The verification and distribution module calculates delay compensation parameters based on the path delay distribution and the execution dead zone of the target device and injects them into the control command. This proactive compensation mechanism replaces the traditional fixed compensation method, improving the reliability of command arrival within the expected time window. The status update module corrects jitter data, clock skew, and time window parameters in the dependency table in the graph model based on the actual arrival timestamp and action completion timestamp returned by the device, ensuring that the system's description of network status and device characteristics continuously approximates actual operating conditions. The assessment and continuation control module uses the updated graph model to calculate the risk of sequential violations. When the risk exceeds the threshold, it switches to a serial distribution sequence arranged according to protection priority. After the risk subsides, it resumes the parallel scheduling mode and generates a collaborative control assessment report, providing data for subsequent operation and maintenance and parameter tuning.

[0059] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for coordinated control of electrical equipment based on a distributed control system, characterized in that, include: S1. Obtain the device action time limit table, linkage dependency table and network status information in the target control network; S2. Map the device action time limit table and linkage dependency table to the node features and logical dependency edge features of the graph structure, respectively. Based on the network state information, map it to the physical communication edge features of the graph structure. Combine the node features, the logical dependency edge features and the physical communication edge features to generate a time limit topology drift attribute graph. S3. Obtain the target electrical equipment and action stage corresponding to the control command to be executed, combine the maximum allowable arrival time limit and protection priority to construct a four-dimensional action window tensor, and aggregate it according to the preset scheduling cycle to form a candidate scheduling batch. S4. Based on the time-limited topology drift attribute graph, perform joint optimization on the candidate scheduling batch, determine the mapping relationship between the candidate sending window and the cross-network segment forwarding path, and generate the final control routing table; S5. Verify the control command based on the device action time limit table and the final control routing table, calculate the delay compensation parameter, write and send the control command to the target electrical equipment; S6. Obtain the execution confirmation information returned by the target electrical equipment, and update the time-limited topology drift attribute graph and the linkage dependency table; S7. Based on the updated time-limited topology drift attribute graph, assess the risk of sequential violation, determine and issue subsequent control instructions, and generate a collaborative control assessment report.

2. The method according to claim 1, characterized in that, S1 includes: Obtain the target control network, which includes the first communication network and the second communication network. Extract the equipment action time limit table, which contains the control cycle, maximum allowable arrival time limit, execution dead zone and protection priority of each electrical device, and the linkage dependency table, which contains the linkage pre- and post-linkage dependencies, from the target control network. Store the equipment action time limit table and the linkage dependency table in the hybrid database. The hybrid database adopts a storage mode that combines time-series database and relational database. The switching path of the first communication network is collected using the link layer discovery protocol, and the bus segment of the second communication network is collected through the bus master station topology scanning function. The hop count of the protocol conversion node in the target control network, the historical jitter percentile of each communication link, and the local clock deviation of each node are collected synchronously. The switching path, bus segment, hop count of protocol conversion nodes, historical jitter percentile value, and local clock deviation are integrated into the network status information of the target control network.

3. The method according to claim 2, characterized in that, S2 includes: The device action time limit table and linkage dependency table are obtained from the hybrid database. Using the attribute graph node feature encoding method, the device action time limit table and linkage dependency table are mapped to the node features and logical dependency edge features of the graph structure, respectively. The switching path, bus segment, hop count of protocol conversion node, historical jitter quantile value and node local clock deviation in the network status information of the target control network are mapped to the physical communication edge features of the graph structure. By using a dual-layer storage fusion method based on attribute graphs, combining node features, physical communication edge features, and logical dependency edge features, an adjacency structure layer and an attribute tensor layer are constructed to generate a time-limited topology drift attribute graph.

4. The method according to claim 1, characterized in that, S3 includes: Obtain the control commands to be executed generated by the master station, and parse the target electrical equipment and action stage corresponding to the control commands to be executed; Extract the maximum permissible arrival time and protection priority of the target electrical equipment from the equipment action time limit table; Using tensor encoding, the target electrical equipment, action stage, maximum permissible arrival time limit corresponding to the allowable arrival window, and the priority level corresponding to the protection priority are encoded into a four-dimensional action window tensor. According to the preset scheduling cycle, all four-dimensional action window tensors generated in the same cycle are categorized and aggregated according to the network segment to which the device belongs and the action type to generate candidate scheduling batches.

5. The method according to claim 1, characterized in that, S4 includes: Based on the expected transmission delay and logical dependency relationship of candidate scheduling batches in the time-limited topology drift attribute graph, calculate the deadline default rate, the number of relative order reversals, and the drift confidence decay value. Based on the deadline default rate, the number of relative order reversals, and the drift confidence decay value, a multi-objective weighted joint optimization objective function is constructed. Using a multi-objective weighted joint optimization objective function as the solution objective, a graph path constraint solution algorithm is used to solve the candidate scheduling batch on the time-limited topology drift attribute graph, and output the mapping relationship between the candidate transmission window and the cross-segment forwarding path corresponding to each control command in the candidate scheduling batch; Collect the link congestion status, protocol conversion node buffer occupancy status, electrical equipment protection status, and action cascading depth on the cross-network segment forwarding path. Combine this with a dynamic weight reallocation algorithm to correct the mapping relationship between candidate sending windows and cross-network segment forwarding paths. Use the corrected mapping relationship as the final control routing table.

6. The method according to claim 1, characterized in that, S5 includes: Extract the execution dead zone of the target electrical equipment from the equipment action time limit table. Based on the instruction sending order and execution dead zone in the final control routing table, verify whether the control instructions meet the preset rigid sequence rules. Add an executable flag to the control instructions that pass the verification. Based on the executable flag and the path transmission delay in the final control routing table, calculate the delay compensation parameters and write the corresponding control instructions. The delay compensation parameters include the local delay compensation amount and the advance triggering amount. Control commands are sent to the target electrical equipment via the indicated path through the sending window corresponding to the final control routing table.

7. The method according to claim 1, characterized in that, S6 includes: Receive the execution confirmation information returned by the target electrical equipment corresponding to the issued control command, combine it with the expected transmission delay in the time-limited topology drift attribute graph, calculate the deviation value between the actual transmission delay and the expected transmission delay, and use the deviation value as a new jitter sample. The execution confirmation information includes: action completion timestamp and execution result code. Based on the new jitter samples and execution confirmation information, the historical jitter quantiles and node local clock deviations of each communication link in the time-limited topology drift attribute map are updated to obtain the updated time-limited topology drift attribute map. Based on the action completion timestamp in the execution confirmation information, calculate the actual linkage interval between devices and update the time window parameters of the pre- and post-linkage dependencies in the linkage dependency table.

8. The method according to claim 1, characterized in that, The S7 includes: Based on the updated time-limited topology drift attribute graph, calculate the order violation risk value and determine whether the order violation risk value exceeds the preset risk threshold. If so, a protection priority degradation sequence is constructed based on the protection priority in the equipment action time limit table, and subsequent control commands are reordered according to the protection priority degradation sequence before being issued. If not, continue to issue subsequent control commands according to the path and sending window policy of the currently effective final control routing table; Execute subsequent control commands, receive execution confirmation information returned by the target electrical equipment, trigger exception handling process based on the execution result code in the execution confirmation information, and statistically analyze the collaborative control performance indicators within a preset period and generate a collaborative control evaluation report.

9. A collaborative control system for electrical equipment based on a distributed control system, used to implement the method as described in any one of claims 1-8, characterized in that, include: The data acquisition module is used to obtain the device action time limit table, linkage dependency table and network status information in the target control network; The topology modeling module is used to map the device action time limit table and linkage dependency table into node features and logical dependency edge features of the graph structure, respectively, and to map the network state information into physical communication edge features of the graph structure. Combining the node features, the logical dependency edge features and the physical communication edge features, a time limit topology drift attribute graph is generated. The window batch construction module is used to obtain the target electrical equipment and action stage corresponding to the control command to be executed, combine the maximum allowable arrival time limit and protection priority to construct a four-dimensional action window tensor, and aggregate it according to the preset scheduling cycle to form a candidate scheduling batch. The joint routing module is used to jointly optimize the candidate scheduling batch based on the time-limited topology drift attribute graph, determine the mapping relationship between candidate sending windows and cross-network segment forwarding paths, and generate the final control routing table. The verification and distribution module is used to verify the control command based on the device action time limit table and the final control routing table, calculate the delay compensation parameters, write and distribute the control command to the target electrical equipment. The status update module is used to obtain the execution confirmation information returned by the target electrical equipment and update the time-limited topology drift attribute graph and the linkage dependency table; The assessment and follow-up control module is used to assess the risk of sequential violations based on the updated time-limited topology drift attribute graph, determine and issue subsequent control commands, and generate a collaborative control assessment report.

Images