A load coordination computing method and system for a power grid edge node

Abstract

The present application relates to the technical field of power system intelligent computing, in particular to a load collaborative computing method and system for power grid edge nodes, which collects power distribution network edge node data to calculate the time sequence change direction, screens voltage continuous change nodes to form a load propagation record sequence, constructs a change fragment set in combination with the relationship between reactive power and power factor, divides the collaborative interval according to the active direction and eliminates the abnormal, generates a propagation order chain according to the time sequence to determine the propagation starting node, aligns the time and maps the parameters to output the collaborative computing result. The present application, by jointly screening the voltage current active and power factor state, combining the topological order connection node track to form a propagation correlation link, using the consistency of reactive power and power factor to divide the fragments to enhance the identification ability, combining the change time sequence to establish the cross-node mapping relationship, effectively distinguishes the real change and random disturbance, weakens the cross fluctuation interference of the feeder, enhances the path recognition ability and the consistency of collaborative computing.

Description

Technical Field

[0001] This invention relates to the field of intelligent computing technology for power systems, and in particular to a load collaborative computing method and system for power grid edge nodes. Background Technology

[0002] The field of intelligent computing technology for power systems encompasses the technologies for power system data acquisition, transmission, storage, and analysis. This technology primarily involves the real-time acquisition and management of grid load, generation, transmission and distribution equipment, and user-side electricity consumption information. Its core components include data monitoring of power system operating status, load forecasting, optimized scheduling, and collaborative processing using edge computing. Overall, this technology focuses on the efficient fusion of multi-source data in the power system, the utilization of distributed computing resources, and data-driven decision support to achieve accurate calculation and analysis of power operating parameters and load changes.

[0003] Among them, a load collaborative computing method and system for power grid edge nodes refers to a technical solution for distributed collaborative computing of load data on power system edge nodes. It involves technical matters of load data collection, distribution and collaborative computing. Specifically, it includes obtaining power load information through edge nodes, exchanging data between nodes according to a preset computing strategy, performing local processing and aggregation computing on load data, and allocating computing tasks among nodes using task partitioning and computing scheduling methods, thereby forming a load collaborative computing system.

[0004] In existing technologies, data changes are mostly analyzed as independent states during edge node load processing. The correlation process between nodes relies on static connection relationships for judgment. In scenarios where the feeder operating status is constantly changing and multiple nodes fluctuate synchronously, there is a lack of unified identification capability for the continuous correlation process of voltage, current and power changes. Different sources of change are prone to superimposed. For example, when adjacent sections generate load fluctuations at the same time, it is difficult to distinguish between continuous propagation changes and short-term random changes. This leads to a mixed phenomenon in the propagation boundary identification process, causing deviations in the judgment of propagation order between nodes, affecting the stability of load collaborative analysis and the accuracy of propagation results in complex power distribution environments. Summary of the Invention

[0005] To address the technical problems existing in the prior art, embodiments of the present invention provide a load coordination calculation method and system for power grid edge nodes. The technical solution is as follows: A load coordination calculation method for power grid edge nodes includes the following steps: S1: Collect voltage, current, active power, reactive power and switch status within the continuous sampling window of the edge node to form initial node data, calculate the power factor and determine the topology of the energized feeder, compare the changes of adjacent sampling windows, filter the node data with continuous voltage changes and connect them according to feeder segments to form a node load propagation record sequence. S2: Based on the node load propagation record sequence, read the reactive power and power factor change direction of the nodes with continuously changing voltage, determine the same or opposite relationship, connect the nodes with consistent state and relationship in time, and form the node data segments according to the topology to obtain the node load change segment set. S3: Extract the feeder segments of the node load change segment set, arrange the segment nodes according to the topology, divide the collaborative interval boundary by comparing the active power change direction, remove the segment nodes where reactive power increases and voltage decreases or reactive power decreases and voltage increases, connect the remaining segment nodes to form collaborative node data, and obtain the node collaborative calculation interval set. S4: Based on the set of node collaborative calculation intervals, read the first sampling time of the active power, voltage, and current changes of the collaborative node data, compare the current and voltage of adjacent collaborative nodes according to the feeder topology, determine the propagation start node data, connect the node data groups in the chain, and obtain the node load propagation sequence chain. S5: Using the connection order of the propagation start node data and the data of the inner edge nodes in the node load propagation sequence chain as the basis for collaborative calculation, extract the active power, voltage and current change process in the chain, align the time of adjacent inner edge nodes and map the voltage and current changes to obtain the collaborative calculation result of the power grid node load.

[0006] A load collaborative computing system for power grid edge nodes, the system comprising: The sampling and identification module collects voltage, current, active power, reactive power and switch status within the continuous sampling window of the edge nodes to form initial node data, calculates the power factor and determines the topology of the energized feeder, compares the changes of adjacent sampling windows, filters the node data with continuous voltage changes and connects them according to feeder segments to form a node load propagation record sequence. The relationship segmentation module, based on the node load propagation record sequence, reads the reactive power and power factor change directions of nodes with continuously changing voltage, determines the same or opposite relationship, connects nodes with consistent states and relationships in time, and forms segment node data according to the topology to obtain a set of node load change segments. The interval screening module extracts the feeder segments of the node load change segment set, arranges the segment nodes according to the topology, divides the collaborative interval boundary by comparing the active power change direction, removes segment nodes where reactive power increases and voltage decreases or reactive power decreases and voltage increases, connects the remaining segment nodes as collaborative node data, and obtains the node collaborative calculation interval set. The timing source determination module, based on the set of node collaborative calculation intervals, reads the first sampling time of the active power, voltage, and current changes of the collaborative node data, compares the current and voltage of adjacent collaborative nodes according to the feeder topology, determines the propagation start node data, connects the node data groups in the chain, and obtains the node load propagation sequence chain. The mapping calculation module uses the connection order of the propagation start node data and the data of the inner edge nodes in the node load propagation sequence chain as the collaborative calculation benchmark. It extracts the active power, voltage and current change process in the chain, aligns the time of adjacent inner edge nodes and maps the voltage and current changes to obtain the collaborative calculation result of the power grid node load.

[0007] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this invention, a joint screening is performed based on the voltage change state, current change direction, active power change direction, and power factor change relationship within a continuous sampling window. Combined with the topological sequence of the energized feeder, the node change trajectories are continuously connected, forming a continuous change link with propagation correlation characteristics. By using a segmentation method that maintains consistency between the reactive power change direction and the power factor change relationship, the correlation identification capability during load change is enhanced. Furthermore, by establishing a cross-node sequential mapping relationship based on the active power change start time, current change time, and voltage change time, the true continuous change and local random disturbances during load propagation can be distinguished, reducing propagation confusion caused by cross-fluctuations in complex feeder environments and enhancing the load propagation path identification capability and the consistency of collaborative calculation results. Attached Figure Description

[0008] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a flowchart illustrating the acquisition process of S1 in this invention; Figure 3 This is a flowchart illustrating the acquisition process of S2 in this invention; Figure 4 This is a flowchart illustrating the acquisition process of S3 in this invention; Figure 5 This is a flowchart illustrating the acquisition process of S4 in this invention; Figure 6 This is a flowchart of the acquisition process for S5 of the present invention. Detailed Implementation

[0009] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0010] Please see Figure 1 This invention provides a technical solution: a load collaborative calculation method for power grid edge nodes, comprising the following steps: S1: Collect voltage, current, active power, reactive power, and switch status of distribution network edge nodes within a continuous sampling window to form initial node data. Calculate the power factor of each node within the sampling window based on voltage, current, and active power. Determine the energized feeder topology of the node based on the switch status. Treat the connection segment between adjacent nodes as feeder segments. Compare the data of the same initial node in the current sampling window with the data in the previous sampling window according to the sampling time sequence to obtain the voltage change status, current, active power, reactive power, and power factor change direction. Filter initial node data that continuously show voltage increases or decreases within consecutive adjacent sampling windows to obtain continuously changing voltage node data. Determine whether the active power change direction and current change direction are consistent in the continuously changing voltage node data. Connect the continuously changing voltage node data with consistent change directions and located in the same feeder segment according to the energized feeder topology order. Retain the reactive power change direction, power factor change direction, and switch status to form a node load propagation record sequence. S2: Based on the voltage continuously changing node data in the node load propagation record sequence, read the reactive power change direction and power factor change direction of the same voltage continuously changing node data in adjacent sampling windows one by one, determine the same or opposite change relationship, connect the voltage continuously changing node data with consistent voltage change state and consistent change relationship in time, and do not connect the same segment if the change relationship is inconsistent. Connect the voltage continuously changing node data after time continuity according to the topology of the live feeder to form a continuous change segment composed of segment node data, and obtain the node load change segment set; S3: Extract the feeder segment to which the segment node data in each continuous change segment in the set of node load change segments belongs. Arrange the segment node data in sequence according to the topology of the energized feeder. Compare the active power change direction of the segment node data with adjacent topologies. Group the segment node data with the same active power change direction and continuous time connection into the same coordination interval. Divide the boundary of the coordination interval according to the active power change direction of the segment node data on both sides of the connection position of adjacent feeder segments. Remove the segment node data with increased reactive power and voltage change state of voltage decrease, or with decreased reactive power and voltage change state of voltage increase. Connect the removed segment node data in each coordination interval in sequence according to the topology of the energized feeder to form a set of node coordination calculation intervals from the coordination node data. S4: Based on the collaborative node data within each collaborative interval in the node collaborative calculation interval set, read the sampling time of the first occurrence of active power change, voltage change, and current change in the adjacent sampling window change record of the collaborative node data to obtain the active power change start time, voltage change time, and current change time. Arrange the collaborative node data within the same collaborative interval according to the topology order of the live feeder. Compare the chronological relationship between the current change time and voltage change time of topologically adjacent collaborative node data. Determine the collaborative node data whose current change time is earlier than the voltage change time and whose active power change direction is consistent with the topology order of the live feeder as the propagation start node data. Starting from the propagation start node data, connect the collaborative node data within the same collaborative interval according to the chronological order of active power change start time, current change time, and voltage change time to form an intra-chain node data group, and obtain the node load propagation sequence chain. S5: Using the propagation start node data and the connection order of the inner edge node data in the node load propagation sequence chain as the basis for collaborative calculation, the active power change process, voltage change process, and current change process of the inner node data are extracted. The active power change process is arranged according to the connection order of the inner node data, and the time alignment between the inner edge node data of the adjacent chain is completed according to the start time of the active power change. The cross-node active power continuous change relationship is generated. According to the voltage change time and current change time, the voltage change process and current change process are mapped to the cross-node active power continuous change relationship. Combined with the topology order of the live feeder, the collaborative calculation result of the grid node load is obtained.

[0011] Please see Figure 2 The steps to obtain S1 are as follows: S11: Collect voltage, current, active power, reactive power and switch status of the edge nodes of the distribution network within the continuous sampling window. Arrange the voltage value, current value, active power value, reactive power value and switch status value according to the node identifier and sampling time sequence. Divide the active power value by the product of the voltage value and the current value to obtain the node power factor value. The system collects voltage, current, active power, reactive power, and switch status data from edge nodes of the distribution network within a continuous sampling window. It employs intelligent distribution terminals deployed at the network edge, combined with high-precision voltage transformers (PTs) and current transformers (CTs) for physical quantity acquisition. The terminal utilizes a built-in high-speed analog-to-digital converter chip (such as a 24-bit precision ADS131 series synchronous sampling chip) to extract analog signals from edge nodes at equal intervals within a continuous sampling time span. Each edge node achieves network-wide time alignment through a built-in GPS / BeiDou high-precision time synchronization module or the IEEE 1588 synchronous clock protocol, ensuring that the absolute error of the sampling timestamp across the entire network is less than 10μs. The system retrieves preset sampling frequency parameters. At 4000Hz, obtain a set of voltage observations. current observation set Active power mapping value set Reactive power mapping value set and the set of switch binary status words obtained using the digital input (DI) port. ,in The value is set to 200 sampling points, corresponding to a single sampling window width of 50ms. The node's physical address encoding is used as the first index dimension, and the timestamp ordinal number of the sampling point is used as the second index dimension. The extracted... , , , , Each component is written into the storage unit of a doubly linked circular list, forming a multidimensional heterogeneous data matrix. For example, for the component numbered... edge nodes, read their position At any given moment, the voltage is 225.42V, the current is 12.56A, the active power is 2.72kW, the reactive power is 0.51kVar, and the switch state is 1. These values ​​are linearly arranged in a floating-point data format in a memory buffer. The floating-point arithmetic unit is then retrieved, and the values ​​are processed for the same time interval. value, Value and The value is subjected to algebraic operations, and the calculation process follows the power factor conversion logic, specifically by calling the multiplication operator to convert the voltage value. With current value Perform a product operation to obtain the intermediate value of apparent power. (Unit is VA), when and At that time, the calculated median apparent power was 2831.2752 VA. The division operator was then used to calculate the active power value. After converting to watts, the value is used as the dividend, and the median apparent power is used as the divisor for proportional calculations. That is, at 2720W, execute The division operation yields a dimensionless value of 0.9607, which is used as the node power factor value corresponding to the sampling point. The calculation operation is then iterated over all points within the sampling window to output the complete power factor time series, thus obtaining the node power factor value.

[0012] S12: Based on the node power factor, call the switch status value and node identifier to determine whether the nodes at both ends of the closed switch belong to the same live feeder path, and whether the nodes at both ends of the open switch belong to different live feeder paths. Mark the connection segment of adjacent nodes in the topology as a feeder segment and generate the feeder segment sequence number value. Based on the node power factor, the switch status value and node identifier are called to search the binary status word of all nodes starting from the first memory address. When the node is found... With neighboring nodes When the switch state value between the two nodes is 1, the logic gate output is true, indicating that there is a closed electrical path between the two nodes. memory address pointer and The memory address pointers are linked to the same logical topology cluster and classified as the same live feeder path. If the retrieval is... and When the switch state value between them is 0, the logic gate output is considered false, representing electrical isolation. In the topology mapping table, the two are divided into independent branches that do not communicate with each other. If an extreme isolation state occurs where all switches in the entire network are open, each node is assigned to an independent single-node network and divided into different live feeder path branches. By traversing the switch state words between all nodes, a complete directed graph adjacency matrix is ​​constructed. In the adjacency matrix, the edge definition between two adjacent nodes with a direct logical connection is defined as a feeder segment. These edges are uniquely encoded using the depth-first search algorithm (DFS) and assigned segment numbers. For example, the first segment of the main line is marked as... Mark the second paragraph as Similarly, the storage width of the segment sequence number value is set to 16 bits to ensure that it can cover all physical connection segments within the area. When the node... Located on the substation outlet side and When located on the first segmented switch downstream, the node identifiers of both are concatenated to form a unique feature string "ID01-ID02", which is then mapped to a decimal integer range. Within, the allocation segment number is 101. If subsequent... and If a connection is established, the sequence number is assigned as 102. This creates a mapping table between the segment sequence number and the physical connection of the node in the topology database, generating the feeder segment sequence number value.

[0013] S13: Based on the feeder section number, call the voltage, current, active power, reactive power, and node power factor values ​​in the current and previous sampling windows of the same node, compare the increase / decrease signs of adjacent sampling window values, filter nodes with continuous voltage changes in the same direction, determine nodes whose active power change direction is consistent with the current change direction, connect them according to the topology of the energized feeder, and obtain the node load propagation record sequence. Based on the feeder section number, the voltage, current, active power, reactive power, and node power factor values ​​within the current and previous sampling windows of the same node are retrieved. Data differential processing is then performed, and the current sampling window... The mean within the previous sampling window Subtract the mean from the mean of the range, let... With the average voltage of the current window at 226.5V and the average voltage of the previous window at 224.2V, execute... The difference is obtained as +2.3V. The sign bit of this difference is extracted as positive, indicating a voltage increase trend. A voltage increase state word is then defined. The voltage is 1, indicating a voltage drop status word. A value of 0 indicates that the voltage remains unchanged when the absolute value of the fluctuation is below the PT accuracy error range. If the sign bit of the voltage difference remains positive for three consecutive sampling windows, the node is marked as a node with continuous voltage change in the same direction. If the voltage difference of a node is detected to alternate between positive and negative and the reactive power change rate is lower than the noise reference, the continuity of the voltage change direction in the monitoring area is determined to be inconsistent, indicating that the node is affected by random disturbances rather than trend load changes. At this time, the characteristic marker bit of the node is cleared, it is removed from the current load tracking chain, the tracking of its current sequence is terminated, the current judgment loop is exited, and the active power change of the node that satisfies the continuous voltage change in the same direction is called. With change in current ,when The change from 1.5kW to 1.8kW is +0.3kW. When the voltage changes from 6.8A to 8.2A by +1.4A, the sign bit comparison operator determines that both the increase and decrease directions are positive, satisfying the direction consistency criterion. At this point, the feeder segment number table is retrieved to find... , , Belonging to consecutive physical segments numbered 101 and 102, and all three satisfying the conditions of continuous voltage change in the same direction and active power and current direction being consistent, the feature vectors of voltage, current, power factor, and switching state of each node are extracted sequentially according to the downstream index order of the energized feeder topology. These vectors are encapsulated in a topological linked list arrangement, and connected end-to-end to form a dynamic data stream sequence. In this sequence, each node's data block contains its segment number and the current timestamp offset. For example, the formed sequence format is as follows: This sequence is then stored in the load dynamic tracking buffer to obtain the node load propagation record sequence.

[0014] Please see Figure 3 The steps to obtain S2 are as follows: S21: Based on the continuous voltage change node data in the node load propagation record sequence, read the reactive power change direction and power factor change direction in adjacent sampling windows of the same node one by one, compare whether the signs of the two directions are the same, record the same signs as the same direction mark, and record the opposite signs as the opposite direction mark to obtain the change relationship mark value; Based on the continuously changing voltage node data in the node load propagation record sequence, the node is located by calling the memory address offset. Extract the current sampling window from snapshots of data recorded within a continuous sampling window. reactive power mapping value For 0.55kVar and the previous sampling window reactive power mapping value The value is 0.51kVar, so the subtraction operator is called to perform the operation. Get the difference The reactive power fluctuation sensitivity baseline is set to +0.04 kVar, and since 0.04 is greater than 0.01, the change is deemed valid and its sign bit is extracted as positive, indicating a positive reactive power change direction. Simultaneously, the power factor sequence is read from the same data record snapshot to obtain... The power factor at time t is 0.9615 and The power factor at time t is 0.9607. Calculate the algebraic difference between the two values. The obtained value is +0.0008. The power factor change threshold is set to 0.0001. The comparator determines that 0.0008 is within the valid range greater than 0.0001. The sign bit of this difference is extracted as positive, indicating a positive power factor change direction. Then, a logic judgment function is called to check the sign values ​​of the reactive power change direction and the power factor change direction. When both sign values ​​are positive or both are negative, the logic gate outputs a high-level signal, marking the corresponding time slice of that node as a same-direction flag. If the reactive power change is +0.04 kVar and the power factor change is -0.0005, the logic gate outputs a low-level signal due to the different sign bits, marking the time slice as a reverse-direction flag. To quantify this correlation, a formula for calculating the correlation flag value with engineering verification significance is introduced: ; in, Represents a node At sampling time The change relationship is indicated by dimensionless values. This is a sign extraction function that extracts the sign of a variable and converts it to +1 (positive), -1 (negative), or 0 (no change). The extracted reactive power change (unit: kVar). The extracted change in power factor (dimensionless), when and When any output is 0, indicating a stationary state with indistinct fluctuations and extreme values, a forced definition is required. The value is 0.5. A result of 1 indicates the node is in the same direction, meaning its reactive power compensation characteristics are synchronized with the load trend. A result of 0 indicates the node is in the opposite direction. This is used to determine the node's direction. , Perform iterative calculations on historical data within all sampling windows to establish a relational feature vector corresponding to the sampling time sequence, and obtain the change relational indicator value.

[0015] S22: Call the change relationship flag value and voltage change status, determine whether the voltage change status in adjacent sampling windows is consistent, and determine whether the change relationship flag is consistent. Connect consistent nodes in the order of sampling time, stop connecting inconsistent nodes, and generate time connection index value. The node is retrieved from the distributed data cache by invoking the change relationship flag value and voltage change status. The feature index table is used to read the node's position in the sampling window sequence. to The voltage change status word (referenced from the sign bit judgment result defined in S13), if in The voltage difference obtained in the window is +2.3V, and its status bit is recorded as a value of 1, indicating an increase. The voltage difference obtained in the window is +1.8V, and the status bit is also 1. At this time, it is determined that the voltage change state is continuous within adjacent time steps. Next, the value calculated in S21 is retrieved. The change relationship indicator value within the corresponding time interval, if to If the segment's change relationship flag value remains constant at 1, it indicates that the physical load characteristics of that node maintain a stable unidirectional evolution during energy fluctuations. In this case, the pointer connection operator is invoked to... The end address of the data block and The starting addresses of the data blocks are logically linked, allowing the two sets of time-series data to enter the same time-continuous index chain. If during traversal to... When the voltage change state changes from increasing to decreasing (status bit changes from 1 to 0) during windowing, or the change relationship flag value jumps from 1 to 0, it is determined as a characteristic interruption, indicating that the load process has abruptly changed or ended. The logic determiner immediately triggers a cutoff action, stopping at... The location segment is extended, at this point targeting the node. , and put it in to This time span that satisfies double consistency is defined as a valid time segment. A binary index consisting of the start sampling number 1 and the end sampling number 2 is generated, and a hexadecimal time concatenation index code is assigned to it. For nodes located in the same feeder section If it is in to If the voltage continuously decreases and the change relationship flag value remains constant at 0, then another set of time connection index values ​​is generated. By performing a linear scan of all edge nodes along the sampling window dimension, the starting and ending boundaries of each node's features that remain stable are identified, the logically continuous intervals of each node on the time axis are recorded, and a time connection index value is generated.

[0016] S23: Based on the time connection index value, call the topology sequence of the energized feeder and the data of the continuously changing voltage nodes, determine the topological adjacency relationship of the time-continuous nodes in the feeder section, connect the topologically adjacent nodes in sequence into continuously changing segments, collect the segment node data according to the segment identifier, and obtain the node load change segment set. Based on the time connection index value, the topology sequence of the energized feeder and the data of nodes with continuously changing voltage are retrieved. The feeder segment number mapping table generated in S12 above is called up, and the feeder segment with sequence number 101 is searched. The physically adjacent nodes connected to this segment are identified as follows: and Then read the node from the memory index area. The time-connected index value shows its valid time range as sampling window ordinal 1 to 2, and the read node... The time connection index value shows its valid time range as sampling window ordinal 1 to 4. The set intersection operator is called to determine the common overlap interval of the two nodes in the time domain as sampling window ordinal 1 to 2. If the intersection operation result is empty, it means that the evolution processes of the two nodes do not overlap in the time domain, and they are determined to be different load fluctuation events. Therefore, this combination is discarded and no connection segment is generated. When the intersection is valid, the live feeder topology sequence table is then searched to confirm that it is in segment 101. upstream node and For downstream nodes, a logical AND gate is used to determine whether two nodes are topologically adjacent in space and undergoing the same change process in time. Node data objects that meet these conditions are pushed into a pre-allocated linear array and defined as continuously changing segments. And assign a unique identifier to the fragment; if the search continues and finds... Variation characteristics under the same sampling window ordinal number and If the consistency criterion is not met, stop expanding the members of the current segment and assign all segments to it. The node data under the identifier is reordered according to the flow order of the feeder topology. Each data entry contains the node ID, the section number, the current voltage, current, active power, and reactive power change gradient values. These spatiotemporally consistent data packets are encapsulated into independent data objects and stored in the fragment set area of ​​the node load change database. By recursively performing searches and matches for all feeder sections in the entire network, the discrete node sampling data is transformed into fragmented data clusters with topological logical association and time series attributes, resulting in the node load change fragment set.

[0017] Please see Figure 4 The steps to obtain S3 are as follows: S31: Extract the feeder segment to which the segment node data in each continuous change segment in the node load change segment set belongs, call the segment node identifier, feeder segment identifier and live feeder topology order, compare the sorting number of the segment node in adjacent feeder segments, arrange the segment node data according to the sorting number, and generate the segment topology sequence number value. Extract the feeder segment to which the segment node data belongs within each continuously changing segment in the node load change segment set, and call the segment identifiers pre-stored in memory. Locate the corresponding fragment node object list and read the node identifiers contained therein. , And their respective associated feeder section identifiers 101, retrieve the energized feeder topology sequence information corresponding to section number 101 from the topology configuration table, extract the spatial sorting weight factor of the nodes, and set the benchmark weight on the substation outlet side as... By sequentially adding the segment length parameters (in meters) provided by the Geographic Information System (GIS) along the direction of the branch road, the desired result can be obtained. The sort number is 1. The sorting number is 2. If there are parallel nodes with the same segment length, the value of the node's physical MAC address (extracted from the device information database using the SNMP protocol) is used as a secondary sorting benchmark to ensure sorting uniqueness. A value comparison operator is then called to perform ascending logical verification on the sorting numbers of adjacent nodes. When execution... When the judgment action output is true, The data block address is stored at the beginning of the linear list. The data block address is stored immediately afterward; if a node is detected... If the sort number is 3 and its segment identifier is 102, then continue to append to the tail in the same sequence. By traversing all nodes in the entire segment set, perform a quick sort operation based on the sort number to make the node data flow in memory completely consistent with the geographical flow of the power grid in physical logic. Encapsulate the topology feature code at the head and tail of the sequence. For example, for a segment containing three nodes, the generated feature code is "SEQ-101-01-02-03", which is used as the unique arrangement certificate of the segment in the topology dimension and written into the extended attribute field of the node to generate the segment topology sequence number value.

[0018] S32: Based on the segment topology index value, call the node data of adjacent segments in the topology, the direction of active power change and the time connection index value, compare the direction signs of adjacent nodes with the time connection flag, classify nodes with the same direction and continuous time into the same cooperative interval, divide the boundary according to the change position of the direction sign, and obtain the cooperative boundary position value. Based on the segment topology index value, the system retrieves the node data of adjacent topology segments, the active power change direction and time connection index value, and uses pointer addressing to retrieve them sequentially. and The active power change characteristic bit is retrieved from the record in S13. Change in active power as well as Change in active power If both signs are positive, perform a sign bit XOR operation. If the result is 0, the directions are considered to be consistent. Then, retrieve the time connection index value generated in S22. and Comparing the sampling window coverage areas of the two, if Coverage sampling windows 1 to 2 and Covering sampling windows 1 to 4, calculate the number of overlapping sampling points between two time intervals. The criterion is that the number of overlapping points must be greater than or equal to 2. When this condition is met, the two intervals are considered to have continuity in the time domain, meaning they satisfy both directional and temporal continuity. and Classified to number Within the collaborative interval storage area, if the time is completely discontinuous, the interval division is interrupted and a new interval is created to continue the comparison. and The characteristics, if obtained The change in active power is Its sign bit is negative, and When the positive sign bit is XORed with a result of 1, the direction is reversed. At this point, the logic unit issues a truncation instruction. and A boundary identifier is inserted at the topological connection point. The boundary identifier is composed of the preceding node ID, the following node ID, and the segment number, such as "ID02-ID03-102". The offset of the boundary in the topological linked list is calculated and recorded as the boundary position index 102.1. This is used as the segmentation benchmark for collaborative computing to obtain the collaborative boundary position value.

[0019] S33: Based on the collaborative boundary position value, call the segment node data, reactive power change direction, voltage change state and segment topology number value, determine the nodes where reactive power increases and voltage decreases, and the nodes where reactive power decreases and voltage increases, connect the remaining nodes in topological order to obtain the node collaborative calculation interval set. Based on the cooperative boundary location value, the segment node data, reactive power change direction, voltage change state, and segment topology number value are retrieved, within the range defined by the cooperative boundary. Perform node attribute filtering operations within the interval and call... The sign of reactive power change is positive, indicating an increase in reactive power, and the voltage change status word is 1, indicating a voltage increase (consistent with load feedback or compensation characteristics within the station). Retrieve The sign of reactive power change is positive, indicating an increase in reactive power, and the voltage change status word is 0, indicating a decrease in voltage. A node removal discrimination function is introduced to identify abnormal or irrelevant load fluctuations, which is of engineering significance. ; in, This is the reactive power change direction function (referenced from S21). When the value is greater than 0, the output is 1 (increases); when the value is less than 0, the output is 0 (decreases). The binary value represents the voltage change state (referenced from S13), where 1 represents an increase and 0 represents a decrease. When the value is null, the function is defined to output 0 to indicate that the node is retained to maintain topology integrity. When the logical operation function... An output of 1 indicates that the node fluctuation is a local, independent load disturbance rather than a systemic propagation event, and a removal action is performed. For Its input vector is If the result of substituting into the function is 0, the data for that node is retained; for Its input vector is Substituting the result into the function yields 1, representing the independent characteristic of the local voltage drop caused by the increase in inductive reactive power. Immediately remove the pointer reference of this node from the data linked list of the current cooperative interval, releasing the memory space it occupies. After completing the traversal and filtering of all nodes in the interval, call the linked list reconstruction operator to re-establish bidirectional links between the remaining valid nodes according to the segment topology sequence number in S31, ensuring that the logical addresses between each node are continuous and without breaks. Encapsulate the processed cooperative interval data into a comprehensive dataset containing topology weights, timing features, and power vectors, and store it in the cooperative calculation result queue to obtain the node cooperative calculation interval set.

[0020] Please see Figure 5 The steps to obtain S4 are as follows: S41: Based on the collaborative node data in each collaborative interval of the node collaborative computing interval set, read the sampling time of the first occurrence of active power change, voltage change and current change in the change record of adjacent sampling windows, record the three types of sampling time according to the node identifier, and obtain the change time series value; Based on the collaborative node data within each collaborative interval of the node collaborative computing interval set, retrieve the memory pointer pointing to the collaborative interval. Starting from the address, iteratively search for records within that range. to The multidimensional temporal characteristics are retrieved by calling the sampling frequency parameter set in S11. Calculate the time step of a single sampling point for The time interval is 0.25 milliseconds. The system retrieves the node's rated power data (from the distribution network resource ledger database) and sets the active power transition threshold. This value is obtained by multiplying the rated power by a sensitivity coefficient of 0.02. If the node's rated power is 10kW, then... The calculated value is 0.2kW. Starting from the first point in the sampling sequence, the subtraction operation is performed between the current active power value and the active power value of the previous sampling point. If the difference is less than 0.2kW throughout the entire sampling window... If no valid fluctuation occurs, the record will be exited when the node... When the active power jumps from 2.72kW to 3.05kW, the calculated difference is 0.33kW, and the numerical comparator is called to determine the difference. Immediately extract the ordinal number of the current sampling point. ; implement The multiplication operation yields 12.5 milliseconds, which is then stored in the active power change start time field. Then, the voltage change discrimination threshold is called. The calculation process is as follows: the rated voltage of 220V is multiplied by the fluctuation coefficient of 0.005 to obtain 1.1V. When the voltage changes from 225.4V to 226.8V, the sampling point number at this time is recorded as 52, and the process is executed. Calculate the voltage change time The threshold for judging current changes is set to 13.0 milliseconds. The current was 0.5A. When the detected current jumped from 12.56A to 13.42A, the difference of 0.86A was greater than 0.5A, so the sampling point number was recorded as 49, and execution was performed. The time of current change was calculated. The time is 12.25 milliseconds, which is the above-mentioned target The extracted 12.5 ms, 13.0 ms, and 12.25 ms values ​​were encapsulated into time vectors, and then... Perform the same difference extraction and threshold comparison operations to obtain The time vectors are 13.25 milliseconds, 14.0 milliseconds, and 13.5 milliseconds. They are stored in a temporary feature matrix according to the mapping relationship between node identifiers and sampling times to obtain the changing time series values. The sampling time for the first occurrence of active power change, voltage change, and current change in the change records of adjacent sampling windows is determined using the following formula: ; Detailed explanation of the formula and its calculation derivation: The total number of collaborative nodes is obtained by counting the number of nodes divided from the collaborative computing interval set, and then determined by the edge gateway data center. 2. The start time of active power change, voltage change time, and current change time of each node are all monitored and obtained through a high-precision miniature synchronous phasor measurement device at the edge node of the distribution network. The active power change start time is extracted from the synchronous sampling of the cooperating node numbered 1 via a hardware clock topology network. The voltage change time is 10.020 seconds. The current change time is 10.050 seconds. The active power change start time is 10.030 seconds. The coordinating node numbered 2, through a network with the same sampling clock, synchronously extracts this information. The voltage change time is 10.025 seconds. The current change time is 10.055 seconds. The time difference is 10.032 seconds. This time difference is determined by the physical characteristics of the distribution network line impedance and load disturbance propagation, and exhibits dynamic oscillations with the intensity of network fluctuations. The actual values ​​obtained through monitoring are substituted into the formula for calculation and derivation: The first step is to calculate the cumulative sum of all time difference terms for the collaborative node numbered 1. The formula is as follows: ; Substituting the corresponding values ​​and expanding, we get: ; Using basic arithmetic operations, evaluate the absolute value term and the square term respectively: ; The second step is to perform a common term operation on the collaborative node numbered 2, using the following formula: ; Substituting the corresponding values ​​and expanding, we get: ; Evaluation of the corresponding calculation terms: ; The third step is to sum the values ​​obtained from each node and calculate the average value. The formula is as follows: ; The fourth step is to perform a square root operation on the mean term to obtain the final time-series characteristic value of the first change of the cooperative node. The formula is as follows: ; The results show that the spatiotemporal dispersion of electrical physical quantity fluctuations at each edge node within the current cooperative interval is 0.0268235 seconds. Its physical significance lies in quantitatively assessing the inherent delay consistency of load impact transmission between different nodes. The smaller the eigenvalue, the more concentrated the electrical quantity response caused by load fluctuation propagation. Based on this eigenvalue, the order of current and voltage changes between adjacent nodes in the topology can be effectively corrected, thereby accurately determining the propagation starting node and ultimately outputting a highly reliable node load propagation sequence chain.

[0021] in, The time-series characteristic value of the first change of the coordinating node, represented in seconds. This represents the total number of cooperating nodes contained within the current cooperation interval. This represents the cyclically incrementing letter identifier used to iterate through each cooperative node within the current cooperative interval. Represents the first term within the current cooperative interval. The sampling time, measured in seconds, is when the first change in active power occurs at each coordinating node. Represents the first term within the current cooperative interval. The sampling time, measured in seconds, is when the voltage change first occurs at each coordinating node. Represents the first term within the current cooperative interval. The sampling time in seconds is the first time when the current changes at each coordinating node. The absolute value symbol represents the positive conversion of the internal time difference term, the summation symbol represents the cyclic accumulation of the calculation terms of all nodes, the square root symbol represents the square root operation of the overall mean, and the superscript 2 represents the square calculation of the difference between the voltage and current sampling time.

[0022] S42: Call the change time series value and the topology order of the energized feeder, arrange the data of the cooperative nodes in the same cooperative interval, compare the current change time and voltage change time of adjacent nodes in the topology, determine the nodes whose current change time is earlier than the voltage change time and whose active power change direction is consistent with the topology order, and obtain the propagation start point flag value. Retrieve the changing time series values ​​and the topology of the energized feeder from memory. The corresponding adjacency matrix identifies the nodes. and The feeder section 101 presents an upstream and downstream connection relationship, in which The topological sort weight is 1. The sorting weight is 2, read The time vector components are used to perform the current change time operation by calling the subtraction operator. With voltage change time The difference operation (unit: ms), i.e. The milliseconds are used to call a logic checker to detect the negative sign of the difference. If the result is less than 0, it is determined that the current response precedes the voltage response, which is consistent with the physical process of grid fluctuations caused by load changes. A logic flag is then set. If the difference is set to 1, it indicates that the current change is caused by passive voltage fluctuation. The logic flag will output 0 and the process will exit. Then, the data recorded in S13 will be retrieved. The direction of active power change, if the direction sign is positive and the adjacency matrix shows the topological flow direction from... point to If the energy flow direction is consistent with the topological growth direction, then the logical flag is set. Set to 1, for implement and The logical AND operation is performed, and when the result is 1, the node is marked in memory as... (Propagation starting point), if targeting downstream nodes When performing the same operation, the time of current change was observed. If the current change time is 13.5 milliseconds, which is significantly later than the current change time of the upstream node and does not meet the topological constraint of the first position of the interval, the logic output is 0. By searching for the node identifier that meets the double consistency criterion and is at the front of the topology in the entire cooperative interval, the starting position with the characteristics of the load disturbance source is located, and the propagation starting point feature code is written into the attribute field of the node to obtain the propagation starting point flag value.

[0023] S43: For the propagation start point flag value, call the change time series value and the topology order of the energized feeder, take the propagation start node data as the starting point, and connect the cooperative node data in the same cooperative interval according to the active power change start time, current change time and voltage change time to obtain the node load propagation sequence chain; For the propagation start point flag value, the changing time series value and the topology order of the charged feeder are called to mark it as... of The data block serves as the root node of the linked list, initiating a topological path traversal and retrieving... Internal time parameters , , Perform a minimum ordinal search on the timeline to determine the starting point of changes within the node as 12.25ms, and then search for neighboring nodes downstream in the topology. ,extract Corresponding time parameters , , Call the comparator to The start time of active power change and The magnitude of the active power change is determined by the start time, and then executed. The subtraction result was checked, and the difference was +0.75ms. The result was determined... The change occurred Subsequently, following the unidirectional evolution pattern on the timeline, representing the downstream propagation of load fluctuations, if the timestamps are the same, a minimal delay is allocated by default according to the topological order to maintain the causal chain. The memory link instruction is then invoked to... The output pointer points to Given the input address, construct a logical connection edge. If there are other nodes within the cooperative interval... Then continue the comparison. and The sequence of the three types of time parameters is determined to ensure that the timestamp of each subsequent node is greater than that of the preceding node. If a node is found to have a reversed time parameter, it is determined that the node is affected by another independent wave source, and the automatic connection of the branch is stopped. The program then searches for the next node that satisfies the monotonically increasing time rule under the topological constraints until all cooperative nodes in the interval have been traversed. According to the sequential response trajectory of active power change, current change and voltage change on the time axis, the discrete node data blocks are connected into a directed acyclic graph (DAG) sequence with strict causal logic. This sequence is serialized and output to the calculation buffer in the form of a linked list to obtain the node load propagation sequence chain.

[0024] Please see Figure 6 The steps to obtain S5 are as follows: S51: Using the connection order of the propagation start node data and the inner edge node data in the node load propagation sequence chain as the basis for collaborative calculation, extract the active power change process, voltage change process and current change process of the node data in the chain, arrange the active power change process according to the connection order, and generate the process sequence value in the chain. Using the connection order of the propagation start node data and the edge node data within the chain in the node load propagation sequence chain as the basis for collaborative computing, the data generated in S43 is retrieved. to A directed sequence list, retrieved from a memory buffer using direct memory access (DMA) instructions. The corresponding multi-dimensional time-series feature array is used to read the active power change process array. The array contains 200 consecutive sampling points covering a 50ms sampling window, and the values ​​cover the complete evolution trajectory from 2.72kW to 3.05kW, while also retrieving the voltage change process array. The value fluctuates between 225.42V and 226.82V; and the array of current change processes The value jumps between 12.56A and 13.42A, and then the pointer is located to the downstream node along the propagation chain. Extract the corresponding active power process array. Voltage process array and current process array In order to construct a multidimensional feature matrix, a classification and ranking operation based on node identifiers is performed. All time-series vectors are used as the first-level feature slices of the matrix. All time-series vectors are used as the second-level feature slices of the matrix. By reading the connection order values ​​1 and 2 from the linked list, the slices are sequentially... and Vertical stacking is performed, and at each sampling time step, the active power values ​​of the two nodes are mapped one-to-one. The power value at the 50th sampling point was 3.05 kW, while If the power value at the 50th sampling point is 2.85kW, then record the set of numerical pairs on the corresponding coordinate axis of the process sequence matrix. Repeat the reading and arranging action until all nodes in the chain are covered, ensuring that the three types of physical quantity processes of each node are precisely bound to their logical positions in the propagation sequence chain. Then, encapsulate the arranged data blocks into floating-point tensors and store them in the preprocessing register of the floating-point arithmetic accelerator to generate the process sequence values ​​in the chain.

[0025] S52: Based on the intra-chain process sequence value, call the active power change start time and the data of the intra-chain edge nodes of the topologically adjacent chain, compare the start time of the adjacent nodes, correct the arrangement position between nodes according to the time difference, complete the time alignment of the active power change process, and generate a cross-node power continuous trend. Based on the intra-chain process sequence values, the active power change start time and the data of the edge nodes in the adjacent intra-chain are retrieved from the time series values ​​recorded by S41. Start time of active power change: as well as Start time of active power change (The reference sampling frequency is 4000Hz). The subtraction operator is called to calculate the time offset, and the starting time difference between adjacent nodes is obtained. for Then retrieve the sampling frequency parameter and its corresponding fixed sampling period The number of translation pulses is obtained through division. One sampling point; The time alignment correction function is set as follows: ; in, The corrected time axis mapped to the reference time axis The node of the first Power values ​​at each sampling point (unit: kW). The original sampled array, The start time (in ms) for the active power change extracted from this node. The start time corresponding to the propagation start point flag value determined in S42 As a global alignment benchmark The sampling step size is 0.25ms. It serves as a sampling index and introduces a rounding symbol to comply with mathematical rigor. This is combined with a 0.5 offset to achieve rounding index positioning. For Substituting the numerical values, we obtain the translation index parameter as 3. Performing a memory address index translation operation, based on the formula logic, makes... The 4th point (index) in the active power process array )and The first point is aligned on the virtual synchronization time axis. If there is a gap at the beginning of the sequence after the shift, the static mean of that node before the shift (2.85kW) is used to fill it. If the tail exceeds the sampling window range, a truncation action is performed. The index value after the shift must meet the following requirements. To prevent memory overflow, the system compares the start times of all adjacent nodes and performs corresponding index shift correction on the data sequence of each node in the chain. This ensures that all nodes are on the same calculation step size basis at the physical instant when the active power changes, thereby eliminating the natural physical delay caused by load propagation in the distribution network. The aligned multi-node active power change curves are then overlaid and mapped to form a dynamic profile, generating a continuous trend of cross-node power.

[0026] S53: Based on the continuous trend of cross-node power, the voltage change time, current change time and the topology order of the energized feeder are called. The time positions of the voltage change process and current change process within the same node number are compared. The two types of processes are mapped to the continuous trend to obtain the load collaborative calculation results of the power grid nodes. Based on the continuous power trend across nodes, the voltage change time, current change time, and the order of the energized feeder topology are used to retrieve the change time series of S41. Voltage change time With time of current change Simultaneously, the active power change time used as the alignment reference in S52 is retrieved. Calculate the relative time offset within the node. For the voltage process, the offset step size is calculated as follows: One sampling point, for the current process; Calculate the offset step size as Each sampling point, using the same index offset operator, will... The voltage change process array is shifted backward by 2 points, and its current change process array is shifted forward by 1 point, so that their relative positions on the time axis are logically synchronized with the already aligned active power continuous trend. Then, the current change process array is sequentially adjusted along the energized feeder topology. Perform the same mapping operation, when you get When the voltage change time is 14.0 ms, the total offset relative to the reference time of 12.5 ms is calculated as follows: Each sampling point is sampled, and its voltage sequence is mapped to the corresponding clock cycle of a continuous trend. By traversing all the indices in the propagation chain, the three-dimensional physical quantity change trajectory of each node is projected onto the same high-dimensional space with power perturbation as the coordinate origin. At each aligned sampling time point, the per-unit voltage amplitude (actual voltage / rated voltage 220V), current increase, and power fluctuation value of all nodes are solved by joint matrix calculation. If the per-unit voltage amplitude of all nodes in the entire chain at a certain moment is found to be within the range of the specified values, then the calculation is complete. If the current change gradient is greater than 0.05A / ms within the rated safe range (compliant with GB / T30137 standard), it is determined to be a valid load collaborative evolution feature. If it exceeds the limit, it is marked as invalid in the result and assigned an over-limit code that conforms to the dispatch system protocol (voltage over-limit is marked as 0x91, current gradient over-limit is marked as 0xB2). Finally, the time sequence characteristics, topological location, and collaborative evolution vector of all nodes are encapsulated into the final distribution network dynamic operation data archive to obtain the load collaborative calculation results of the power grid nodes.

[0027] A load collaborative computing system for power grid edge nodes, the system comprising: The sampling and identification module collects voltage, current, active power, reactive power and switch status within the continuous sampling window of the edge nodes to form initial node data, calculates the power factor and determines the topology of the energized feeder, compares the changes of adjacent sampling windows, filters the node data with continuous voltage changes and connects them according to feeder segments to form a node load propagation record sequence. The relationship segmentation module, based on the node load propagation record sequence, reads the reactive power and power factor change direction of nodes with continuously changing voltage, determines the same or opposite relationship, connects nodes with consistent state and relationship in time, and forms segment node data according to the topology to obtain a set of node load change segments. The interval screening module extracts the feeder segments of the node load change segment set, arranges the segment nodes according to the topology, divides the collaborative interval boundary by comparing the active power change direction, removes segment nodes where reactive power increases and voltage decreases or reactive power decreases and voltage increases, connects the remaining segment nodes as collaborative node data, and obtains the node collaborative calculation interval set. The timing source determination module, based on the node collaborative calculation interval set, reads the first sampling time of the active power, voltage and current changes of the collaborative node data, compares the current and voltage of adjacent collaborative nodes according to the feeder topology, determines the propagation start node data, connects the node data groups in the chain, and obtains the node load propagation sequence chain. The mapping calculation module uses the connection order of the propagation start node data and the data of the inner edge nodes in the node load propagation sequence chain as the basis for collaborative calculation. It extracts the active power, voltage and current change process in the chain, aligns the time of adjacent inner edge nodes and maps the voltage and current changes to obtain the collaborative calculation results of the power grid node load.

[0028] It is understandable that the above system and method have the same execution process and the same effect, so they will not be described again here.

[0029] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A load coordination calculation method for power grid edge nodes, characterized in that, Includes the following steps: S1: Collect voltage, current, active power, reactive power and switch status within the continuous sampling window of the edge node to form initial node data, calculate the power factor and determine the topology of the energized feeder, compare the changes of adjacent sampling windows, filter the node data with continuous voltage changes and connect them according to feeder segments to form a node load propagation record sequence. S2: Based on the node load propagation record sequence, read the reactive power and power factor change direction of the nodes with continuously changing voltage, determine the same or opposite relationship, connect the nodes with consistent state and relationship in time, and form the node data segments according to the topology to obtain the node load change segment set. S3: Extract the feeder segments of the node load change segment set, arrange the segment nodes according to the topology, divide the collaborative interval boundary by comparing the active power change direction, remove the segment nodes where reactive power increases and voltage decreases or reactive power decreases and voltage increases, connect the remaining segment nodes to form collaborative node data, and obtain the node collaborative calculation interval set. S4: Based on the set of node collaborative calculation intervals, read the first sampling time of the active power, voltage, and current changes of the collaborative node data, compare the current and voltage of adjacent collaborative nodes according to the feeder topology, determine the propagation starting node data, connect the node data groups in the chain, and obtain the node load propagation sequence chain.

2. The load coordination calculation method for power grid edge nodes according to claim 1, characterized in that: The method further includes: S5: Using the connection order of the propagation start node data and the data of the inner edge nodes in the node load propagation sequence chain as the basis for collaborative calculation, extract the active power, voltage and current change process in the chain, align the time of adjacent inner edge nodes and map the voltage and current changes to obtain the collaborative calculation result of the power grid node load.

3. The load collaborative calculation method for power grid edge nodes according to claim 1, characterized in that: The steps to obtain S1 are as follows: S101: Collect voltage, current, active power, reactive power and switch status of the edge nodes of the distribution network within a continuous sampling window, arrange the sampled values ​​according to the node identifier and sampling time sequence, and divide the active power value by the product of the voltage value and the current value to obtain the node power factor value. S102: Based on the node power factor value, call the switch status value and node identifier to determine whether the nodes at both ends of the closed switch belong to the same live feeder path, and whether the nodes at both ends of the open switch belong to different live feeder paths. Mark the connection segment of the adjacent nodes in the topology as a feeder segment and generate the feeder segment sequence value. S103: Based on the feeder section number, retrieve the voltage, current, active power, reactive power, and node power factor values ​​from adjacent sampling windows of the same node. Compare the signs of increase and decrease of each value, filter nodes with continuous voltage changes in the same direction, determine nodes whose active power change direction is consistent with the current change direction, connect them according to the energized feeder topology, and obtain the node load propagation record sequence.

4. The load collaborative calculation method for power grid edge nodes according to claim 1, characterized in that: The steps to obtain S2 are as follows: S201: Based on the continuous voltage change node data in the node load propagation record sequence, read the reactive power change direction and power factor change direction in adjacent sampling windows of the same node one by one, compare the signs of the two directions, record the same signs as the same direction mark, and record the opposite signs as the opposite direction mark to obtain the change relationship mark value; S202: Call the change relationship flag value and voltage change status, determine whether the voltage change status and change relationship flag in adjacent sampling windows are consistently consistent, connect consistent nodes in the order of sampling time, stop connecting inconsistent nodes, and generate time connection index value. S203: Based on the time connection index value, call the topology sequence of the live feeder and the data of the continuously changing voltage nodes, determine the topological adjacency relationship between the time-continuous nodes, connect the topologically adjacent nodes in sequence into continuously changing segments, collect the segment node data according to the segment identifier, and obtain the node load change segment set.

5. The load collaborative calculation method for power grid edge nodes according to claim 1, characterized in that: The steps to obtain S3 are as follows: S301: Extract the feeder section to which the segment node data in each continuous change segment in the set of node load change segments belongs, call the segment node identifier, feeder section identifier and live feeder topology order, compare the segment node sort number, arrange the segment node data according to the sort number, and generate the section topology sequence number value. S302: Based on the segment topology sequence number value, call the node data of adjacent topology segments, the active power change direction and time connection index value, compare the direction sign and time connection flag, and classify nodes with the same sign and continuous time into the same cooperative interval to obtain the cooperative boundary position value. S303: Based on the collaborative boundary position value, call the segment node data, reactive power change direction, voltage change state and segment topology number value, determine the combined state of reactive power and voltage change, connect the reserved nodes in topological order, and obtain the set of node collaborative calculation intervals.

6. The load collaborative calculation method for power grid edge nodes according to claim 1, characterized in that: The steps to obtain S4 are as follows: S401: Based on the collaborative node data in each collaborative interval of the node collaborative computing interval set, read the change records of adjacent sampling windows, compare the first change positions of active power, voltage, and current, record the sampling time according to the node identifier, and obtain the change time series value; S402: Call the change time series value and the topology order of the energized feeder, arrange the data of the cooperating nodes in the same cooperative interval, compare the current change time and voltage change time of adjacent nodes in the topology, determine the nodes whose active power change direction is consistent with the topology order, and obtain the propagation start point flag value. S403: For the propagation start point flag value, call the change time series value and the topology order of the energized feeder, take the propagation start node data as the starting point, and connect the cooperative node data according to the start time of active power change, current change time, and voltage change time to obtain the node load propagation sequence chain.

7. The load coordination calculation method for power grid edge nodes according to claim 2, characterized in that: The steps to obtain S5 are as follows: S501: Using the connection order of the propagation start node data and the edge node data in the node load propagation sequence chain as the basis for collaborative calculation, extract the active power change process, voltage change process and current change process of the node data in the chain, arrange the active power change process according to the connection order, and generate the process sequence value in the chain. S502: Based on the process sequence value within the chain, call the start time of active power change and the data of adjacent chain edge nodes in the topology, compare the start time of adjacent nodes, adjust the node arrangement position according to the time difference, and generate a continuous trend of cross-node power. S503: Based on the continuous trend of cross-node power, the voltage change time, current change time and the topology order of the energized feeder are called. The time positions of the voltage change process and current change process in the same node sequence are compared. The two types of processes are mapped to the continuous trend to obtain the load collaborative calculation results of the power grid nodes.

8. A load collaborative computing system for power grid edge nodes, characterized in that, The system is used in the load collaborative calculation method for grid edge nodes as described in any one of claims 1-7, and the system comprises: The sampling and identification module collects voltage, current, active power, reactive power and switch status within the continuous sampling window of the edge nodes to form initial node data, calculates the power factor and determines the topology of the energized feeder, compares the changes of adjacent sampling windows, filters the node data with continuous voltage changes and connects them according to feeder segments to form a node load propagation record sequence. The relationship segmentation module, based on the node load propagation record sequence, reads the reactive power and power factor change directions of nodes with continuously changing voltage, determines the same or opposite relationship, connects nodes with consistent states and relationships in time, and forms segment node data according to the topology to obtain a set of node load change segments. The interval screening module extracts the feeder segments of the node load change segment set, arranges the segment nodes according to the topology, divides the collaborative interval boundary by comparing the active power change direction, removes segment nodes where reactive power increases and voltage decreases or reactive power decreases and voltage increases, connects the remaining segment nodes as collaborative node data, and obtains the node collaborative calculation interval set. The timing source determination module, based on the set of node collaborative calculation intervals, reads the first sampling time of the active power, voltage, and current changes of collaborative node data, compares the current and voltage of adjacent collaborative nodes according to the feeder topology, determines the propagation start node data, connects the node data groups in the chain, and obtains the node load propagation sequence chain. The mapping calculation module uses the connection order of the propagation start node data and the data of the inner edge nodes in the node load propagation sequence chain as the collaborative calculation benchmark. It extracts the active power, voltage and current change process in the chain, aligns the time of adjacent inner edge nodes and maps the voltage and current changes to obtain the collaborative calculation result of the power grid node load.

Images