An edge-computing-based power distribution network adaptive protection method and system

Abstract

The application relates to the technical field of power distribution network protection, and discloses a power distribution network adaptive protection method and system based on edge computing, which comprises the following steps: acquiring multi-source operation data through an edge terminal to sense the working condition and topology of a power distribution network, identifying abnormal features to locate the fault type and section, dynamically adjusting the adaptive protection triggering threshold and delay, executing distributed tripping isolation based on peer-to-peer communication and coordination of adjacent nodes, controlling switch action, and reconstructing the global operation structure. According to the application, multi-source data is acquired by relying on an edge terminal to sense the topology of a power distribution network, the fault section is accurately located by fusing transient and steady state features, the protection setting value is dynamically adjusted according to the operation mode and the state of a distributed power source, the distributed defense system is constructed based on the coordination of adjacent nodes, the real-time response and collaboration of protection are enhanced, the defects of traditional centralized protection response delay and setting value mismatch are overcome, and the power supply reliability and flexibility under high proportion of distributed power source access are improved.

Description

Technical Field

[0001] This invention relates to the field of distribution network protection technology, and specifically to a distribution network adaptive protection method and system based on edge computing. Background Technology

[0002] The field of distribution network protection technology involves methods related to power supply safety and operation control in the context of distributed energy access. With the continuous development of smart grids, the high proportion of distributed power sources, new energy grid integration, and diversified load access have transformed distribution networks into complex nonlinear systems. Traditional centralized protection schemes focus on the unified scheduling and calculation of the master station system. However, when faced with frequent changes in topology and the access of numerous nonlinear power electronic devices, they often suffer from increased response delays, weak environmental adaptability, and high dependence on communication bandwidth. Therefore, optimizing the real-time performance and accuracy of protection using a distributed architecture has become an important research direction in the current field of power system protection.

[0003] Chinese patent CN113437730B discloses an adaptive topology change distribution network protection method based on an edge gateway system. This scheme deploys an edge gateway system at each node of the distribution network, utilizes sensing changes in switch values ​​to trigger current request signals between nodes, and determines new topology relationships and fault sections through correlation analysis. However, this scheme relies heavily on frequent data interactions between nodes to reconstruct the topology, and its real-time response within a very short time is limited by the stability of the communication link. Furthermore, its protection criteria primarily focus on the correlation of current data, and it remains insufficient in comprehensively identifying and deeply fusing multi-source information under various complex abnormal conditions such as islanded operation and overload.

[0004] Chinese patent CN113325264B discloses a distribution network fault protection method based on an adaptive differential grounding algorithm. This scheme installs edge computing devices at feeder monitoring points and combines them with a central edge computing device to collect transient current signals, then uses an adaptive differential algorithm to determine grounding faults. However, this technical solution focuses on grounding faults in distribution networks, and has limitations in covering all operating conditions such as short circuits, overloads, and islanding. Its collaborative mechanism often adopts a master-slave centralized monitoring mode, and the potential for autonomous decision-making and collaborative linkage of distributed edge computing units at the node side has not been fully realized. When dealing with the dynamic adjustment requirements of protection strategies brought about by a high proportion of distributed energy grid connection, its flexibility shows room for improvement.

[0005] In summary, existing power distribution network protection technologies still need improvement in terms of edge-side intelligence, multi-condition comprehensive identification capabilities, and distributed collaborative response speed. Summary of the Invention

[0006] This invention aims to provide a distribution network adaptive protection method and system based on edge computing. By deploying edge computing units at key nodes and combining them with multi-source information fusion intelligent algorithms, it achieves rapid identification of complex operating conditions and adaptive adjustment of protection strategies, thereby improving the power supply reliability and operational flexibility of the distribution network in the scenario of new energy access.

[0007] The basic solution provided by this invention is: an adaptive protection method for distribution networks based on edge computing, comprising: S1: By acquiring multi-source operation data in real time through edge terminals deployed at key nodes of the distribution network, identifying feeder segment information, determining the current operating conditions and topological connection relationship of the distribution network, extracting state feature vectors that characterize the system operation features, and generating multi-source information perception results of the distribution network; feeder segment information includes current vector, node voltage waveform, power factor, and switch position status; S2: Call the multi-source information sensing results of the distribution network to identify the transient change characteristics and steady-state deviation characteristics of the power of each node, distinguish the manifestations of abnormal operating conditions, summarize the area range and severity of the fault, and generate fault type and section location identification results; abnormal operating conditions include short circuit faults, overload operation and islanding operation; S3: Based on the fault type and section location identification results, determine the validity of the protection criteria under the current topology, combine the preset protection setting adjustment algorithm, determine the adaptive protection trigger threshold and delay parameters that match the current operating conditions, and generate the adaptive protection strategy adjustment results; S4: Invoke the adaptive protection strategy to adjust the results, determine the collaborative protection needs of adjacent nodes based on the peer-to-peer communication protocol between edge terminals, identify the fault isolation boundary and load transfer path, select the distributed collaborative tripping command suitable for the current fault condition, and generate the collaborative linkage protection execution result.

[0008] This invention also provides an edge computing-based adaptive protection system for distribution networks to execute an edge computing-based adaptive protection method for distribution networks; the system includes: The multi-source information sensing module is used to acquire multi-source operating data in real time through edge terminals deployed at key nodes of the distribution network, identify feeder segment information, determine the current operating conditions and topological connection relationship of the distribution network, extract state feature vectors that characterize the system's operating features, and generate multi-source information sensing results for the distribution network; feeder segment information includes current vector, node voltage waveform, power factor, and switch position status; The fault location and identification module is used to call the multi-source information sensing results of the distribution network, identify the transient change characteristics and steady-state deviation characteristics of the power of each node, distinguish the manifestation of abnormal operating conditions, summarize the area and severity of the fault, and generate fault type and section location identification results; abnormal operating conditions include short circuit faults, overload operation and islanding operation; The strategy adaptive module is used to call the fault type and section location identification results, determine the validity of the protection criteria under the current topology, combine the preset protection setting adjustment algorithm, determine the adaptive protection trigger threshold and delay parameters that match the current operating conditions, and generate the adaptive protection strategy adjustment results. The distributed coordination module is used to call the adaptive protection strategy adjustment results, determine the coordination protection needs of adjacent nodes based on the peer-to-peer communication protocol between edge terminals, identify the fault isolation boundary and load transfer path, select the distributed coordination trip command suitable for the current fault condition, and generate the coordination linkage protection execution result. The reconfiguration module is used to call the execution results of the coordinated protection. Based on the operating characteristics of the switching equipment and the load transfer demand of the distribution network, it adjusts the operating sequence of each switch, controls the switch actuator to complete fault isolation and non-fault zone restoration, and generates dynamic optimization results for the adaptive protection of the distribution network.

[0009] The working principle and advantages of this invention are as follows: This invention deploys edge terminals at key nodes of the distribution network to acquire multi-source power data in real time, divides the topology change area and establishes multi-dimensional information perception relationships, combines transient and steady-state feature fusion algorithms to identify the manifestation of abnormal operating conditions, summarizes fault types and accurately locates fault sections, dynamically adjusts protection setting curves and defines adaptive judgment logic based on system operation mode and distributed power source access status, identifies the collaborative decision-making status and power transfer spatial deployment mode between adjacent nodes, determines distributed collaborative paths and completes spatial mapping of control commands, and constructs a distributed defense system with highly consistent power rhythm characteristics and protection action logic. This enhances the real-time response, judgment accuracy and control coordination in the protection process, significantly improves the power supply reliability and operational flexibility of the distribution network under the background of high proportion of distributed power source access, and effectively solves the problems of response delay and setting mismatch of traditional centralized protection under complex operating conditions. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating an edge computing-based adaptive protection method for distribution networks provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of an edge computing-based adaptive protection system for power distribution networks provided in an embodiment of the present invention. Detailed Implementation

[0011] The following detailed explanation illustrates the specific implementation methods: The basic implementation examples are as follows: Figure 1 As shown: An adaptive protection method for distribution networks based on edge computing, comprising: S1: By acquiring multi-source operation data in real time through edge terminals deployed at key nodes of the distribution network, identifying feeder segment information, determining the current operating conditions and topological connection relationship of the distribution network, extracting state feature vectors that characterize the system operation features, and generating multi-source information perception results of the distribution network; feeder segment information includes current vector, node voltage waveform, power factor, and switch position status; Specifically, S1 includes: S101: Acquire the raw electrical signals from each monitoring point in the distribution network, perform analog-to-digital conversion on the current and voltage signals through a high-frequency sampling circuit, extract the fundamental component, specific harmonic components, and aperiodic components from the sampling sequence, and call the preset fast Fourier transform algorithm to calculate the effective value and phase angle of each component to obtain the basic electrical characteristic dataset; as detailed below: Acquire the analog voltage and current signals of phases A, B, and C at monitoring points N1, N2, and N3 in the distribution network. For the high-frequency sampling circuit front end, extract the analog quantity processed by the low-pass filter. Configure the analog-to-digital converter to 16-bit resolution and set the sampling frequency to [value missing]. .

[0012] Extracting the period from a continuous analog signal The 256 discrete sampling points are written into a one-dimensional array to form the voltage sequence at the current moment. With current sequence , where index .

[0013] Extracting the measured extreme value range of analog voltage Linearly map it to in the analog-to-digital converter register. to Within the digital quantization space.

[0014] For the sampled sequence, after adding a Hanning window function to truncate the segment, substitute it into the radix-2 Fast Fourier Transform formula: Perform iterative multiplication and addition operations. Substitute the measured data sequence for the current period and extract the index from the frequency domain array. The fundamental component is obtained from the parameter unit at the location. The fundamental voltage amplitude of phase A at node N1 is measured and read. Phase angle A-phase fundamental current amplitude Phase angle .

[0015] For a specific subharmonic component, extract the frequency domain array. The unit value is used to obtain the amplitude of the third harmonic current. ,extract The unit value is used to obtain the fifth harmonic voltage amplitude. .

[0016] For the aperiodic component, the absolute values ​​of all 256 sampling points within a single period are extracted, and an arithmetic mean is calculated to obtain the overall mean. The root mean square value of the AC component extracted earlier is then subtracted from this mean to obtain the effective value of the DC bias component. .

[0017] The extracted harmonic amplitudes, phase angles, and DC bias parameters are stored in the multidimensional database fields according to the node number order to generate a basic power characteristic dataset.

[0018] S102: A basic feature dataset of power consumption for adaptive protection of distribution networks based on edge computing. This dataset calls the geospatial location indexes of each edge terminal to compare the current vector difference and voltage drop values ​​between adjacent monitoring points in real time. For branches whose current increment exceeds a preset activation threshold, topology connectivity verification is performed according to their current flow direction and phase relationship to obtain a dynamic topology connection matrix. Details are as follows: The hardware configuration information corresponding to each edge terminal T1 and T2 is retrieved from the read-only memory to read the geospatial location coordinate data. The coordinates of T1 are then extracted. With T2 coordinate point .

[0019] For physically adjacent terminals T1 and T2 in the topology connection, extract the A-phase current vector parameters of the two in-phase currents from the aforementioned dataset, and perform complex subtraction modulus calculation to obtain the current vector difference. Substitute the T1 parameters. With T2 parameters Seeking .

[0020] Extract the voltage phasors between the nodes and perform vector interpolation modulo operation to obtain the voltage drop value. .

[0021] Preset startup threshold The specific settings are based on the following: Extract the no-load to full-load switching record sequence once per minute within the historical 30-day period of the target power distribution line; sort the total of 43,200 transient impulse current sample points in ascending order of value; and extract the measured value at the 95th quantile index. Introducing a reliability coefficient The value of this coefficient is based on the predicted value of the maximum short-circuit current of the line. With rated current carrying capacity The ratio curve is obtained by looking up the table, and the current value is... Perform multiplication operations .

[0022] The instantaneous increment of the phase current in branch L1 was measured in real time. ,right and Perform a size comparison; the value is in Within the interval. For this branch L1 that meets the conditions, extract the phase angle parameters of the current at both ends and perform absolute difference calculation to obtain the phase angle difference. .

[0023] Perform a conditional judgment operation on this phase angle difference, because it is in Within the physical connectivity phase angle difference interval, it is determined to be in a connected state, and the number 1 is written into the corresponding row and column coordinate position in the connection matrix to generate a dynamic topology connection matrix.

[0024] S103: Based on the dynamic topology connection matrix of the distribution network adaptive protection based on edge computing, the output parameters of the grid-connected inverters at the distributed power source access points and the synchronous switch status information are extracted. The power flow direction change status and load balance within the distribution network are linked for judgment, identifying locations with reverse power flow or local power gaps. Their status labels are then encapsulated to obtain the multi-source information perception results of the distribution network; specifically as follows: According to the network packet parsing protocol, extract the grid-connected inverter operation message of distributed power source DG1 at access point N3. Read the active power field output within the packet data frame. reactive power field Extract the hardware status register value of the corresponding synchronous switch to 0x01.

[0025] Extract active power from the measuring points at the entrance of the corresponding area of ​​the distribution network. Extract the total power consumption of the entire area load. Perform algebraic summation of node power: Substituting the aforementioned measured parameters and performing the calculation yields... .

[0026] The polarity of the calculation result is determined, and the value is in the positive or negative range. The interval is marked with a power gap identifier of 1. An extraction operation is performed on the active power flow parameters fed back upstream from the DG1 grid connection point; the measured feedback value is... Because the value is in For each interval, the status register of that interval is overwritten with the "current flow reverse" encoding code.

[0027] The node number code "N3", switch status bit "0x01", power difference value "-100.0" and reverse encoding code are combined and encapsulated into a fixed-length data packet using a string concatenation function to obtain the multi-source information sensing results of the distribution network.

[0028] The advantage of this approach is that it establishes a numerical relationship between state logic quantities and physical analog quantities by directly performing algebraic addition and subtraction operations between the inverter's instantaneous output parameters and the power difference state of the topology nodes.

[0029] The results of multi-source information perception in the distribution network include current amplitude and phase sequence, voltage harmonic distortion rate, frequency offset, and distributed power source access point status. The results of fault type and section location identification include fault phase identifier, transient energy distribution characteristics, and topology branch correlation matrix. The results of adaptive protection strategy adjustment include setting correction coefficient, action time ladder sequence, and criterion logic switching identifier. The results of coordinated protection execution include circuit breaker closing and opening sequence, logic blocking signal, and islanding operation control command. The results of dynamic optimization of adaptive protection in the distribution network include switch action sequence record, fault isolation range map, and non-faulty section power transfer scheme.

[0030] S2: Call the multi-source information sensing results of the distribution network, identify the transient change characteristics and steady-state deviation characteristics of the power of each node, distinguish the manifestations of three abnormal operating conditions: short circuit fault, overload operation and islanding operation, summarize the area range and severity of the fault, and generate fault type and section location identification results.

[0031] Specifically, S2 includes: S201: The system invokes the multi-source information sensing results of the distribution network adaptive protection system based on edge computing. It sequentially retrieves the derivative rate of change and voltage sag depth of the current sequence in each zone. It performs continuous comparison of the differential current value and braking current value between adjacent edge terminals within the same feeder segment. If the differential current value remains within the preset internal fault determination interval, it is determined to be a fault state within the zone, and the fault initiation feature matrix is ​​obtained. Specifically: Retrieve the current sequence buffered in the data register With voltage sequence Regarding the sampling interval Perform the difference quotient operation between adjacent sampling points to calculate the rate of change of the derivative. Substitute the measured values ​​from two consecutive sampling points at the moment of the fault. and The derivative value at that moment is obtained as follows: .

[0032] Extracting the rated reference voltage parameters Extract the real-time effective value of the fault phase voltage. Substitute into the formula The calculation yielded the voltage dip depth value as follows: .

[0033] Extract the instantaneous current phasor at the edge terminal measurement points at both ends of feeder segment L2. and The differential current value is obtained by performing a complex addition modulo operation. Perform complex subtraction, find the modulus, and multiply by the coefficient. Obtain braking current value , respectively obtain , .

[0034] The braking equation presupposes the relationship between the internal decision intervals. Braking coefficient The experimental setup was as follows: A three-phase metallic short circuit was applied to the nodes outside the transformer area at the test site, and the maximum through current was recorded. The corresponding unbalanced differential current caused by transformer saturation is Perform division Superimposed error tolerance constant ,set up Fixed threshold value According to the specifications of the 5P20 class instrument transformer, the inherent error limit is set as follows: Substitute the aforementioned real-time calculation results into the relational expression and perform the subtraction operation: .

[0035] Compare the results Perform numerical comparison. In The interval is defined as the range in which the differential current value falls within the internal fault determination interval. The derivative, indentation depth, and determination flag (set to 1) are extracted and written into a matrix structure in column order to obtain the fault initiation feature matrix.

[0036] S202: Based on the fault initiation feature matrix of the distribution network adaptive protection system based on edge computing, extract the polarity relationship between the node index number and the internal electrical characteristics of all nodes with transient voltage waveform distortion, perform direction determination on the phase offset values ​​of the voltage zero-sequence component and the current zero-sequence component at the fault initiation time, retrieve the location of the fault point on the bus side or the line side, and mark the fault phase in sequence to obtain a multi-dimensional fault feature classification set; as detailed below: Extract the index number of the transient voltage waveform distortion node corresponding to the flag bit 1 in the matrix, and obtain the label N1. Retrieve the three-phase waveform buffer data corresponding to this node, and perform the summation and division of the three-phase quantities. The operation to obtain the zero-sequence instantaneous value: voltage zero-sequence Zero sequence current .

[0037] Perform another Fourier frequency domain transform on the obtained zero-order sequence to extract the timestamp. Voltage zero-sequence phasor at location zero-sequence phasor of current Extract the phase angle parameters of both and perform a subtraction operation to obtain the offset value: .

[0038] Set the positive direction boundary interval as The opposite direction interval is .Will Perform a size comparison with the interval endpoints. [ , Within the specified interval. According to the network coordinate system definition, output direction scalar code 1, representing the fault point pointing towards the bus side. Retrieve the effective value sequence of the three-phase voltage at this node, extracting those values ​​where the voltage drop exceeds the rated value. The string is merged using the corresponding identifier character (such as "A_PHASE").

[0039] As shown in Table 1, the zero-sequence amplitude, phase angle difference, and direction state values ​​of multiple nodes are extracted and stored in the database records to obtain a multi-dimensional fault feature classification set.

[0040] Table 1: Multidimensional Fault Feature Classification Data Table

[0041] S203: The multi-dimensional fault feature classification set for adaptive protection of the distribution network based on edge computing is invoked. Spatial correlation statistics are performed on the fault identification status reported by all edge terminals. Based on the continuity and temporal consistency of the distribution of similar fault statuses on the topology branches in the statistical sequence, the specific physical location and affected area of ​​the fault are determined. This feature is then combined into the location array according to the node order to obtain the fault type and section location identification results; specifically as follows: Read the hexadecimal fault status records uploaded periodically by all edge terminals from the communication gateway buffer. T2 reports 0x01, T3 reports 0x01, and T4 reports 0x00.

[0042] Based on the above records, construct a one-dimensional state array State_Arr[k]. Traverse the elements of this array, and perform a Boolean AND operation on adjacent array elements with the same topological level pointer, where the operation between T2 and T3 is 1 AND 1 = 1. Extract the terminal index numbers where two or more consecutive operations result in 1, forming a continuous set {T2, T3}.

[0043] Extracting time-series imprint parameters from records T2 and T3: T2 start timestamp T3 Startup Timestamp Perform subtraction to find the absolute difference. Extracting the timing consistency tolerance range parameter. Substitute the difference into the comparison. If the value falls within this interval, output the timing consistency scalar value 1.

[0044] Extract the physical coordinate data of the first and last ends of this continuous set and the associated line segment number L3, using these as the boundary of the affected area. Perform bit combination and arrangement of the text fields "Fault Segment L3", "Type: Single-phase grounding", and "Node Set: N2-N3", mapping them to a specific node according to the spatial topology. The fault type and segment location identification results are obtained from the structure array memory block.

[0045] The advantage of this approach is that it constructs a direct mapping of the spatial correlation of nonlinear states by comparing the direct Boolean logic AND operation of the state words of adjacent nodes with the numerical value of the timing difference.

[0046] S3: Based on the fault type and section location identification results, determine the validity of the protection criteria under the current topology, and combine the preset protection setting adjustment algorithm to determine the adaptive protection trigger threshold and delay parameters that match the current operating conditions, and generate the adaptive protection strategy adjustment results.

[0047] Specifically, S3 includes: S301: Based on the fault type and section location identification results of the distribution network adaptive protection based on edge computing, the hierarchical position of each node in the distribution network topology sequence is extracted. The short-circuit current contribution, system impedance variation, and protection coordination relationship of distributed generation in the corresponding section are compared and retrieved to monitor whether they meet the operating conditions of instantaneous overcurrent protection, time-limited instantaneous overcurrent protection, or overcurrent protection. Based on the intersection relationship between system operating parameters and setting curves during the comparison process, a protection setting correction annotation set is obtained; specifically as follows: Based on the aforementioned identification results, “segment L3” is read, and the topology hierarchy database is queried to extract its hierarchy parameter Level=3.

[0048] Read the nameplate parameter table of the distributed power source DG2 connected within section L3 and extract the rated output current. Current limiting factor parameter The setting is based on the hard-wire overload cutoff setting of the IGBT module inside the inverter, and the limit tolerance coefficient given in the factory test verification report is read. Times, setting Perform a multiplication operation to obtain the contribution of the short-circuit current. .

[0049] Extract the system equivalent impedance before short circuit from the substation terminal waveform file. Extract the system equivalent impedance during the short-circuit period considering DG2 in the constant current source model. Perform complex subtraction to obtain the impedance change. .

[0050] Extracting the instantaneous trip setting register data of the main protection relay The instantaneous value of the total fault current flowing through the protection installation point after extracting and superimposing the contribution of DG (Disruptive Gauge) is calculated. .right and Perform size comparison. The branch is determined to meet the triggering conditions for instantaneous overcurrent protection without delay.

[0051] Substitute into the mathematical formula of the inverse time-limit property Set the benchmark value ,constant Time multiplier and Perform exponentiation and division operations to calculate the time of action intersection. .

[0052] The above calculations obtained , and The parameters are arranged in address order and written into the data dictionary to obtain the protection setting correction annotation set.

[0053] S302: Call the edge computing-based distribution network adaptive protection setting correction annotation set, extract the action threshold corresponding to each operating condition and its priority number in the protection logic, monitor the protection function type currently executed by each edge terminal, and compare the mapping relationship of its corresponding setting calculation parameter set in the adaptive algorithm library. Based on the attribute consistency relationship between fault severity and power supply reliability requirements, obtain the protection parameter matching index set; as detailed below: Access the data dictionary and read the currently matched second system operating condition. Extract the corresponding rated action threshold parameter from this condition's table entry. Extract its priority control bit value 2.

[0054] Read the current function code register word of the edge terminal T4 protection execution module. The return value is 0x02, which corresponds to the time-limited current instantaneous overcurrent protection mechanism.

[0055] The set of constant value calculation parameters is called from the adaptive algorithm library corresponding to the second working condition according to the mapping rule library.

[0056] Table 2: Adaptive Correction Parameter Table for Protection Settings

[0057] As shown in Table 2, the set of constant multipliers is extracted. For the severity attribute, perform division to calculate the severity index. Extracting the power supply reliability coefficient requirement constant. This constant is determined by the proportion of primary load capacity attached to the node, based on the measured proportion. Multiplied by a fixed weight , set as Perform a product operation to obtain the comprehensive attribute index. Consistency is extracted using interval parameters. ,Will Compare with the interval endpoints to confirm that it falls within the interval. Output a matching flag of 1, extract the multiplier set pointer addresses to form a linked list, and obtain the protection parameter matching index set.

[0058] S303: Based on the edge computing-based distribution network adaptive protection parameter matching index set, the setpoint configuration parameter pointed to by each item is called. The hardware execution mechanism and logic operation unit of each edge terminal perform parameter writing operations, and the time-series state mapping processing is performed between the fault duration prediction value and the action delay configuration item in the same section to obtain the adaptive protection strategy adjustment result; as follows: Based on the memory addresses pointed to by the aforementioned list of links, read the specific configuration items. Extract the new operating current reference value. Action delay baseline value .

[0059] The memory allocation table of the DSP control chip built into the T4 hardware actuator of the terminal is used to determine the operating current. Converted to hexadecimal code 0x01C2, a data write operation is performed to address 0x1000 via the SPI bus; a delay will be applied. Multiply by the proportionality factor Convert to millisecond integer This is converted to hexadecimal code 0x012C, and a write operation is performed to address 0x1002.

[0060] Retrieve the predicted limit value of the L3 fault duration calculated by the thermal stability limit location model. Extract the newly written action delay configuration parameters. Perform a direct subtraction operation on the two time-series data points mentioned above: .

[0061] Compare the result with the zero point. Compare, Output a positive margin flag. Package the completed write receipt code and the margin flag into a JSON format parameter message and return it to the main process to obtain the adaptive protection strategy adjustment result.

[0062] The advantage of this approach is that it directly compares the predicted duration with the hardware configuration latency through subtraction operations, thereby constructing an adaptive verification mathematical logic based on the time sequence dimension.

[0063] S4: Invoke the adaptive protection strategy to adjust the results, determine the collaborative protection needs of adjacent nodes based on the peer-to-peer communication protocol between edge terminals, identify the fault isolation boundary and load transfer path, select the distributed collaborative tripping command suitable for the current fault condition, and generate the collaborative linkage protection execution result.

[0064] Specifically, S4 includes: S401: The system invokes the adjustment results of the adaptive protection strategy for the distribution network based on edge computing, retrieves the protection action logic parameters and communication neighbor node set corresponding to each edge terminal, monitors the round-trip delay and data packet drop rate of the communication links between nodes, performs logical AND operations on the trip warning signals and blocking request signals issued by each node, classifies and expresses the decision consistency status of each edge terminal under the distributed architecture, and obtains a collaborative decision-making sequence; specifically as follows: Extract the parameter list from the JSON message. Read the pre-configured routing table information of terminal T2 and extract the set of MAC addresses of its connected communication neighbor nodes {T1_MAC, T3_MAC}.

[0065] The network controller sends a fixed-length 64-byte link test packet to the T3_MAC port, reads the transmission timestamp from the local hardware high-precision timer, and reads the timestamp again when the receiving end returns an ACK response. The round-trip time is then calculated by subtracting the two timestamps. .

[0066] Retrieve the network card register records and extract the total number of data packets sent within the previous 5-minute period. Extract the counter value for each item that has not received a response confirmation. The packet drop rate is obtained by performing division and percentage conversion. .

[0067] Read the trip warning signal bit value cached in the T2 local logic controller. Extract the bit values ​​of the interlocking request signal sent by T3 via the network. Execute bitwise operation rules: for After performing the logical NOT operation, and Perform a logical AND operation; the expression is: Substitute in specific values: .

[0068] Extracting the collaborative classification rule set: setting the time delay And packet loss rate and At that time, the classification status code is determined and assigned as 0x0A. Current measured data. , and If all conditions are met, write code 0x0A to the end of the queue to obtain the collaborative decision-making sequence.

[0069] S402: A collaborative decision-making sequence for adaptive protection of distribution networks based on edge computing. This sequence detects the rated breaking capacity and current load current of the boundary switches in the fault section, retrieves the corresponding backup power capacity and current carrying capacity of the transfer branches within that range, determines the recovery scheme for the non-faulty area after the fault section is cleared, and maps this scheme to the protection action sequence to obtain the linkage control associated state set; specifically as follows: Parse the classification codes in the sequence to extract the hardware device code CB1 for the fault section boundary switch. Retrieve its nominal rated short-circuit breaking capacity from the equipment ledger database. Read the effective value of the load current flowing through the primary side of the current transformer (CT). .

[0070] Table 3: Mapping Table of Execution Status of Cooperative Control Logic

[0071] As shown in Table 3, the real-time available power supply capacity of the surrounding topologically connected backup substations is extracted. Extract the network nominal line voltage. Substitute the values ​​into the apparent power formula to obtain the maximum support current. Substitute the data and perform multiplication and division to get Read the rated current-carrying capacity parameters of the conductors for the backup branch. .

[0072] Perform a three-way numerical comparison operation: Compared with current load , The gap value is calculated by subtracting the two values. Based on this difference status word, the recovery instruction set sequence for "partial load reduction followed by power transfer" is retrieved from the recovery scheme mapping table.

[0073] Extract the absolute time-scale sequence data of protection actions: Trip the circuit breaker. Perform load reduction. The recovery instruction set string is appended to the end of the above action timestamp string using an underscore, stored in a hash table, and the associated state set of the linkage control is retrieved.

[0074] S403: Based on the edge computing-based distribution network adaptive protection linkage control associated state set, for each edge terminal, the corresponding actuator control parameters are called to determine the consistency between the trip coil operating current and the auxiliary contact feedback state. The determination result is mapped to the corresponding position in the execution command sequence, and the switch action control signal is output according to the timing requirements to generate the coordinated linkage protection execution result; specifically as follows: Traverse the associated records in the hash table and retrieve the corresponding closed-loop control parameters of the edge terminal actuator by device code. For the embedded relay circuit of T2, read the operating current value after voltage conversion at both ends of the DC shunt in the trip coil circuit. .

[0075] Extract the upper and lower limit parameter ranges of the operating current reference from the datasheet. .Will If the result is compared with the endpoint of the interval and found to be within the interval, the hardware allows the action.

[0076] The hardware feedback level of the circuit breaker's auxiliary normally closed contact was read through the optocoupler isolation interface. The measured input pin voltage was... The state is converted to digital logic state 0 (representing complete disconnection). The target state 0x00 of the software control register issued before the instruction was sent is retrieved. A bitwise XOR operation is performed on both state words: 0XOR0 = 0.

[0077] The XOR result is The flag instruction is consistent within the closed loop. This digital flag is directly written to offset address 0x20 of the memory block instruction sequence. The system clock signal is read. When the predetermined action is reached The interrupt function is triggered on time, and the duty cycle is written to the address bus corresponding to the PWM pulse generator controller. The drive command word continuously outputs a high level to the external pins, generating the result of the coordinated protection execution.

[0078] The advantage of this approach is that it eliminates the latency loss associated with complex state machines by directly comparing the expected state word in the software with the logic level in the hardware feedback through a bitwise XOR operation.

[0079] S5: Based on the results of coordinated protection, control the distribution network switching equipment to perform actions and monitor the system feedback after the actions in real time. Combine the operating status update information of each node and match it to the global power grid model to complete the reconstructed operating structure after the distribution network protection actions and generate the dynamic optimization results of the distribution network adaptive protection.

[0080] S501: Based on the execution results of the edge computing-based adaptive protection and coordinated linkage protection of the distribution network, retrieve the physical location and corresponding logical number of each switch in the distribution network coordinate system, execute pulse output operation for the closing and opening drive circuit corresponding to each switch, switch the relative order of their closing or opening states according to the protection strategy, and synchronously correct the electrical connection relationship between adjacent sections to obtain the distribution network operation state rearrangement sequence; as follows: Parse the pulse drive command response sequence and retrieve the spatial coordinates and logical index entries of each operating switch from the network database. Read the coordinate data of the tie switch CB2 in the closed-position change state. And its corresponding topology pointer number Node_5.

[0081] The main control chip sends an address set operation to the base pin of the corresponding closing / opening driver transistor CB2 via the GPIO control register, and the pin outputs a high level. At the same time, it activates the internal timer to count, and when the count reaches equal to When the value is equal to the specified value, the set bit is canceled, and a complete rectangular pulse output operation is performed.

[0082] Read the relative priority flag array in the policy table: CB2 action sequence number must take precedence over CB3. After the monitoring interrupt captures a feedback message indicating that the CB2 status has changed to 1 (closed), load the timer. The count value triggers a pulse output of the same type to the CB3 pin after the count is full.

[0083] Read the original dynamic topology connection matrix structure, locate the coordinates of the 5th row and 6th column corresponding to CB2, overwrite the element value 0 representing disconnection with the element value 1 representing connection, and synchronously overwrite the coordinate values ​​of all state change associated switches. Then, export the updated matrix multidimensional array sequence to obtain the power distribution network operation state rearrangement sequence.

[0084] S502: An edge computing-based adaptive protection system for distribution networks rearranges the network operation status sequence, extracts the power flow distribution parameters and voltage levels corresponding to each feeder branch, performs position comparison judgment on its safety constraint boundaries in the target operation mode, issues adjustment commands to the reactive power output of distributed generation sources based on the voltage deviation range, and establishes a mapping relationship between the adjusted operation parameters and their corresponding topology nodes to obtain the operation condition positioning mapping set; details are as follows: Load the updated matrix sequence and power flow calculation module interface.

[0085] Extract the power flow distribution measurement parameters of feeder L4 branch and read its three-phase total active power variable. Read the effective value of the node line voltage Extract the node voltage lower limit values ​​set in the safety constraint database. The threshold value is taken as the system's rated line voltage. The ratio of the lower limit The product is calculated to obtain Perform direct subtraction operation: The sign of the result value is determined; a negative sign indicates that the voltage level at that node has exceeded the limit and dropped significantly. The equivalent reactance parameter of that branch is then extracted. .

[0086] Substituting the reactive power-voltage sensitivity simplified relationship, we can derive the reactive power demand. Substitute the measured values ​​and parameters into the arithmetic operations: That is, converted to .

[0087] Reactive power increase is issued to the distributed power reactive power control channel connected to this node. The absolute value is given in the instruction frame.

[0088] In the local memory structure, the original value of the reactive power field in the running parameter record corresponding to this node is changed from... Revised to Establish a key-value pair linked list with the corresponding node label address to obtain the operating condition positioning mapping set.

[0089] S503: The edge computing-based distribution network adaptive protection operation condition location mapping set is invoked to perform consistency checks on the power balance status of each node across the entire network. Based on the topology order, the system structure hierarchy is updated, and the operation results of each node are written into a unified power grid operation description sequence to generate dynamic optimization results for distribution network adaptive protection. Details are as follows: The information of each key-value pair node in the storage linked list is parsed sequentially.

[0090] Targeting the entire network A topological node is identified, and the measured active power values ​​of the meters at each node are extracted to form a one-dimensional array. The actual elements extracted from the array are: (Assume the power injection is positive and the load power consumption is negative). Perform a summation arithmetic operation on this array: .

[0091] Substituting the above The floating-point elements are added together to calculate the total value. Retrieve the estimated total line loss obtained from the real-time estimation of the network loss model. Performing a subtraction operation on the two yields the difference as: .

[0092] Extract the preset tolerance range for the network-wide active power imbalance. Perform interval comparison, because Within this tolerance limit, write the Boolean truth value 1 to the global check status word register.

[0093] Traverse the topological node matrix in ascending order from low to high, and synchronously overwrite the latest closed switch index bit in the hierarchical relationship dictionary.

[0094] The power, voltage and current measurements, state balance flag, and switch connectivity identifier of each node are written byte by byte into a unified hexadecimal power grid operation description string stack structure. The calculation process ends and the string text is output to generate the dynamic optimization result of the distribution network adaptive protection.

[0095] like Figure 2 As shown, this embodiment also provides a distribution network adaptive protection system based on edge computing, which executes a distribution network adaptive protection method based on edge computing; the system includes: The multi-source information sensing module is used to acquire multi-source operating data in real time through edge terminals deployed at key nodes of the distribution network, identify feeder segment information, determine the current operating conditions and topology of the distribution network, extract state feature vectors representing the system's operating characteristics, and generate multi-source information sensing results for the distribution network. Feeder segment information includes current vectors, node voltage waveforms, power factors, and switch position status. Specifically, it acquires power sampling data and switch status data from each node of the distribution network, filters and scales current and voltage signals through signal conditioning circuits, detects the secondary outputs of current transformers and voltage transformers, identifies the access capacity and grid connection status of distributed power sources, determines the current topology of the distribution network, and generates multi-source information sensing results for the distribution network.

[0096] The fault location and identification module is used to call upon the multi-source information sensing results of the distribution network, identify the transient change characteristics and steady-state deviation characteristics of the electrical quantities of each node, distinguish the manifestations of abnormal operating conditions, summarize the regional range and severity of the fault, and generate fault type and section location identification results. Abnormal operating conditions include short-circuit faults, overload operation, and islanded operation. Specifically, it calls upon the multi-source information sensing results of the distribution network adaptive protection based on edge computing, and determines the location and type of the fault based on the differential current characteristics and sequence component characteristics calculated by each edge terminal. Furthermore, based on the arrival time difference of the transient traveling wave signal, it performs precise location analysis of the fault point, summarizes the regional range affected by the fault, and generates fault type and section location identification results.

[0097] The adaptive strategy module is used to call the fault type and section location identification results, determine the validity of the protection criteria under the current topology, and combine the preset protection setting adjustment algorithm to determine the adaptive protection trigger threshold and delay parameters that match the current operating conditions, generating adaptive protection strategy adjustment results. Specifically, it calls the fault type and section location identification results of the distribution network adaptive protection based on edge computing, analyzes the matching relationship between the fault current level and the protection setting based on the current distribution network operation mode, and refers to the preset protection setting library and coordination logic, dynamically adjusts the action threshold of each level of protection, and generates adaptive protection strategy adjustment results.

[0098] The distributed coordination module is used to call the adaptive protection strategy adjustment results, determine the coordinated protection needs of adjacent nodes based on the peer-to-peer communication protocol between edge terminals, identify fault isolation boundaries and load transfer paths, select the distributed coordinated tripping command suitable for the current fault condition, and generate the coordinated linkage protection execution results; that is, it calls the distribution network adaptive protection strategy adjustment results based on edge computing, analyzes the switching action logic and interlocking relationship required for fault isolation based on the status of the peer-to-peer communication link and the information exchange content between edge terminals, determines the most suitable coordinated linkage scheme, and generates the coordinated linkage protection execution results.

[0099] The reconfiguration module is used to call the execution results of the collaborative protection system. Based on the operating characteristics of the switching equipment and the load transfer requirements of the distribution network, it adjusts the operating timing of each switch, controls the switch actuators to complete fault isolation and non-fault zone restoration, and generates dynamic optimization results for the adaptive protection of the distribution network. In other words, it calls the execution results of the collaborative protection system based on edge computing, adjusts the operating timing of each switch based on the operating characteristics of the switching equipment and the load transfer requirements of the distribution network, controls the switch actuators to complete fault isolation and non-fault zone restoration, and generates dynamic optimization results for the adaptive protection of the distribution network.

[0100] 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.

[0101] This embodiment provides a distribution network adaptive protection method and system based on edge computing. By deploying edge computing units at key nodes and combining them with multi-source information fusion intelligent algorithms, it achieves rapid identification of complex operating conditions and adaptive adjustment of protection strategies, thereby improving the power supply reliability and operational flexibility of the distribution network in the scenario of new energy access.

[0102] The above descriptions are merely embodiments of the present invention. Commonly known structures and characteristics of the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent.

Claims

1. A distribution network adaptive protection method based on edge computing, characterized in that, include: S1: By acquiring multi-source operation data in real time through edge terminals deployed at key nodes of the distribution network, identifying feeder segment information, determining the current operating conditions and topological connection relationship of the distribution network, extracting state feature vectors that characterize the system operation features, and generating multi-source information perception results of the distribution network; feeder segment information includes current vector, node voltage waveform, power factor, and switch position status; S2: Call the multi-source information sensing results of the distribution network to identify the transient change characteristics and steady-state deviation characteristics of the power of each node, distinguish the manifestations of abnormal operating conditions, summarize the area range and severity of the fault, and generate fault type and section location identification results; abnormal operating conditions include short circuit faults, overload operation and islanding operation; S3: Based on the fault type and section location identification results, determine the validity of the protection criteria under the current topology, and combine the preset protection setting adjustment algorithm to determine the adaptive protection trigger threshold and delay parameters that match the current operating conditions, and generate the adaptive protection strategy adjustment results; S4: Invoke the adaptive protection strategy adjustment result, determine the collaborative protection needs of adjacent nodes based on the peer-to-peer communication protocol between edge terminals, identify the fault isolation boundary and load transfer path, select the distributed collaborative tripping command suitable for the current fault condition, and generate the collaborative linkage protection execution result.

2. The adaptive protection method for distribution networks based on edge computing according to claim 1, characterized in that, The results of multi-source information perception of the distribution network in S1 include current amplitude and phase sequence, voltage harmonic distortion rate, frequency offset, and status of distributed power source access points; in S2, the results of fault type and section location identification include fault phase identifier, transient energy distribution characteristics, and topology branch correlation matrix; in S3, the results of adaptive protection strategy adjustment include setting correction coefficient, action time ladder sequence, and criterion logic switching identifier. In S4, the results of coordinated protection execution include circuit breaker closing and opening sequences, logic blocking signals, and islanding operation control commands.

3. The adaptive protection method for distribution networks based on edge computing according to claim 1, characterized in that, The steps to obtain S1 include: S101: Calculate the effective value and phase angle of the preset components in the original current and voltage signal sampling sequence of each monitoring point in the distribution network to obtain the basic characteristic dataset of the power quantity; S102: Based on the basic power characteristic dataset, compare the current vector difference and voltage drop value between adjacent monitoring points in real time, and perform topology connectivity verification processing on the set of branches whose current increment exceeds the preset start threshold to obtain a dynamic topology connection matrix. S103: Based on the dynamic topology connection matrix, extract the grid connection parameters and synchronous switching information of the distributed power source access point, and make a linkage judgment on the power flow change status and load balance within the distribution network to obtain the multi-source information perception results of the distribution network.

4. The adaptive protection method for distribution networks based on edge computing according to claim 1, characterized in that, The steps to obtain S2 include: S201: Call the multi-source information sensing results of the distribution network, calculate the differential current characteristics and sequence component characteristics in sequence, and obtain the fault initiation feature matrix when the fault is determined to be an intra-zone fault state. S202: Based on the fault initiation feature matrix, extract the arrival time difference of all transient traveling wave signals, label the fault phases in sequence, and obtain a multi-dimensional fault feature classification set; S203: Call the multi-dimensional fault feature classification set, perform spatial correlation statistics on the fault identification status reported by all edge terminals, and determine the specific physical location and affected range of the fault based on the continuity and temporal consistency of the distribution of the same type of fault status on the topology branch in the statistical sequence. Combine the features into the positioning array according to the node order to obtain the fault type and segment positioning identification results.

5. The adaptive protection method for distribution networks based on edge computing according to claim 1, characterized in that, The steps to obtain S3 include: S301: Based on the fault type and section location identification results, extract the hierarchical position of each node in the distribution network topology sequence, and compare the relevant parameters of distributed power source contribution in the corresponding section to determine whether they meet the protection action conditions. Based on the intersection relationship between the system operating parameters and the set value curve during the comparison process, obtain the protection set value correction label set. S302: Call the protection setting correction annotation set, monitor the protection function type currently executed by each edge terminal, and compare the mapping relationship of the corresponding setting calculation parameter set in the adaptive algorithm library to obtain the protection parameter matching index set; S303: Based on the protection parameter matching index set, call the fixed configuration parameter pointed to by each item, perform parameter writing operation for each edge terminal, and perform time-series state mapping processing between the fault duration prediction value and the action delay configuration item in the same segment to obtain the adaptive protection strategy adjustment result.

6. The adaptive protection method for distribution networks based on edge computing according to claim 1, characterized in that, The steps to obtain S4 include: S401: Invoke the adaptive protection strategy adjustment results, retrieve the protection action logic parameters and communication neighbor node set corresponding to each edge terminal, classify and express the decision consistency status of each edge terminal under the distributed architecture according to the communication link status information between each node and the published tripping and blocking related signals, and obtain the collaborative decision judgment sequence. S402: Based on the collaborative decision-making sequence, detect the rated breaking capacity and current load current value of the fault section boundary switch, perform corresponding retrieval of the backup power capacity and transfer branch current within the range, determine the non-fault area recovery scheme after the fault section is cleared, and associate and map it with the protection action sequence to obtain the linkage control associated state set. S403: Based on the linkage control associated state set, for each edge terminal, call its corresponding actuator control parameters, make a consistency judgment on the trip coil operating current and auxiliary contact feedback state, map the judgment result to the corresponding position in the execution instruction sequence, output the switch action control signal according to the timing requirements, and generate the collaborative linkage protection execution result.

7. The adaptive protection method for distribution networks based on edge computing according to claim 1, characterized in that, It also includes S5: Based on the results of coordinated protection, it controls the distribution network switching equipment to perform actions and monitors the system feedback after the actions in real time, combines the operating status update information of each node and matches it to the global power grid model, completes the reconstructing of the operating structure after the distribution network protection actions, and generates the dynamic optimization results of the distribution network adaptive protection.

8. The adaptive protection method for distribution networks based on edge computing according to claim 7, characterized in that, The steps to obtain S5 include: S501: Based on the results of the coordinated protection execution, the relative order of each switch in the closed or open state is switched according to the protection strategy, and the electrical connection relationship between adjacent sections is simultaneously corrected to obtain the power distribution network operation status rearrangement sequence. S502: Based on the reordering sequence of distribution network operation status, extract the power flow distribution parameters and voltage level corresponding to each feeder branch, perform position comparison judgment on its safety constraint boundary in the target operation mode, issue adjustment commands to the reactive power output of distributed power sources according to the voltage deviation range, and establish a mapping relationship between the adjusted operation parameters and their corresponding topology nodes to obtain the operation condition positioning mapping set. S503: Call the operating condition location mapping set to perform consistency verification on the power balance status of each node in the entire network, perform status update on the system structure hierarchy according to the topology order, write the operating results of each node into a unified power grid operation description sequence, and generate the dynamic optimization results of the distribution network adaptive protection.

9. The adaptive protection method for distribution networks based on edge computing according to claim 7, characterized in that, In S5, the dynamic optimization results of the distribution network adaptive protection include switch action sequence records, fault isolation range maps, and power transfer schemes for non-faulty sections.

10. A distribution network adaptive protection system based on edge computing, characterized in that, The system is used in the edge computing-based adaptive protection method for distribution networks according to any one of claims 1-9, and the system comprises: The multi-source information sensing module is used to acquire multi-source operating data in real time through edge terminals deployed at key nodes of the distribution network, identify feeder segment information, determine the current operating conditions and topological connection relationship of the distribution network, extract state feature vectors that characterize the system's operating features, and generate multi-source information sensing results for the distribution network; feeder segment information includes current vector, node voltage waveform, power factor, and switch position status; The fault location and identification module is used to call the multi-source information sensing results of the distribution network, identify the transient change characteristics and steady-state deviation characteristics of the power of each node, distinguish the manifestation of abnormal operating conditions, summarize the area and severity of the fault, and generate fault type and section location identification results; abnormal operating conditions include short circuit faults, overload operation and islanding operation; The strategy adaptive module is used to call the fault type and section location identification results, determine the validity of the protection criteria under the current topology, combine the preset protection setting adjustment algorithm, determine the adaptive protection trigger threshold and delay parameters that match the current operating conditions, and generate the adaptive protection strategy adjustment results. The distributed coordination module is used to call the adaptive protection strategy adjustment results, determine the coordination protection needs of adjacent nodes based on the peer-to-peer communication protocol between edge terminals, identify the fault isolation boundary and load transfer path, select the distributed coordination trip command suitable for the current fault condition, and generate the coordination linkage protection execution result. The reconfiguration module is used to call the execution results of the coordinated protection. Based on the operating characteristics of the switching equipment and the load transfer demand of the distribution network, it adjusts the operating sequence of each switch, controls the switch actuator to complete fault isolation and non-fault zone restoration, and generates dynamic optimization results for the adaptive protection of the distribution network.

Images