A method and system for terminal control of power distribution automation based on multi-criteria fusion
By using multi-criteria fusion and distributed collaborative control, the problems of insufficient reliability of single criteria and insufficient information fusion in fault detection and control of distribution automation terminals are solved, and efficient fault identification, rapid location and power restoration are achieved.
Patent Information
- Application Number
- CN202611366823.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-04
- Publication Date
- 2026-11-10
AI Technical Summary
Existing fault detection and control technologies for distribution automation terminals suffer from several problems, including insufficient reliability of single criteria, lack of information fusion between criteria, long fault location time, criterion failure due to distributed power source access, lack of adaptability in recovery strategies, weak inter-terminal coordination capabilities, and insufficient utilization of environmental and non-electrical quantity information.
A multi-criteria fusion method is adopted, which integrates electrical quantities, non-electrical quantities and communication criteria. Information fusion is performed through Dempster-Shafer evidence theory, combined with adaptive weight allocation and distributed cooperative control, to achieve fault diagnosis and power restoration.
It significantly improves the accuracy of fault identification and positioning, enhances the adaptability to complex working conditions, reduces the risk of false operation and failure to operate, achieves rapid fault isolation and power restoration, and reduces dependence on the master station.
Smart Images

Figure CN123073752A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of power distribution automation technology, specifically relating to a power distribution automation terminal control method and system based on multi-criteria fusion. Background Technology
[0002] Distribution automation is one of the core technologies of smart distribution networks. By deploying automated terminal devices at the switches of distribution network lines, it enables the monitoring of the distribution network's operating status, the detection and isolation of faults, and the rapid restoration of power supply. The distribution automation terminal is the "nerve ending" of the distribution automation system, directly collecting line electrical quantities and executing switch control operations.
[0003] Currently, fault detection and control in distribution automation terminals mainly adopt methods based on a single criterion or a simple combination of criteria. In existing technologies, the fault detection criteria of distribution automation terminals are usually independent and unrelated. For example, overcurrent protection and zero-sequence overcurrent protection have their own independent setting values and action logic, without information fusion between criteria. Fault location mainly relies on the "reclosing-test" method or on centralized fault information analysis at the master station, which is time-consuming and inefficient. Recovery strategies usually adopt preset fixed schemes and are not optimized according to real-time operating status.
[0004] Existing distribution automation terminal control technologies based on a single criterion or a simple combination of criteria have the following drawbacks: Defect 1: Insufficient reliability of a single criterion. When distribution automation terminals rely solely on overcurrent criteria for fault detection, the overcurrent criterion may fail to trigger correctly in cases of changes in system operation (such as altered short-circuit current distribution due to distributed power source integration) or when fault characteristics are not obvious (such as high-resistance grounding faults where the short-circuit current is only tens of amperes), leading to missed fault detection. Similarly, relying solely on zero-sequence overcurrent criteria may also fail in systems where the neutral point is grounded via an arc suppression coil (where zero-sequence current is compensated and suppressed). Furthermore, the false tripping rate and failure-to-trip rate of a single criterion are difficult to meet requirements simultaneously.
[0005] Defect 2: Lack of information fusion among criteria leads to low fault diagnosis accuracy. In existing terminals, each fault criterion operates independently, without information exchange or comprehensive judgment between them. When the outputs of multiple criteria are inconsistent (e.g., the overcurrent criterion is triggered but the zero-sequence criterion is not), the terminal cannot make a comprehensive decision and can only select the result of a single criterion according to a preset priority, easily leading to misjudgment. The lack of a multi-criterion fusion mechanism prevents the terminal from fully utilizing the complementarity of multi-source information, limiting the potential for improving fault detection accuracy.
[0006] Defect 3: Fault location is time-consuming and inaccurate. Existing fault location methods mainly employ two approaches: one is the "reclosing-trial" method, which determines the faulty section through multiple reclosing attempts. Each trial takes 5-15 seconds, and the entire process can take more than 30 seconds, with multiple trials impacting the system; the other is the centralized analysis method from the master station, which requires waiting for all terminals to send fault information before the master station performs comprehensive analysis. Due to communication delays and limitations in the master station's processing capacity, the location time is typically 1-3 minutes. The accuracy of both methods depends on the accuracy of the trigger criteria; if the criteria are mistriggered or fail to trigger, the location results are unreliable.
[0007] Defect 4: Distributed power source integration renders traditional criteria ineffective. With the integration of numerous distributed photovoltaic and wind power sources into the distribution network, the network has transformed from a single-source radial network to a multi-source network. Traditional fault detection criteria and location logic based on the assumption of a single-source radial network are no longer applicable. For example, the overcurrent criterion setting is calculated based on the short-circuit current supplied by a single power source in the substation. After the integration of distributed power sources, the short-circuit current distribution changes, which may lead to changes in the protection range; directional elements may misjudge the direction when the fault current supplied by the distributed power source flows bidirectionally.
[0008] Defect 5: Lack of Adaptability in Restoration Strategies. Existing power restoration strategies typically employ pre-defined, fixed schemes that are not optimized based on real-time conditions such as fault type, fault location, load distribution, and distributed generation output. In complex distribution networks (such as multi-interconnection, multi-segment structures), fixed schemes may fail to achieve optimal restoration results, potentially leading to excessively long restoration delays or even failure to restore power to some non-faulty areas. Furthermore, the restoration process lacks differentiated handling of load priorities and power supply reliability.
[0009] Defect 6: Weak inter-terminal coordination. Current distribution automation terminals primarily rely on centralized dispatching from the master station for coordination, with no direct information exchange between terminals. When master station communication is interrupted or processing is delayed, terminals cannot coordinate autonomously and must rely on local control logic for fault handling, resulting in limited processing capacity and efficiency. The lack of a distributed coordination mechanism between terminals restricts the speed and reliability of fault handling.
[0010] Defect 7: Insufficient utilization of environmental and non-electrical quantity information. Existing terminals primarily rely on electrical quantities for fault diagnosis, failing to fully utilize non-electrical quantity information such as temperature, humidity, partial discharge, and the mechanical characteristics of switching equipment. In certain fault scenarios (such as mechanical faults in switching equipment or faults caused by slow insulation degradation), electrical quantity characteristics are not obvious, but non-electrical quantity information may indicate abnormal conditions in advance. The lack of comprehensive judgment based on both electrical and non-electrical quantities limits the comprehensiveness of fault detection. Summary of the Invention
[0011] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a power distribution automation terminal control method and system based on multi-criteria fusion, so as to solve the technical problems existing in the related technologies mentioned above.
[0012] To achieve the above objectives, the present invention provides the following technical solution: A microprocessor-based protection control method for adaptive line protection includes the following steps: S1: Collect line electrical quantities, non-electrical quantities, and neighborhood communication data, and perform preprocessing; S2: Generates multiple fault criteria from preprocessed data, including electrical quantity criteria, non-electrical quantity criteria, and communication criteria, with each criterion outputting an independent fault confidence value; S3: Dynamically calculate the fusion weight of each criterion based on the current operating conditions, data quality, and historical performance; S4: Map the confidence values of each criterion to the basic probability allocation function, use the Dempster combination rule to recursively fuse them, and monitor the conflict coefficient between criteria in real time. S5: Calculate the comprehensive fault confidence based on the fusion results, and output the fault determination and fault type discrimination results; S6: After a fault is determined, exchange direction criterion results with adjacent terminals and determine the faulty section through distributed reasoning; S7: Terminals at both ends of the faulty section coordinate to perform isolation operations; S8: Generate and evaluate multiple power restoration plans, select the optimal plan, and execute them in a distributed manner; S9: Perform a post-evaluation of the fault handling process and update the criterion weights and strategy parameters.
[0013] In step S2, the electrical quantity criteria include overcurrent criteria, sudden change criteria, zero-sequence direction criteria, negative-sequence direction criteria, undervoltage criteria, impedance criteria, and harmonic criteria; the non-electrical quantity criteria include temperature anomaly criteria, partial discharge criteria, and switch mechanical characteristic criteria; the communication criteria include neighborhood overcurrent criteria, neighborhood direction criteria, and master station confirmation criteria; and each criterion uses a soft threshold function to calculate the fault confidence value.
[0014] In step S3, the adaptive weight calculation formula is: wi = α×wi_cond + β×wi_qual + γ×wi_hist, where wi_cond is the operating condition weight component, determined based on the neutral grounding method, distributed power supply output status, and load level; wi_qual is the data quality weight component, determined based on data integrity rate, timeliness, and sensor health status; wi_hist is the historical performance weight component, determined based on the correct trigger rate and false trigger rate of the criterion; α, β, and γ are weighting coefficients, and α+β+γ=1; after weight normalization, Σwi = 1 is satisfied.
[0015] In step S6, the reasoning rule for distributed fault location is as follows: if the direction criterion of this terminal points downstream and the direction criterion of the downstream adjacent terminal points upstream, then the fault is in the segment between this terminal and the downstream adjacent terminal; if the direction criterion is uncertain, a comprehensive inference is made by combining the overcurrent status of the neighborhood; the location result is consistent across all terminals, and if there is inconsistency, a consensus is reached through multiple rounds of information exchange.
[0016] In step S8, the evaluation indicators for the power supply restoration scheme include the restored load, restoration time, post-restoration load rate, voltage deviation, and number of switch actions; the scheme scoring function is: Score=w1×P_restore-w2×T_restore-w3×N_switch-w4×V_deviation; before closing the tie switch, a no-voltage check and synchronization check are performed, and real-time monitoring and adaptive adjustment are carried out during the restoration process.
[0017] This invention also provides a technical solution: a distribution automation terminal control system based on multi-criteria fusion. The system is deployed at the switchgear of the distribution network line (including pole-mounted switches, ring main units, substations, etc.) and includes five main functional modules: a multi-source data acquisition module, a multi-criteria generation module, a fusion decision engine module, a distributed collaborative control module, and an execution and communication module. The system adopts an edge computing architecture, with each terminal having local intelligent decision-making capabilities, while achieving distributed collaboration between terminals through peer-to-peer communication.
[0018] The multi-source data acquisition module is the data foundation for fusion decision-making. It is responsible for acquiring three types of information sources: electrical quantities, non-electrical quantities, and communication data. The module includes an electrical quantity acquisition subunit, a non-electrical quantity acquisition subunit, a neighborhood communication data receiving subunit, and a data preprocessing subunit.
[0019] The multi-criteria generation module is the source of "multi-criteria" in "multi-criteria fusion", including electrical quantity criteria subunit, non-electrical quantity criteria subunit and communication criteria subunit.
[0020] The fusion decision engine module adopts Dempster-Shafer (DS) evidence theory as the fusion framework and combines it with an adaptive weight allocation mechanism. The module includes a basic probability allocation (BPA) construction subunit, an adaptive weight allocation subunit, an evidence fusion subunit, a conflict handling subunit, and a decision output subunit.
[0021] The distributed collaborative control module includes a neighborhood discovery and topology maintenance subunit, a distributed fault location subunit, a collaborative fault isolation subunit, an adaptive power supply recovery subunit, and a consistency protocol subunit.
[0022] The execution and communication module includes a switch control execution subunit, an action recording subunit, and a communication subunit.
[0023] Compared with the prior art, the present invention has the following beneficial effects: It significantly improves the fault identification accuracy and positioning precision of distribution automation terminals under complex operating conditions, effectively reducing the risk of false operation and failure to operate due to the failure of a single criterion or data anomalies; by introducing a multi-source heterogeneous information fusion mechanism, it enhances the system's adaptability to special scenarios such as high-resistance grounding, weak feeder lines, and networks containing distributed power sources; the adaptive weight adjustment strategy enables each criterion to dynamically exert its optimal performance under different operating states, improving the robustness and real-time performance of overall decision-making; the distributed collaborative control architecture avoids strong dependence on the master station, and can still achieve rapid local isolation and power restoration in the event of communication interruption or master station failure; the combination of edge intelligence and lightweight inference balances response speed and computing resource constraints, making it suitable for low-cost deployment and efficient operation and maintenance of large-scale distribution network terminals. Attached Figure Description
[0024] Figure 1 This is a flowchart of the control method of the present invention.
[0025] Figure 2 This is a block diagram of the control system architecture of the present invention. Detailed Implementation
[0026] To make the above-mentioned objectives, features and advantages of the present invention more readily understood, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0027] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0028] like Figure 1 -like Figure 2 As shown, a microprocessor-based adaptive line protection control method includes the following steps: S1: Collect line electrical quantities, non-electrical quantities, and neighborhood communication data, and perform preprocessing; S2: Generates multiple fault criteria from preprocessed data, including electrical quantity criteria, non-electrical quantity criteria, and communication criteria, with each criterion outputting an independent fault confidence value; S3: Dynamically calculate the fusion weight of each criterion based on the current operating conditions, data quality, and historical performance; S4: Map the confidence values of each criterion to the basic probability allocation function, use the Dempster combination rule to recursively fuse them, and monitor the conflict coefficient between criteria in real time. S5: Calculate the comprehensive fault confidence based on the fusion results, and output the fault determination and fault type discrimination results; S6: After a fault is determined, exchange direction criterion results with adjacent terminals and determine the faulty section through distributed reasoning; S7: Terminals at both ends of the faulty section coordinate to perform isolation operations; S8: Generate and evaluate multiple power restoration plans, select the optimal plan, and execute them in a distributed manner; S9: Perform a post-evaluation of the fault handling process and update the criterion weights and strategy parameters.
[0029] In step S2, the electrical quantity criteria include overcurrent criteria, sudden change criteria, zero-sequence direction criteria, negative-sequence direction criteria, undervoltage criteria, impedance criteria, and harmonic criteria; the non-electrical quantity criteria include temperature anomaly criteria, partial discharge criteria, and switch mechanical characteristic criteria; the communication criteria include neighborhood overcurrent criteria, neighborhood direction criteria, and master station confirmation criteria; and each criterion uses a soft threshold function to calculate the fault confidence value.
[0030] In step S3, the adaptive weight calculation formula is: wi = α×wi_cond + β×wi_qual + γ×wi_hist, where wi_cond is the operating condition weight component, determined based on the neutral grounding method, distributed power supply output status, and load level; wi_qual is the data quality weight component, determined based on data integrity rate, timeliness, and sensor health status; wi_hist is the historical performance weight component, determined based on the correct trigger rate and false trigger rate of the criterion; α, β, and γ are weighting coefficients, and α+β+γ=1; after weight normalization, Σwi = 1 is satisfied.
[0031] In step S6, the reasoning rule for distributed fault location is as follows: if the direction criterion of this terminal points downstream and the direction criterion of the downstream adjacent terminal points upstream, then the fault is in the segment between this terminal and the downstream adjacent terminal; if the direction criterion is uncertain, a comprehensive inference is made by combining the overcurrent status of the neighborhood; the location result is consistent across all terminals, and if there is inconsistency, a consensus is reached through multiple rounds of information exchange.
[0032] In step S8, the evaluation indicators for the power supply restoration scheme include the restored load, restoration time, post-restoration load rate, voltage deviation, and number of switch actions; the scheme scoring function is: Score=w1×P_restore-w2×T_restore-w3×N_switch-w4×V_deviation; before closing the tie switch, a no-voltage check and synchronization check are performed, and real-time monitoring and adaptive adjustment are carried out during the restoration process.
[0033] This invention also provides a technical solution: a microcomputer protection control system for adaptive line protection. The system is deployed at the switchgear of the distribution network (including pole-mounted switches, ring main units, substations, etc.) and includes five main functional modules: a multi-source data acquisition module, a multi-criteria generation module, a fusion decision engine module, a distributed collaborative control module, and an execution and communication module. The system adopts an edge computing architecture, with each terminal having local intelligent decision-making capabilities, while achieving distributed collaboration between terminals through peer-to-peer communication.
[0034] The multi-source data acquisition module is the data foundation for fusion decision-making. It is responsible for acquiring three types of information sources: electrical quantities, non-electrical quantities, and communication data. The module includes an electrical quantity acquisition subunit, a non-electrical quantity acquisition subunit, a neighborhood communication data receiving subunit, and a data preprocessing subunit.
[0035] Furthermore, the electrical quantity acquisition subunit acquires three-phase voltage, three-phase current, zero-sequence voltage, and zero-sequence current signals of the line through voltage transformers and current transformers, with a sampling rate of not less than 32 points per cycle (for a 50Hz system, the sampling rate is not less than 1600Hz). The acquired data is digitally filtered (FIR low-pass filter + 50Hz notch filter) and then output to the multi-criteria generation module. At the same time, this subunit has a synchronous phasor measurement function, which uses GPS / BeiDou timing signals to achieve network-wide synchronous sampling, providing a time reference for distributed collaboration.
[0036] Furthermore, the non-electrical quantity acquisition subunit acquires non-electrical quantities such as the temperature of the switching equipment (via a fiber optic temperature sensor or an infrared temperature sensor), partial discharge signals (via an ultra-high frequency UHF sensor or a high-frequency current HFCT sensor), switch mechanical characteristics (obtaining opening and closing times and travel curves via limit switches and vibration sensors), and ambient temperature, humidity, and air pressure. The sampling period for non-electrical quantities is 100ms-1s (adjusted according to the signal type), and the sampling frequency is automatically increased in case of abnormalities.
[0037] Furthermore, the neighborhood communication data receiving subunit receives status information (including electrical quantity characteristic values, fault criterion outputs, switch positions, etc.) sent by adjacent distribution automation terminals via P2P communication, as well as system topology information, operating parameters, and policy instructions issued by the distribution automation master station. The neighborhood communication data is timestamped for time alignment during multi-criteria fusion.
[0038] Furthermore, the data preprocessing subunit performs outlier removal (based on the 3σ criterion or box plot method), data alignment (time synchronization and spatial mapping), data normalization (mapping data of different dimensions to a unified confidence space), and data validity verification on various types of raw data, providing high-quality input for subsequent multi-criteria generation and fusion.
[0039] The multi-criteria generation module generates multiple fault criteria from the preprocessed multi-source data. Each criterion is independently calculated and outputs a fault confidence value (within the range of [0,1], where 0 indicates certainty of no fault and 1 indicates certainty of a fault). This module is the source of the "multi-criteria" in "multi-criteria fusion", including electrical quantity criterion subunit, non-electrical quantity criterion subunit and communication criterion subunit.
[0040] Furthermore, the electrical quantity criterion subunit generates the following criteria: (1) Overcurrent criterion (C_OC): Calculate the effective values of the three-phase current and zero-sequence current and compare them with the adaptive setting value; the fault confidence is calculated using a soft threshold function: when the current I ≤ I_set_low, C_OC=0; when I ≥ I_set_high, C_OC=1; when I_set_low < I < I_set_high, C_OC = (I - I_set_low) / (I_set_high- I_set_low); where I_set_low and I_set_high are adaptively adjusted according to the system operating mode and load level.
[0041] (2) Sudden change criterion (C_DF): Calculate the current sudden change ΔI = |I(t) - I(tT)| and the voltage sudden change ΔU = |U(t) - U(tT)| (T is one power frequency cycle); when the sudden change exceeds the set threshold, the high confidence level is output. The sudden change criterion is sensitive to the fault start time and has a fast response speed (results can be output within 1-2 cycles).
[0042] (3) Zero-sequence direction criterion (C_ZS): Calculate the phase difference between zero-sequence voltage 3U0 and zero-sequence current 3I0 to determine the direction of zero-sequence power; when the direction of zero-sequence power points to the line (positive direction), output high confidence; when it is in the opposite direction, output low confidence. This criterion can effectively distinguish between faults in this line and reverse faults.
[0043] (4) Negative sequence direction criterion (C_NS): Calculate the phase difference between negative sequence voltage 2U2 and negative sequence current 2I2 to determine the direction of negative sequence power; mainly used for direction discrimination of asymmetrical faults (single-phase grounding, two-phase short circuit).
[0044] (5) Undervoltage criterion (C_UV): Calculate the effective value of the three-phase voltage and output a high confidence level when the voltage is lower than the set value; adopt a soft threshold function similar to the overcurrent criterion.
[0045] (6) Impedance criterion (C_Z): Calculate the measured impedance Z = U / I and determine whether the measured impedance falls within the preset impedance action area; the impedance criterion can reflect the fault distance information and provide a basis for fault location.
[0046] (7) Harmonic Criterion (C_HM): Calculate the 2nd, 3rd and 5th harmonic content in the fault current and use the differences in harmonic characteristics of different fault types to assist in the fault type identification; for example, transformer inrush current contains a large number of 2nd harmonics, which can be used to distinguish between inrush current and fault current.
[0047] Furthermore, the non-electrical quantity criterion subunit generates the following criteria: (1) Temperature anomaly criterion (C_TEMP): Monitor the temperature of switchgear contacts, cable joints, etc. When the temperature exceeds the threshold or the temperature rise rate is abnormal, output a high confidence level; temperature anomalies may indicate poor contact or overload.
[0048] (2) Partial discharge criterion (C_PD): Monitor the intensity and frequency of partial discharge signals in the switch cabinet. When the partial discharge signal exceeds the threshold, output high confidence; abnormal partial discharge may indicate insulation deterioration.
[0049] (3) Switch mechanical characteristic criterion (C_MECH): Analyze the switch opening and closing stroke curves and vibration signals. When the mechanical characteristics deviate from the normal mode, output a medium confidence level (as an early warning rather than a fault criterion).
[0050] Furthermore, the communication criterion subunit generates the following criterion: (1) Neighborhood overcurrent criterion (C_NOC): Receive the overcurrent criterion result sent by the neighboring terminal. When the neighboring terminal reports an overcurrent but this terminal does not report an overcurrent, the confidence of the neighboring overcurrent of this terminal decreases (indicating that the fault may not be in this segment); when both the neighboring terminal and this terminal report an overcurrent, the fault may be in the segment between the two terminals.
[0051] (2) Neighborhood Direction Criterion (C_NDR): Receive the direction element results of neighboring terminals and infer the fault section by combining the direction information of multiple terminals; for example, when the direction elements of all terminals upstream of the fault point point downstream (positive direction), and the direction elements of all terminals downstream of the fault point point upstream (opposite direction), the fault section can be accurately located.
[0052] (3) Master station confirmation criterion (C_MS): Receives fault confirmation information sent by the master station and outputs high confidence when the master station confirms the fault; this criterion serves as a supplement to the communication criterion and provides a reference when the local criterion is uncertain.
[0053] The fusion decision engine module is the core of this invention. It is responsible for fusing the confidence levels of each criterion output by the multi-criteria generation module to output a comprehensive fault confidence level and decision result. The fusion decision engine module adopts Dempster-Shafer (DS) evidence theory as the fusion framework and combines it with an adaptive weight allocation mechanism. This module includes a basic probability allocation (BPA) construction subunit, an adaptive weight allocation subunit, an evidence fusion subunit, a conflict handling subunit, and a decision output subunit. DS evidence theory is a multi-source fusion theory for handling uncertain information. It can handle the "unknown" and "uncertain" situations and is more suitable than Bayesian methods for handling the uneven confidence levels of each criterion in distribution automation terminals.
[0054] Furthermore, the Basic Probability Allocation (BPA) construction subunit maps the confidence value of each criterion to the basic probability allocation function in DS evidence theory. For each criterion Ci, the identification framework Θ = {F, N} (F = fault, N = normal) is defined, and the BPA function mi is constructed: mi({F}) = Ci × wi, mi({N}) = (1-Ci) × wi, mi({F,N}) = 1 - wi, where wi is the weight of criterion i (takes values [0,1]), mi({F,N}) represents the uncertainty of "don't know", and the weight wi reflects the credibility of criterion i under the current working condition and is dynamically calculated by the adaptive weight allocation subunit.
[0055] Furthermore, the adaptive weight allocation subunit dynamically calculates the weights of each criterion based on the current operating conditions, and the weight allocation adopts the following strategy: (1) Weight adjustment based on operating conditions: Adjust the weight of each criterion according to the system operation mode (neutral grounding mode, distributed power output status, load level, etc.); for example, in a neutral point grounded system via an arc suppression coil, the weight of the zero-sequence criterion is reduced (because the arc suppression coil compensates for the zero-sequence current), while the weights of the sudden change criterion and the negative-sequence criterion are increased; during the period of high output of distributed photovoltaic power, the weight of the overcurrent criterion is reduced (because the current limiting characteristics of the photovoltaic inverter limit the amplitude of the fault current), while the weights of the sudden change criterion and the direction criterion are increased.
[0056] (2) Weight adjustment based on data quality: The weights are adjusted according to the quality indicators of the input data for each criterion (such as data integrity rate, data timeliness, sensor health status, etc.), and the weights of criteria with poor data quality are reduced.
[0057] (3) Weight adjustment based on historical performance: The correct trigger rate and false trigger rate of each criterion in historical faults are statistically analyzed. The weight of the criterion with a high correct rate is increased, and the weight of the criterion with a high false trigger rate is decreased. The weight adjustment is performed by exponential weighted moving average (EWMA) for smooth update.
[0058] Specifically, the adaptive weight calculation formula is: wi = α × wi_cond + β × wi_qual + γ × wi_hist, where wi_cond is the working condition weight component, wi_qual is the data quality weight component, wi_hist is the historical performance weight component, and α, β, and γ are weighting coefficients (α+β+γ=1, typical values are α=0.5, β=0.2, γ=0.3). After weight normalization, Σwi = 1 is satisfied.
[0059] Furthermore, the evidence fusion subunit employs Dempster's combination rule to fuse the BPA functions of multiple criteria. The combination rule for two criteria i and j is: mij(A) = Σ_{B∩C=A} mi(B) × mj(C) / (1- K), where K = Σ_{B∩C= mi(B) × mj(C) is the conflict coefficient, which reflects the degree of conflict between the two criteria. When the value of K is large (e.g., K > 0.7), it indicates that there is a serious conflict between the criteria. At this time, the credibility of the fusion result is reduced, triggering the conflict handling mechanism.
[0060] For the fusion of multiple criteria, a recursive approach is adopted for step-by-step fusion: first, criteria 1 and criteria 2 are fused to obtain m_12, then m_12 is fused with criteria 3 to obtain m_123, and so on. The fusion order is arranged in descending order of the criteria weights, prioritizing the fusion of high-weight criteria to reduce the impact of low-weight criteria on the fusion result.
[0061] Furthermore, the conflict handling subunit activates a conflict handling mechanism when the evidence fusion subunit detects that the inter-criteria conflict coefficient K exceeds a threshold. The conflict handling adopts the following strategy: (1) Conflict analysis: Identify the criterion pairs that cause conflict and analyze the causes of conflict (such as sensor failure, data delay, sudden change in operating conditions, etc.).
[0062] (2) Weight reduction processing: Reduce or remove the criteria that are identified as abnormal (such as data abnormality caused by sensor failure) and recalculate the fusion.
[0063] (3) Delayed confirmation: When the conflict cannot be resolved immediately, a delayed confirmation strategy is adopted. During the delay period (e.g., 2-3 cycles), the output of each criterion is continuously monitored, and a final decision is made after the conflict is eliminated or new information arrives.
[0064] (4) Conservative strategy: When the conflict continues to be unresolved and cannot be delayed, a conservative strategy is adopted (i.e., it tends to be judged as a fault, with safety as the priority), and an alarm is triggered and sent to the main station.
[0065] Furthermore, the decision output subunit outputs the final decision result based on the fused BPA function, defining the fault confidence CF = m({F}) / (m({F}) + m({N})). When CF ≥ CF_threshold_high (e.g., 0.8), it is determined to be a fault and the fault handling process is triggered; when CF ≤ CF_threshold_low (e.g., 0.2), it is determined to be normal; when CF_threshold_low < CF < CF_threshold_high, it is determined to be uncertain, and a delayed confirmation or requesting information from neighboring terminals to assist in the judgment is initiated. The decision output simultaneously includes the fault type discrimination result (inferring the fault type by combining the fault characteristics of various criteria) and the time series of the fault confidence (for trend analysis).
[0066] The distributed collaborative control module enables peer-to-peer collaboration among power distribution automation terminals, autonomously completing fault location, isolation, and recovery without the participation of a master station; this module is the key to the rapid fault handling of this invention; the distributed collaborative control module includes a neighborhood discovery and topology maintenance subunit, a distributed fault location subunit, a collaborative fault isolation subunit, an adaptive power supply recovery subunit, and a consistency protocol subunit.
[0067] Furthermore, the neighborhood discovery and topology maintenance subunit identifies neighboring terminals and establishes a neighborhood topology table by broadcasting discovery messages when a terminal starts up or when the network topology changes. The neighborhood topology table records the ID, communication address, connection direction (upstream / downstream / connection), segment information, etc. of each neighboring terminal. The topology information is updated periodically (e.g., every 5 minutes) and immediately when a switch change is detected.
[0068] Furthermore, the distributed fault location subunit, based on the criterion output of multiple terminals and neighborhood direction information, employs a distributed inference algorithm to locate the fault segment; the location logic is as follows: (1) Each terminal sends its own fusion fault confidence (CF) and direction criterion results to the neighboring terminals.
[0069] (2) After each terminal receives information from the adjacent terminal, it infers the faulty section according to the following rules: if the direction criterion of this terminal points downstream and the direction criterion of the downstream adjacent terminal points upstream (in the opposite direction), then the fault is in the section between this terminal and the downstream adjacent terminal.
[0070] (3) If the confidence level of the fault in this terminal is high but the direction criterion is uncertain, then a comprehensive inference is made by combining the neighborhood overcurrent criterion (overcurrent status of upstream and downstream terminals): if the upstream terminal is overcurrent while the downstream terminal is not overcurrent, then the fault is in the downstream section of this terminal; if both upstream and downstream terminals are overcurrent, then the fault is in the more downstream section; if neither upstream nor downstream terminals are overcurrent, then the fault is in the upstream section of this terminal or is a high-resistance fault.
[0071] (4) The positioning results are consistent across terminals: the positioning conclusions of all terminals should be consistent (i.e., only one segment is identified as a faulty segment). If there is any inconsistency, a negotiation mechanism is triggered to reach a consensus through multiple rounds of information exchange.
[0072] Furthermore, in the collaborative fault isolation subunit, after the fault segment is determined, the terminals at both ends of the fault segment collaboratively perform isolation operations; the isolation strategy is as follows: (1) The upstream terminal of the fault section trips the downstream switch, and the downstream terminal of the fault section trips the upstream switch, thus achieving double-end isolation of the fault section.
[0073] (2) The isolation operation adopts the "confirmation-execution" mode: Terminal A sends an isolation request to Terminal B. After Terminal B confirms, both parties execute the tripping at the same time to avoid non-selective disconnection caused by single-end isolation.
[0074] (3) After isolation is completed, the two terminals send isolation confirmation information to each other and notify the neighboring terminals and the main station.
[0075] (4) If a terminal fails to perform isolation operation (e.g., the switch refuses to operate), the adjacent terminal starts backup isolation logic to expand the isolation range to the next level switch.
[0076] Furthermore, the adaptive power supply recovery subunit, after fault isolation, restores power to the non-faulty area through automatic switching of the tie switch; the recovery strategy adopts a distributed optimization method. (1) Generation of recovery candidate schemes: Based on the current network topology and load distribution, generate multiple feasible recovery schemes (i.e., which non-faulty sections to restore power supply by closing which tie switches).
[0077] (2) Evaluation of restoration schemes: Each candidate scheme is evaluated, and the evaluation indicators include: the total load of power supply restored (maximum), restoration time (minimum), load rate of the line after restoration (not exceeding the safety limit), voltage quality after restoration (meeting voltage deviation requirements), and number of switching operations during restoration (minimum).
[0078] (3) Recovery scheme selection: The optimal scheme is selected by weighted scoring method. The scoring function is: Score = w1 × P_restore - w2 × T_restore - w3 × N_switch - w4 × V_deviation, where P_restore is the amount of load to be restored, T_restore is the estimated recovery time, N_switch is the number of switching actions, V_deviation is the voltage deviation, and w1-w4 are weighting coefficients (adjusted according to load priority and reliability requirements).
[0079] (4) Execution of the recovery plan: The selected recovery plan is executed in a distributed manner among the terminals. The terminal at the tie switch is responsible for performing the closing operation. Before closing, a "no voltage check - synchronization check" confirmation is performed to ensure the safety of closing. During the recovery process, the load and voltage are monitored in real time. If an overload or voltage exceeds the limit, the recovery strategy is automatically adjusted (such as disconnecting some low-priority loads).
[0080] Furthermore, the consensus protocol subunit, during the distributed collaboration process, employs a consensus protocol to ensure consistency in fault location and isolation decisions among all terminals; when disagreements arise among terminals, consensus is reached through multiple rounds of information exchange and voting mechanisms; the consensus protocol is based on the following principles: terminals with higher fault confidence have greater say in decision-making; when a majority of terminals reach a consensus, a minority of terminals accept the majority opinion; when a consensus cannot be reached, a conservative strategy is adopted (preferring to implement isolation to ensure security).
[0081] The execution and communication module includes a switch control execution subunit, an action recording subunit, and a communication subunit.
[0082] Furthermore, the switch control execution subunit receives commands from the fusion decision engine or the distributed collaborative control module to drive the switch to perform opening / closing operations; it has anti-pumping circuit, control circuit disconnection monitoring and operation circuit self-holding functions; the switch operation response time does not exceed 50ms.
[0083] Furthermore, the action recording subunit records all action information of the terminal, including fault waveforms, criterion outputs, fusion decision-making processes, collaborative communication processes, and switch operation records, which facilitates post-event analysis.
[0084] Furthermore, the communication subunit supports multiple communication methods, including: P2P communication with adjacent terminals (based on IEC 61850 GOOSE messages or custom lightweight protocols, via 5G or local wireless mesh networks, with a transmission latency of no more than 20ms); communication with the distribution automation master station (based on IEC 60870-5-104 or IEC 61850 MMS messages, via 5G / 4G / fiber optic networks); the communication has message authentication and encryption functions to ensure communication security.
[0085] This invention presents a multi-criteria fusion framework based on DS evidence theory, applying DS evidence theory to fault detection in distribution automation terminals for the first time. It unifies electrical quantity criteria, non-electrical quantity criteria, and communication criteria into a unified evidence fusion framework. Compared to existing technologies where each criterion operates independently and is unrelated, this invention fully utilizes the complementarity of multi-source information through evidence fusion. Even when a single criterion fails, the overall detection reliability can be maintained by supplementing it with other criteria. Furthermore, the explicit modeling of "unknown" and "uncertain" conditions by DS evidence theory makes the fusion results more reasonable.
[0086] Simultaneously, a three-dimensional adaptive weight allocation mechanism based on operating conditions, data quality, and historical performance is employed. This invention's adaptive weight allocation not only considers system operating conditions (such as neutral point grounding method and distributed power supply status) but also comprehensively considers data quality (sensor health status, data integrity rate) and historical performance (criteria accuracy and false alarm rate), achieving multi-dimensional dynamic adjustment of weights. Compared to existing technologies that use fixed weights or adjust weights based on only a single factor, this invention's weight allocation is more refined and reliable.
[0087] This invention introduces a fault-tolerance mechanism based on evidence conflict detection. In the DS evidence fusion process, real-time monitoring of the conflict coefficient K is implemented. When serious conflicts occur between criteria, conflict analysis and weight reduction are automatically initiated, preventing conflicting criteria from contaminating the fusion results. Simultaneously, delayed confirmation and conservative strategies are provided as fallback solutions to ensure that reasonable decisions can still be made when conflicts cannot be resolved. This represents a significant improvement over the lack of conflict handling mechanisms in existing technologies.
[0088] Based on a distributed fault location method using multi-terminal directional information fusion, this invention exchanges directional criteria results through P2P communication between terminals and utilizes the symmetrical characteristics of directional information between upstream and downstream terminals at the fault point (upstream points to downstream, downstream points to upstream) to accurately locate the faulty section. Compared with existing "reclosing-test" methods or centralized analysis methods at the master station, the location method of this invention does not require multiple test operations, reduces the location time from 30 seconds to 3 minutes to 1-2 seconds, and the location accuracy does not depend on master station communication.
[0089] The distributed adaptive power restoration strategy based on weighted scoring comprehensively considers multiple optimization objectives such as restoration load, restoration time, load rate, voltage quality, and number of switching actions. It uses a weighted scoring method to distribute the optimal solution among terminals. Compared with existing fixed restoration schemes, this invention can adaptively adjust the restoration strategy according to real-time operating status, maximizing the restoration effect. Real-time monitoring and adaptive adjustment during the restoration process further ensure the safety of the restoration.
[0090] This invention integrates electrical and non-electrical quantity criteria for the first time, incorporating non-electrical quantity criteria such as switchgear temperature, partial discharge, and mechanical characteristics into the fault detection fusion framework of power distribution automation terminals. Non-electrical quantity criteria can capture faults with inconspicuous electrical quantity characteristics (such as slow insulation degradation and switch mechanical failure), complementing electrical quantity criteria and significantly improving the comprehensiveness and early warning capability of fault detection.
[0091] Finally, it should be noted that the descriptions of each embodiment in the above embodiments have different focuses. Parts not described in detail in a certain embodiment can be referred to in the relevant descriptions of other embodiments. The above descriptions are merely preferred embodiments of this application and explanations of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to the technical solutions formed by specific combinations of the above technical features, but should also cover other technical solutions formed by arbitrary combinations of the above technical features or their equivalent features without departing from the inventive concept.
Claims
1. A power distribution automation terminal control method based on multi-criteria fusion, characterized in that, The method includes the following steps: S1: Collect line electrical quantities, non-electrical quantities, and neighborhood communication data, and perform preprocessing; S2: Generates multiple fault criteria from preprocessed data, including electrical quantity criteria, non-electrical quantity criteria, and communication criteria, with each criterion outputting an independent fault confidence value; S3: Dynamically calculate the fusion weight of each criterion based on the current operating conditions, data quality, and historical performance; S4: Map the confidence values of each criterion to the basic probability allocation function, use the Dempster combination rule to recursively fuse them, and monitor the conflict coefficient between criteria in real time. S5: Calculate the comprehensive fault confidence based on the fusion results, and output the fault determination and fault type discrimination results; S6: After a fault is determined, exchange direction criterion results with adjacent terminals and determine the faulty section through distributed reasoning; S7: Terminals at both ends of the faulty section coordinate to perform isolation operations; S8: Generate and evaluate multiple power restoration plans, select the optimal plan, and execute them in a distributed manner; S9: Perform a post-evaluation of the fault handling process and update the criterion weights and strategy parameters.
2. The distribution automation terminal control method based on multi-criteria fusion according to claim 1, characterized in that, In step S2, the electrical quantity criteria include overcurrent criteria, sudden change criteria, zero-sequence direction criteria, negative-sequence direction criteria, undervoltage criteria, impedance criteria, and harmonic criteria; the non-electrical quantity criteria include temperature anomaly criteria, partial discharge criteria, and switch mechanical characteristic criteria; the communication criteria include neighborhood overcurrent criteria, neighborhood direction criteria, and master station confirmation criteria; and each criterion uses a soft threshold function to calculate the fault confidence value.
3. The distribution automation terminal control method based on multi-criteria fusion according to claim 1, characterized in that, In step S3, the adaptive weight calculation formula is: wi = α×wi_cond + β×wi_qual + γ×wi_hist, where wi_cond is the operating condition weight component, determined based on the neutral grounding method, distributed power supply output status, and load level; wi_qual is the data quality weight component, determined based on data integrity rate, timeliness, and sensor health status; wi_hist is the historical performance weight component, determined based on the correct trigger rate and false trigger rate of the criterion; α, β, and γ are weighting coefficients, and α+β+γ=1; After weight normalization, Σwi = 1 is satisfied.
4. The distribution automation terminal control method based on multi-criteria fusion according to claim 1, characterized in that, In step S6, the reasoning rule for distributed fault location is as follows: if the direction criterion of this terminal points downstream and the direction criterion of the downstream adjacent terminal points upstream, then the fault is in the segment between this terminal and the downstream adjacent terminal; if the direction criterion is uncertain, a comprehensive inference is made by combining the overcurrent status of the neighborhood; the location result is consistent across all terminals, and if there is inconsistency, a consensus is reached through multiple rounds of information exchange.
5. The distribution automation terminal control method based on multi-criteria fusion according to claim 1, characterized in that, In step S8, the evaluation indicators for the power restoration plan include the restored load, restoration time, post-restoration load rate, voltage deviation, and number of switching actions; the plan scoring function is: Score = w1 × P_restore - w2 × T_restore - w3 × N_switch - w4 × V_deviation; Before closing the tie switch, a check for no voltage and a check for synchronization are performed. During the restoration process, real-time monitoring and adaptive adjustment are performed.
6. A power distribution automation terminal control system based on multi-criteria fusion, characterized in that, The system is deployed at the switching points of the power distribution network (including pole-mounted switches, ring main units, and substations), and includes five main functional modules: a multi-source data acquisition module, a multi-criteria generation module, a fusion decision engine module, a distributed collaborative control module, and an execution and communication module. The system adopts an edge computing architecture, with each terminal having local intelligent decision-making capabilities, while achieving distributed collaboration between terminals through peer-to-peer communication.
7. The power distribution automation terminal control system based on multi-criteria fusion according to claim 6, characterized in that, The multi-source data acquisition module is the data foundation for fusion decision-making. It is responsible for acquiring three types of information sources: electrical quantities, non-electrical quantities, and communication data. The module includes an electrical quantity acquisition subunit, a non-electrical quantity acquisition subunit, a neighborhood communication data receiving subunit, and a data preprocessing subunit.
8. The power distribution automation terminal control system based on multi-criteria fusion according to claim 6, characterized in that, The multi-criteria generation module is the source of "multi-criteria" in "multi-criteria fusion", including electrical quantity criteria subunit, non-electrical quantity criteria subunit and communication criteria subunit.
9. The power distribution automation terminal control system based on multi-criteria fusion according to claim 6, characterized in that, The fusion decision engine module adopts Dempster-Shafer (DS) evidence theory as the fusion framework and combines it with an adaptive weight allocation mechanism. The module includes a basic probability allocation (BPA) construction subunit, an adaptive weight allocation subunit, an evidence fusion subunit, a conflict handling subunit, and a decision output subunit.
10. The power distribution automation terminal control system based on multi-criteria fusion according to claim 6, characterized in that, The distributed collaborative control module includes a neighborhood discovery and topology maintenance subunit, a distributed fault location subunit, a collaborative fault isolation subunit, an adaptive power supply recovery subunit, and a consistency protocol subunit.
11. The power distribution automation terminal control system based on multi-criteria fusion according to claim 6, characterized in that, The execution and communication module includes a switch control execution subunit, an action recording subunit, and a communication subunit.
